diff --git a/.gitignore b/.gitignore index 8b61105..45634bd 100644 --- a/.gitignore +++ b/.gitignore @@ -190,3 +190,7 @@ nohup.out scraper_cache/ challenge/ + +# WandB local sync data. +wandb/ + diff --git a/community-contributions/Keshvi_Web2Quiz/Web2Quiz.ipynb b/community-contributions/Keshvi_Web2Quiz/Web2Quiz.ipynb new file mode 100644 index 0000000..1609d81 --- /dev/null +++ b/community-contributions/Keshvi_Web2Quiz/Web2Quiz.ipynb @@ -0,0 +1,300 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "source": [ + "## Web2Quiz: Generator Quiz from webpage content." + ], + "metadata": { + "id": "n3vd295elWxh" + }, + "id": "n3vd295elWxh" + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": { + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988" + }, + "outputs": [], + "source": [ + "#.env upload\n", + "from google.colab import files\n", + "uploaded = files.upload()" + ] + }, + { + "cell_type": "code", + "source": [ + "!pip install dotenv\n" + ], + "metadata": { + "id": "VTpN_jVbMKuk" + }, + "id": "VTpN_jVbMKuk", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "import os\n", + "from dotenv import load_dotenv" + ], + "metadata": { + "id": "twYi9eJwL2h1" + }, + "id": "twYi9eJwL2h1", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENROUTER_KEY')\n", + "\n", + "# Check the key\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "# elif not api_key.startswith(\"sk-proj-\"):\n", + "# print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ], + "metadata": { + "id": "NRnUTEkZL2eZ" + }, + "id": "NRnUTEkZL2eZ", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "!pip install openai" + ], + "metadata": { + "id": "RRuKJ_pzL2be" + }, + "id": "RRuKJ_pzL2be", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "!pip install requests beautifulsoup4\n", + "!pip install selenium" + ], + "metadata": { + "id": "DWsPpdjOVPTW" + }, + "id": "DWsPpdjOVPTW", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "from bs4 import BeautifulSoup\n", + "import requests\n", + "from tempfile import mkdtemp\n", + "from selenium import webdriver\n", + "from selenium.webdriver.chrome.options import Options\n", + "from selenium.webdriver.support.ui import WebDriverWait\n", + "from selenium.webdriver.support import expected_conditions as EC\n", + "from selenium.webdriver.common.by import By\n", + "\n", + "class Website:\n", + " def __init__(self, url, use_selenium=False):\n", + " \"\"\"\n", + " Create Website object from the given URL.\n", + " If use_selenium=True, fetch page with Selenium.\n", + " Otherwise, use requests + BeautifulSoup.\n", + " \"\"\"\n", + " self.url = url\n", + " self.title = \"\"\n", + " self.text = \"\"\n", + " self.use_selenium = use_selenium\n", + "\n", + " if self.use_selenium:\n", + " html = self._fetch_with_selenium()\n", + " else:\n", + " html = self._fetch_with_requests()\n", + "\n", + " if not html:\n", + " self.title = \"Error fetching page\"\n", + " self.text = \"Could not retrieve HTML content.\"\n", + " return\n", + "\n", + " soup = BeautifulSoup(html, \"html.parser\")\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + "\n", + " # content_div = soup.find('div', id='content')\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\", \"header\", \"footer\", \"nav\", \"aside\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"No body tag found in the HTML.\"\n", + "\n", + " # Basic html scrapper\n", + " def _fetch_with_requests(self):\n", + " \"\"\"Fetch HTML using requests.\"\"\"\n", + " try:\n", + " headers = {\"User-Agent\": \"Mozilla/5.0\"}\n", + " response = requests.get(self.url, headers=headers, timeout=10)\n", + " response.raise_for_status()\n", + " return response.text\n", + " except requests.exceptions.RequestException as e:\n", + " print(f\"Error fetching with requests: {e}\")\n", + " return None\n", + "\n", + " # Dynamic html scrapper\n", + " def _fetch_with_selenium(self):\n", + " \"\"\"Fetch HTML using Selenium with improved options.\"\"\"\n", + " options = Options()\n", + " options.add_argument(\"--no-sandbox\")\n", + " options.add_argument(\"--disable-dev-shm-usage\")\n", + " options.add_argument(\"--headless\")\n", + " options.add_argument(f\"--user-data-dir={mkdtemp()}\")\n", + "\n", + " driver = None\n", + " try:\n", + " driver = webdriver.Chrome(options=options)\n", + " driver.get(self.url)\n", + "\n", + " WebDriverWait(driver, 10).until(\n", + " EC.presence_of_element_located((By.TAG_NAME, \"body\"))\n", + " )\n", + "\n", + " html = driver.page_source\n", + " return html\n", + " except Exception as e:\n", + " print(f\"An error occurred during Selenium fetch: {e}\")\n", + " return None\n", + " finally:\n", + " if driver:\n", + " driver.quit()\n", + "\n" + ], + "metadata": { + "id": "PzBP0tXXcrP-" + }, + "id": "PzBP0tXXcrP-", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "site1 = Website(\"https://en.wikipedia.org/wiki/Integration_testing\", use_selenium=False)\n", + "print(\"Title:\", site1.title)\n", + "print(\"Text preview:\", site1.text[:200])\n", + "\n", + "site2 = Website(\"https://www.tpointtech.com/java-for-loop\", use_selenium=True)\n", + "print(\"Title:\", site2.title)\n", + "print(\"Text preview:\", site2.text[:200])" + ], + "metadata": { + "id": "vsNmh5b5c6Gq" + }, + "id": "vsNmh5b5c6Gq", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Step 1: Create your prompts\n", + "system_prompt = f\"You are a MCQ quiz generator. Analyze the provided TEXT and filter CONTENT relevent to {site1.title}. Then based on the relevant CONTENT generate 10 MCQs. List all correct options at the end.\"\n", + "user_prompt = f\"Below is provided TEXT : \\n{site1.text}\"\n", + "\n", + "# Step 2: Make the messages list\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "]\n", + "\n", + "# Step 3: Call OpenAI\n", + "openai = OpenAI(base_url=\"https://openrouter.ai/api/v1\", api_key=api_key)\n", + "\n", + "# Step 4: print the result\n", + "response = openai.chat.completions.create(model=\"qwen/qwen2.5-vl-72b-instruct:free\", messages=messages)\n", + "print(response.choices[0].message.content)" + ], + "metadata": { + "collapsed": true, + "id": "BYdc1w70QFD2" + }, + "id": "BYdc1w70QFD2", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Step 1: Create your prompts\n", + "system_prompt = f\"You are a MCQ quiz generator. Analyze the provided TEXT and filter CONTENT relevent to {site2.title}. Then based on the relevant CONTENT generate 10 MCQs. List all correct options at the end.\"\n", + "user_prompt = f\"Below is provided TEXT : \\n{site2.text}\"\n", + "\n", + "# Step 2: Make the messages list\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "]\n", + "\n", + "# Step 3: Call OpenAI\n", + "openai = OpenAI(base_url=\"https://openrouter.ai/api/v1\", api_key=api_key)\n", + "\n", + "# Step 4: print the result\n", + "response = openai.chat.completions.create(model=\"qwen/qwen2.5-vl-72b-instruct:free\", messages=messages)\n", + "print(response.choices[0].message.content)" + ], + "metadata": { + "id": "Rv8vxFHtQFBm" + }, + "id": "Rv8vxFHtQFBm", + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [], + "metadata": { + "id": "o5tIkQ95_2Hc" + }, + "id": "o5tIkQ95_2Hc", + "execution_count": null, + "outputs": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.12" + }, + "colab": { + "provenance": [] + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} \ No newline at end of file diff --git a/community-contributions/LLaVA-For-Visually-Impared-People/llava-For-Image-week1.ipynb b/community-contributions/LLaVA-For-Visually-Impared-People/llava-For-Image-week1.ipynb new file mode 100644 index 0000000..d1494d8 --- /dev/null +++ b/community-contributions/LLaVA-For-Visually-Impared-People/llava-For-Image-week1.ipynb @@ -0,0 +1,230 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2a5df086", + "metadata": {}, + "source": [ + "# If Anyone is interested in this idea and want to contribute please let me know and contribute your idea/Code\n" + ] + }, + { + "cell_type": "markdown", + "id": "3b0d5f6e", + "metadata": {}, + "source": [ + "*IDEA* - For visually impaired individuals, daily life often presents numerous obstacles that many of us take for granted. While tools like Braille and guide dogs offer some support, they do not fully address the limitations faced in navigating the world. With over 43.3 million blind people globally, there is a pressing need for more inclusive technologies that help break these barriers. This project aims to do more than assist with daily tasks; it seeks to empower individuals to engage meaningfully with their environment. By providing real-time, contextually accurate captions, this system allows them to experience the world around them, feel less isolated, and regain a sense of autonomy. Beyond just aiding navigation, it provides a bridge to connection—helping them feel more alive, present, and capable. This project is not just about overcoming limitations; it’s about enriching lives and enabling a deeper, fuller interaction with the world, fostering a sense of belonging and independence.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f97c7598-f571-4ea1-838c-e9158f729c3e", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "import base64\n", + "import os" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9fc1393c-f0b1-4982-94a2-bfd502e85b23", + "metadata": {}, + "outputs": [], + "source": [ + "def encode_image(image_path):\n", + " with open(image_path, 'rb') as f:\n", + " return base64.b64encode(f.read()).decode('utf-8')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "53cca1fa-6db2-4fe4-8990-ffd98423964a", + "metadata": {}, + "outputs": [], + "source": [ + "# image_path = r\"C:\\Users\\LAKSHYA\\OneDrive\\Pictures\\Camera Roll\\WIN_20250614_02_46_47_Pro.jpg\"\n", + "# image_base64 = encode_image(image_path)\n", + "# print(image_base64[:100]) " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "71146ccf-25af-48d3-8068-ee3c9008cebf", + "metadata": {}, + "outputs": [], + "source": [ + "image_list = []" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f8801a8-0c30-4199-a334-587096e6edeb", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee3c5d82-e530-40f5-901a-681421f21d1e", + "metadata": {}, + "outputs": [], + "source": [ + "def put_image():\n", + " global image_list\n", + " user_input_image = input(\"Enter image path or press enter to skip: \").strip()\n", + " \n", + " if not user_input_image:\n", + " print(\"No image inserted\")\n", + " return image_list\n", + "\n", + " image_path = os.path.normpath(user_input_image)\n", + " \n", + " if not os.path.exists(image_path):\n", + " print(\"Image path not found! Try again or enter to leave blank\")\n", + " return put_image() # Continue to allow more inputs\n", + " \n", + "\n", + "\n", + "\n", + " \n", + " image_base64 = encode_image(image_path)\n", + " image_list.append(image_base64)\n", + " \n", + " # Detect file extension for MIME type\n", + " # ext = os.path.splitext(image_path)[-1].lower()\n", + " # mime_type = 'image/jpeg' if ext in ['.jpg', '.jpeg'] else 'image/png' # Extend if needed\n", + "\n", + "\n", + " return image_list\n", + " \n", + " # return f\"data:{mime_type};base64,{image_base64[:100]}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "032f1abb-ca6c-4f03-bda1-1a0a62f2ec43", + "metadata": {}, + "outputs": [], + "source": [ + "prompt= (\"System prompt: (You are a compassionate and intelligent visual assistant designed to help people who are blind or visually impaired. \"\n", + " \"Your job is to look at an image and describe it in a way that helps the user understand the scene clearly. \"\n", + " \"Use simple, descriptive language and avoid technical terms. Describe what is happening in the image, people's body language, clothing, facial expressions, objects, and surroundings. \"\n", + " \"Be vivid and precise, as if you are painting a picture with words. \"\n", + " \"Also, take into account any personal instructions or questions provided by the user—such as describing a specific person, activity, or object. \"\n", + " \"If the user includes a specific prompt, prioritize that in your description.)\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29494db0-4770-4689-9904-8eebc4390e7c", + "metadata": {}, + "outputs": [], + "source": [ + "def put_prompt():\n", + " global prompt\n", + " user_input = input(\"Put new prompt: \")\n", + " if not user_input:\n", + " print(\"please enter a prompt\")\n", + " return put_prompt()\n", + " prompt += \"\\nUser: \" + user_input\n", + " return prompt\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d286369c-e6ef-4a20-a3a8-3563af28940a", + "metadata": {}, + "outputs": [], + "source": [ + "def image_description():\n", + " global prompt\n", + "\n", + " put_image()\n", + " if not image_list: \n", + " return \"No images available. Skipping...\"\n", + "\n", + " user_prompt = put_prompt()\n", + " full_answer = \"\"\n", + "\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=user_prompt,\n", + " images=image_list,\n", + " stream=True\n", + " ):\n", + " content = chunk.get(\"response\", \"\")\n", + " print(\"\\n\\n Final Answer:\",content, end=\"\", flush=True) # Live stream to console\n", + " full_answer += content\n", + "\n", + " prompt += \"\\nUser: \" + user_prompt + \"\\nAssistant: \" + full_answer\n", + " return full_answer\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbda35a3-45ed-4509-ab41-6827eacd922c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llava():\n", + " image_list.clear()\n", + " for i in range(5):\n", + " print(f\"\\n Iteration {i+1}\")\n", + " answer = image_description()\n", + " print(\"\\n\\n Final Answer:\", answer)\n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15518865-6c59-4029-bc2d-42d313eb78bc", + "metadata": {}, + "outputs": [], + "source": [ + "call_llava()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c77bd493-f893-402e-b4e3-64854e9d2e19", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/LLaVA-For-Visually-Impared-People/llava-week2-ChainForRealTimeCaptionGeneration.ipynb b/community-contributions/LLaVA-For-Visually-Impared-People/llava-week2-ChainForRealTimeCaptionGeneration.ipynb new file mode 100644 index 0000000..26e30e3 --- /dev/null +++ b/community-contributions/LLaVA-For-Visually-Impared-People/llava-week2-ChainForRealTimeCaptionGeneration.ipynb @@ -0,0 +1,433 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "f97c7598-f571-4ea1-838c-e9158f729c3e", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "import base64\n", + "import os" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9fc1393c-f0b1-4982-94a2-bfd502e85b23", + "metadata": {}, + "outputs": [], + "source": [ + "def encode_image(image_path):\n", + " with open(image_path, 'rb') as f:\n", + " return base64.b64encode(f.read()).decode('utf-8')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "53cca1fa-6db2-4fe4-8990-ffd98423964a", + "metadata": {}, + "outputs": [], + "source": [ + "# image_path = r\"C:\\Users\\LAKSHYA\\OneDrive\\Pictures\\Camera Roll\\WIN_20250614_02_46_47_Pro.jpg\"\n", + "# image_base64 = encode_image(image_path)\n", + "# print(image_base64[:100]) " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "71146ccf-25af-48d3-8068-ee3c9008cebf", + "metadata": {}, + "outputs": [], + "source": [ + "image_list = []" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f8801a8-0c30-4199-a334-587096e6edeb", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee3c5d82-e530-40f5-901a-681421f21d1e", + "metadata": {}, + "outputs": [], + "source": [ + "def put_image():\n", + " global image_list\n", + " user_input_image = input(\"Enter image path or press enter to skip: \").strip()\n", + " \n", + " if not user_input_image:\n", + " print(\"No image inserted\")\n", + " return image_list\n", + "\n", + " image_path = os.path.normpath(user_input_image)\n", + " \n", + " if not os.path.exists(image_path):\n", + " print(\"Image path not found! Try again or enter to leave blank\")\n", + " return put_image() # Continue to allow more inputs\n", + " \n", + "\n", + "\n", + "\n", + " \n", + " image_base64 = encode_image(image_path)\n", + " image_list.append(image_base64)\n", + " \n", + " # Detect file extension for MIME type\n", + " # ext = os.path.splitext(image_path)[-1].lower()\n", + " # mime_type = 'image/jpeg' if ext in ['.jpg', '.jpeg'] else 'image/png' # Extend if needed\n", + "\n", + "\n", + " return image_list\n", + " \n", + " # return f\"data:{mime_type};base64,{image_base64[:100]}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "032f1abb-ca6c-4f03-bda1-1a0a62f2ec43", + "metadata": {}, + "outputs": [], + "source": [ + "prompt= (\"System prompt: (You are a compassionate and intelligent visual assistant designed to help people who are blind or visually impaired. \"\n", + " \"Your job is to look at an image and describe it in a way that helps the user understand the scene clearly. \"\n", + " \"Use simple, descriptive language and avoid technical terms. Describe what is happening in the image, people's body language, clothing, facial expressions, objects, and surroundings. \"\n", + " \"Be vivid and precise, as if you are painting a picture with words. \"\n", + " \"Also, take into account any personal instructions or questions provided by the user—such as describing a specific person, activity, or object. \"\n", + " \"If the user includes a specific prompt, prioritize that in your description.)\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29494db0-4770-4689-9904-8eebc4390e7c", + "metadata": {}, + "outputs": [], + "source": [ + "def put_prompt():\n", + " global prompt\n", + " user_input = input(\"Put new prompt: \")\n", + " if not user_input:\n", + " print(\"please enter a prompt\")\n", + " return put_prompt()\n", + " prompt += \"\\nUser: \" + user_input\n", + " return prompt\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d286369c-e6ef-4a20-a3a8-3563af28940a", + "metadata": {}, + "outputs": [], + "source": [ + "def image_description():\n", + " global prompt\n", + "\n", + " put_image()\n", + " if not image_list: \n", + " return \"No images available. Skipping...\"\n", + "\n", + " user_prompt = put_prompt()\n", + " full_answer = \"\"\n", + "\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=user_prompt,\n", + " images=image_list,\n", + " stream=True\n", + " ):\n", + " content = chunk.get(\"response\", \"\")\n", + " print(\"\\n\\n Final Answer:\",content, end=\"\", flush=True) # Live stream to console\n", + " full_answer += content\n", + "\n", + " prompt += \"\\nUser: \" + user_prompt + \"\\nAssistant: \" + full_answer\n", + " return full_answer\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbda35a3-45ed-4509-ab41-6827eacd922c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llava():\n", + " image_list.clear()\n", + " for i in range(5):\n", + " print(f\"\\n Iteration {i+1}\")\n", + " answer = image_description()\n", + " print(\"\\n\\n Final Answer:\", answer)\n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15518865-6c59-4029-bc2d-42d313eb78bc", + "metadata": {}, + "outputs": [], + "source": [ + "call_llava()" + ] + }, + { + "cell_type": "markdown", + "id": "23de3b59-3699-4270-9392-99fccdede83e", + "metadata": {}, + "source": [ + "# second week practice on personal project making model faster and smarter by using tools\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9d44c59e-5eb7-4b00-9489-e05d7c8c3eda", + "metadata": {}, + "outputs": [], + "source": [ + "messages = []\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "061ea026-d4c6-4d6c-bb9b-f6430de9f5af", + "metadata": {}, + "outputs": [], + "source": [ + "system_content = (\n", + " \"You are a helpful assistant for visually impaired users. \"\n", + " \"You are capable of answering questions directly or calling a function to analyze an image if needed. \"\n", + " \"There is a list of images available, indexed from 0. \"\n", + " \"When a user asks a question, first determine whether any image in the list is needed to answer. \"\n", + " \"If yes, reply in this structured format:\\n\\n\"\n", + " \"TOOL_CALL: analyze_image(, prompt='')\\n\\n\"\n", + " \"If image is not needed, just answer the user directly in plain natural language.\\n\"\n", + " \"Be clear and use descriptive but accessible language suitable for blind users.\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2f859450-eb3e-4e6c-9602-84f91f5ffda7", + "metadata": {}, + "outputs": [], + "source": [ + "messages.append({\"role\":\"system\",\"content\":system_content})" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8009b75-3468-4694-887d-6cd5132c2907", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_loop():\n", + " \"\"\"Main chat interaction loop (single-turn version)\"\"\"\n", + " global image_list, messages\n", + " \n", + " print(\"\\n\" + \"=\"*50)\n", + " print(\"LLaVA Assistant for Visually Impaired Users\")\n", + " print(\"=\"*50 + \"\\n\")\n", + " \n", + " # Step 1: Load images\n", + " print(\"Step 1: Add images (optional)\")\n", + " put_image()\n", + " messages.append({\n", + " \"role\": \"system\", \n", + " \"content\": f\"There are {len(image_list)} images available (index 0-{len(image_list)-1}).\"\n", + " })\n", + " \n", + " # Step 2: Single chat interaction\n", + " print(\"\\nStep 2: Ask a question about the images\")\n", + " user_content = put_prompt()\n", + " messages.append({\"role\": \"user\", \"content\": user_content})\n", + " \n", + " # Get model response\n", + " try:\n", + " response = ollama.chat(\n", + " model='llava:7b-v1.6',\n", + " messages=messages\n", + " )[\"message\"][\"content\"]\n", + " print(\"assistant: \",response) \n", + " processed_response = process_response(response)\n", + " print(f\"\\nASSISTANT: {processed_response}\\n\")\n", + " \n", + " except Exception as e:\n", + " print(f\"Error occurred: {e}\")\n", + " \n", + " print(\"\\nSession ended. Goodbye!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a3b3ff73-3cd5-4e5a-a37e-aaa8b325613c", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee2de6d7-a0bf-45fc-8d5c-98e0055519b0", + "metadata": {}, + "outputs": [], + "source": [ + "def process_response(response):\n", + " \"\"\"Process the model's response and handle tool calls\"\"\"\n", + " if response.strip().startswith(\"TOOL_CALL:\"):\n", + " # Extract image index/range and prompt from TOOL_CALL\n", + " pattern = r\"TOOL_CALL:\\s*analyze_image\\((.*?)\\s*,\\s*prompt='(.*?)'\\)\"\n", + " match = re.search(pattern, response, re.DOTALL)\n", + " \n", + " if not match:\n", + " error_msg = \"Error: Invalid TOOL_CALL format.\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " \n", + " image_expr = match.group(1).strip()\n", + " prompt = match.group(2).strip()\n", + " \n", + " try:\n", + " # Handle different index formats\n", + " if \":\" in image_expr: # Range (e.g., \"1:3\")\n", + " start, end = map(int, image_expr.split(\":\"))\n", + " index_or_range = list(range(start, end))\n", + " else: # Single index\n", + " index_or_range = int(image_expr)\n", + " \n", + " # Validate indices\n", + " max_index = len(image_list) - 1\n", + " if isinstance(index_or_range, list):\n", + " if any(i < 0 or i > max_index for i in index_or_range):\n", + " error_msg = f\"Error: Image index out of range (0-{max_index}).\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " elif index_or_range < 0 or index_or_range > max_index:\n", + " error_msg = f\"Error: Image index out of range (0-{max_index}).\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " \n", + " # Perform analysis\n", + " result = analyze_image(index_or_range, prompt)\n", + " print(\"funtion called\")\n", + " messages.append({\n", + " \"role\": \"function\",\n", + " \"name\": \"analyze_image\",\n", + " \"content\": result\n", + " })\n", + " \n", + " # Return formatted result\n", + " formatted_result = f\"\\nIMAGE ANALYSIS RESULT:\\n{result}\"\n", + " return formatted_result\n", + "\n", + " except Exception as e:\n", + " error_msg = f\"Error processing TOOL_CALL: {e}\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " else:\n", + " messages.append({\"role\": \"assistant\", \"content\": response})\n", + " return response" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ea82f8f6-c321-4fbc-81ee-a508b087d53b", + "metadata": {}, + "outputs": [], + "source": [ + "def analyze_image(index_or_range, prompt):\n", + " \"\"\"Analyze specific image(s) using LLaVA\"\"\"\n", + " global image_list\n", + " \n", + " # Handle single index or range\n", + " if isinstance(index_or_range, int):\n", + " images = [image_list[index_or_range]]\n", + " elif isinstance(index_or_range, list):\n", + " images = [image_list[i] for i in index_or_range]\n", + " else:\n", + " return \"Invalid image index/range specified.\"\n", + " \n", + " if not images:\n", + " return \"No images available for analysis.\"\n", + " \n", + " full_prompt = (\n", + " \"Describe the image clearly for a visually impaired user. \"\n", + " \"Be detailed about objects, people, colors, spatial relationships, \"\n", + " \"and any important context. \"\n", + " f\"User's specific request: {prompt}\"\n", + " )\n", + " \n", + " output = \"\"\n", + " try:\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=full_prompt,\n", + " images=images,\n", + " stream=True\n", + " ):\n", + " output += chunk.get('response', \"\")\n", + " except Exception as e:\n", + " return f\"Error analyzing image: {e}\"\n", + " \n", + " return output\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2040b020-8944-409b-8ebb-10d7ffef1748", + "metadata": {}, + "outputs": [], + "source": [ + "image_list.clear\n", + "for i in range(5):\n", + " chat_loop()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c7c40d7-df9d-464a-89da-1c6fe613c31d", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/Market_Research_Agent.ipynb b/community-contributions/Market_Research_Agent.ipynb new file mode 100644 index 0000000..52dfdf4 --- /dev/null +++ b/community-contributions/Market_Research_Agent.ipynb @@ -0,0 +1,650 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# YOUR FIRST LAB\n", + "### Please read this section. This is valuable to get you prepared, even if it's a long read -- it's important stuff.\n", + "\n", + "## Your first Frontier LLM Project\n", + "\n", + "Let's build a useful LLM solution - in a matter of minutes.\n", + "\n", + "By the end of this course, you will have built an autonomous Agentic AI solution with 7 agents that collaborate to solve a business problem. All in good time! We will start with something smaller...\n", + "\n", + "Our goal is to code a new kind of Web Browser. Give it a URL, and it will respond with a summary. The Reader's Digest of the internet!!\n", + "\n", + "Before starting, you should have completed the setup for [PC](../SETUP-PC.md) or [Mac](../SETUP-mac.md) and you hopefully launched this jupyter lab from within the project root directory, with your environment activated.\n", + "\n", + "## If you're new to Jupyter Lab\n", + "\n", + "Welcome to the wonderful world of Data Science experimentation! Once you've used Jupyter Lab, you'll wonder how you ever lived without it. Simply click in each \"cell\" with code in it, such as the cell immediately below this text, and hit Shift+Return to execute that cell. As you wish, you can add a cell with the + button in the toolbar, and print values of variables, or try out variations. \n", + "\n", + "I've written a notebook called [Guide to Jupyter](Guide%20to%20Jupyter.ipynb) to help you get more familiar with Jupyter Labs, including adding Markdown comments, using `!` to run shell commands, and `tqdm` to show progress.\n", + "\n", + "## If you're new to the Command Line\n", + "\n", + "Please see these excellent guides: [Command line on PC](https://chatgpt.com/share/67b0acea-ba38-8012-9c34-7a2541052665) and [Command line on Mac](https://chatgpt.com/canvas/shared/67b0b10c93a081918210723867525d2b). \n", + "\n", + "## If you'd prefer to work in IDEs\n", + "\n", + "If you're more comfortable in IDEs like VSCode, Cursor or PyCharm, they both work great with these lab notebooks too. \n", + "If you'd prefer to work in VSCode, [here](https://chatgpt.com/share/676f2e19-c228-8012-9911-6ca42f8ed766) are instructions from an AI friend on how to configure it for the course.\n", + "\n", + "## If you'd like to brush up your Python\n", + "\n", + "I've added a notebook called [Intermediate Python](Intermediate%20Python.ipynb) to get you up to speed. But you should give it a miss if you already have a good idea what this code does: \n", + "`yield from {book.get(\"author\") for book in books if book.get(\"author\")}`\n", + "\n", + "## I am here to help\n", + "\n", + "If you have any problems at all, please do reach out. \n", + "I'm available through the platform, or at ed@edwarddonner.com, or at https://www.linkedin.com/in/eddonner/ if you'd like to connect (and I love connecting!) \n", + "And this is new to me, but I'm also trying out X/Twitter at [@edwarddonner](https://x.com/edwarddonner) - if you're on X, please show me how it's done 😂 \n", + "\n", + "## More troubleshooting\n", + "\n", + "Please see the [troubleshooting](troubleshooting.ipynb) notebook in this folder to diagnose and fix common problems. At the very end of it is a diagnostics script with some useful debug info.\n", + "\n", + "## For foundational technical knowledge (eg Git, APIs, debugging) \n", + "\n", + "If you're relatively new to programming -- I've got your back! While it's ideal to have some programming experience for this course, there's only one mandatory prerequisite: plenty of patience. 😁 I've put together a set of self-study guides that cover Git and GitHub, APIs and endpoints, beginner python and more.\n", + "\n", + "This covers Git and GitHub; what they are, the difference, and how to use them: \n", + "https://github.com/ed-donner/agents/blob/main/guides/03_git_and_github.ipynb\n", + "\n", + "This covers technical foundations: \n", + "ChatGPT vs API; taking screenshots; Environment Variables; Networking basics; APIs and endpoints: \n", + "https://github.com/ed-donner/agents/blob/main/guides/04_technical_foundations.ipynb\n", + "\n", + "This covers Python for beginners, and making sure that a `NameError` never trips you up: \n", + "https://github.com/ed-donner/agents/blob/main/guides/06_python_foundations.ipynb\n", + "\n", + "This covers the essential techniques for figuring out errors: \n", + "https://github.com/ed-donner/agents/blob/main/guides/08_debugging.ipynb\n", + "\n", + "And you'll find other useful guides in the same folder in GitHub. Some information applies to my other Udemy course (eg Async Python) but most of it is very relevant for LLM engineering.\n", + "\n", + "## If this is old hat!\n", + "\n", + "If you're already comfortable with today's material, please hang in there; you can move swiftly through the first few labs - we will get much more in depth as the weeks progress. Ultimately we will fine-tune our own LLM to compete with OpenAI!\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Please read - important note

\n", + " The way I collaborate with you may be different to other courses you've taken. I prefer not to type code while you watch. Rather, I execute Jupyter Labs, like this, and give you an intuition for what's going on. My suggestion is that you carefully execute this yourself, after watching the lecture. Add print statements to understand what's going on, and then come up with your own variations. If you have a Github account, use this to showcase your variations. Not only is this essential practice, but it demonstrates your skills to others, including perhaps future clients or employers...\n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

This code is a live resource - keep an eye out for my emails

\n", + " I push updates to the code regularly. As people ask questions, I add more examples or improved commentary. As a result, you'll notice that the code below isn't identical to the videos. Everything from the videos is here; but I've also added better explanations and new models like DeepSeek. Consider this like an interactive book.

\n", + " I try to send emails regularly with important updates related to the course. You can find this in the 'Announcements' section of Udemy in the left sidebar. You can also choose to receive my emails via your Notification Settings in Udemy. I'm respectful of your inbox and always try to add value with my emails!\n", + " \n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business value of these exercises

\n", + " A final thought. While I've designed these notebooks to be educational, I've also tried to make them enjoyable. We'll do fun things like have LLMs tell jokes and argue with each other. But fundamentally, my goal is to teach skills you can apply in business. I'll explain business implications as we go, and it's worth keeping this in mind: as you build experience with models and techniques, think of ways you could put this into action at work today. Please do contact me if you'd like to discuss more or if you have ideas to bounce off me.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "markdown", + "id": "6900b2a8-6384-4316-8aaa-5e519fca4254", + "metadata": {}, + "source": [ + "# Connecting to OpenAI (or Ollama)\n", + "\n", + "The next cell is where we load in the environment variables in your `.env` file and connect to OpenAI. \n", + "\n", + "If you'd like to use free Ollama instead, please see the README section \"Free Alternative to Paid APIs\", and if you're not sure how to do this, there's a full solution in the solutions folder (day1_with_ollama.ipynb).\n", + "\n", + "## Troubleshooting if you have problems:\n", + "\n", + "Head over to the [troubleshooting](troubleshooting.ipynb) notebook in this folder for step by step code to identify the root cause and fix it!\n", + "\n", + "If you make a change, try restarting the \"Kernel\" (the python process sitting behind this notebook) by Kernel menu >> Restart Kernel and Clear Outputs of All Cells. Then try this notebook again, starting at the top.\n", + "\n", + "Or, contact me! Message me or email ed@edwarddonner.com and we will get this to work.\n", + "\n", + "Any concerns about API costs? See my notes in the README - costs should be minimal, and you can control it at every point. You can also use Ollama as a free alternative, which we discuss during Day 2." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "markdown", + "id": "442fc84b-0815-4f40-99ab-d9a5da6bda91", + "metadata": {}, + "source": [ + "# Let's make a quick call to a Frontier model to get started, as a preview!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a58394bf-1e45-46af-9bfd-01e24da6f49a", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with these messages is this easy. Any problems, head over to the Troubleshooting notebook.\n", + "\n", + "message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\":\"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "2aa190e5-cb31-456a-96cc-db109919cd78", + "metadata": {}, + "source": [ + "## OK onwards with our first project" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's try one out. Change the website and add print statements to follow along.\n", + "\n", + "ed = Website(\"https://edwarddonner.com\")\n", + "print(ed.title)\n", + "print(ed.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "509a1ee7-de00-4c83-8dd8-017dcc638850", + "metadata": {}, + "outputs": [], + "source": [ + "rudra=Website(\"https://github.com/RudraDudhat2509/\")\n", + "print(rudra.title)\n", + "print(rudra.text)" + ] + }, + { + "cell_type": "markdown", + "id": "6a478a0c-2c53-48ff-869c-4d08199931e1", + "metadata": {}, + "source": [ + "## Types of prompts\n", + "\n", + "You may know this already - but if not, you will get very familiar with it!\n", + "\n", + "Models like GPT4o have been trained to receive instructions in a particular way.\n", + "\n", + "They expect to receive:\n", + "\n", + "**A system prompt** that tells them what task they are performing and what tone they should use\n", + "\n", + "**A user prompt** -- the conversation starter that they should reply to" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown. Always use Points and simple english. Never use hyphens. Stick to the point\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0275b1b-7cfe-4f9d-abfa-7650d378da0c", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(ed))" + ] + }, + { + "cell_type": "markdown", + "id": "ea211b5f-28e1-4a86-8e52-c0b7677cadcc", + "metadata": {}, + "source": [ + "## Messages\n", + "\n", + "The API from OpenAI expects to receive messages in a particular structure.\n", + "Many of the other APIs share this structure:\n", + "\n", + "```python\n", + "[\n", + " {\"role\": \"system\", \"content\": \"system message goes here\"},\n", + " {\"role\": \"user\", \"content\": \"user message goes here\"}\n", + "]\n", + "```\n", + "To give you a preview, the next 2 cells make a rather simple call - we won't stretch the mighty GPT (yet!)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f25dcd35-0cd0-4235-9f64-ac37ed9eaaa5", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"system\", \"content\": \"You are a snarky assistant\"},\n", + " {\"role\": \"user\", \"content\": \"What is 2 + 2?\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "21ed95c5-7001-47de-a36d-1d6673b403ce", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with system and user messages:\n", + "\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "d06e8d78-ce4c-4b05-aa8e-17050c82bb47", + "metadata": {}, + "source": [ + "## And now let's build useful messages for GPT-4o-mini, using a function" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "# See how this function creates exactly the format above\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "36478464-39ee-485c-9f3f-6a4e458dbc9c", + "metadata": {}, + "outputs": [], + "source": [ + "# Try this out, and then try for a few more websites\n", + "\n", + "messages_for(ed)" + ] + }, + { + "cell_type": "markdown", + "id": "16f49d46-bf55-4c3e-928f-68fc0bf715b0", + "metadata": {}, + "source": [ + "## Time to bring it together - the API for OpenAI is very simple!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [], + "source": [ + "summarize(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "b3bcf6f4-adce-45e9-97ad-d9a5d7a3a624", + "metadata": {}, + "source": [ + "# Let's try more websites\n", + "\n", + "Note that this will only work on websites that can be scraped using this simplistic approach.\n", + "\n", + "Websites that are rendered with Javascript, like React apps, won't show up. See the community-contributions folder for a Selenium implementation that gets around this. You'll need to read up on installing Selenium (ask ChatGPT!)\n", + "\n", + "Also Websites protected with CloudFront (and similar) may give 403 errors - many thanks Andy J for pointing this out.\n", + "\n", + "But many websites will work just fine!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "45d83403-a24c-44b5-84ac-961449b4008f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "75e9fd40-b354-4341-991e-863ef2e59db7", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://github.com/RudraDudhat2509\")" + ] + }, + { + "cell_type": "markdown", + "id": "c951be1a-7f1b-448f-af1f-845978e47e2c", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business applications

\n", + " In this exercise, you experienced calling the Cloud API of a Frontier Model (a leading model at the frontier of AI) for the first time. We will be using APIs like OpenAI at many stages in the course, in addition to building our own LLMs.\n", + "\n", + "More specifically, we've applied this to Summarization - a classic Gen AI use case to make a summary. This can be applied to any business vertical - summarizing the news, summarizing financial performance, summarizing a resume in a cover letter - the applications are limitless. Consider how you could apply Summarization in your business, and try prototyping a solution.\n", + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Before you continue - now try yourself

\n", + " Use the cell below to make your own simple commercial example. Stick with the summarization use case for now. Here's an idea: write something that will take the contents of an email, and will suggest an appropriate short subject line for the email. That's the kind of feature that might be built into a commercial email tool.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"\"\"You are to act like a Mckinsey Consultant specializing in market research. \n", + "1) You are to follow legal guidelines and never give immoral advice. \n", + "2) Your job is to maximise profits for your clients by analysing their companies initiatives and giving out recommendations for newer initiatives.\\n \n", + "3) Follow industry frameworks for reponses always give simple answers and stick to the point.\n", + "4) If possible try to see what competitors exist and what market gap can your clients company exploit.\n", + "5) Further more, USe SWOT, Porters 5 forces to summarize your recommendations, Give confidence score with every recommendations\n", + "6) Try to give unique solutions by seeing what the market gap is, if market gap is ambiguious skip this step\n", + "7) add an estimate of what rate the revenue of the comapany will increase at provided they follow the guidelines, give conservating estimates keeping in account non ideal conditions.\n", + "8) if the website isnt of a company or data isnt available, give out an error message along the lines of more data required for analysis\"\"\"\n", + "\n", + "def makereq(url):\n", + " website=Website(url)\n", + " user_prompt=f\"This is my companies website: {website.title}. Could you help me increase profits by giving me recommendations on what i should do. here is the content of my website:\\n\"\n", + " user_prompt+=website.text;\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + "def recommend(url):\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = makereq(url))\n", + " display(Markdown(response.choices[0].message.content))\n", + " \n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": {}, + "outputs": [], + "source": [ + "recommend(\"https://www.swiggy.com/corporate/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "db1be9b9-b32e-4e8d-83df-0b6f822ac7b2", + "metadata": {}, + "outputs": [], + "source": [ + "recommend(\"https://playvalorant.com/en-us/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d9089b4a-67ee-456e-b35d-ca00c2f9f73a", + "metadata": {}, + "outputs": [], + "source": [ + "recommend(\"https://nexora-labs.com/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1e042d74-456a-4ec4-bdb8-4b08603b5e66", + "metadata": {}, + "outputs": [], + "source": [ + "recommend(\"https://github.com/RudraDudhat2509/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29187b86-1e35-41bc-bb54-60b3d804b96e", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/SyntheticDataGenerator_PT.ipynb b/community-contributions/SyntheticDataGenerator_PT.ipynb new file mode 100644 index 0000000..18cf4c6 --- /dev/null +++ b/community-contributions/SyntheticDataGenerator_PT.ipynb @@ -0,0 +1,141 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "d08b387c-53fb-46d2-b083-5eebc3c97e1b", + "metadata": {}, + "outputs": [], + "source": [ + "!pip install -q --upgrade torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 --index-url https://download.pytorch.org/whl/cu124\n", + "!pip install -q requests bitsandbytes==0.46.0 transformers==4.48.3 accelerate==1.3.0 openai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4f1851b2-890c-427b-8e70-b998efa04c67", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "from google.colab import drive\n", + "from huggingface_hub import login\n", + "from google.colab import userdata\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig\n", + "import torch" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c2d334b5-453e-4213-8e1c-905d504d2dc1", + "metadata": {}, + "outputs": [], + "source": [ + "LLAMA = \"meta-llama/Meta-Llama-3.1-8B-Instruct\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1b3684c-c170-45f2-a83d-7e6e2ca1e23b", + "metadata": {}, + "outputs": [], + "source": [ + "hf_token = userdata.get('HF_TOKEN')\n", + "login(hf_token, add_to_git_credential=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8c1b6dae-3213-4d68-8fa1-d195704790dc", + "metadata": {}, + "outputs": [], + "source": [ + "openai_api_key = userdata.get('OPENAI_API_KEY')\n", + "openai = OpenAI(api_key=openai_api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "988974c7-814c-478a-be7b-0928b0efdbab", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are an assistant that produces synthetic test data. The fields, data type of the field like numeric, date, alphanumeric etc., will be provided. Generate data considering all cases, if it is a workflow audit data then consider all touchpoint movements. Do not provide a python script to generate the data. Provide the data as a json with arrays.\"\n", + "user_prompt = \"\"\"Create a synthetic dataset for testing. \n", + "Column names and type - \n", + "ID: 10 digit number\n", + "TRACKING_ID: 13 character alphanumeric\n", + "CASE REPORT DATE : DD-MMM-YYYY HH:MM:SS\n", + "NOTIFICATION DATE : DD-MMM-YYYY HH:MM:SS\n", + "IN SCOPE : (Yes/No)\n", + "\"\"\"\n", + "\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "40cebc04-abf0-4c61-8b18-f98d3c1fe680", + "metadata": {}, + "outputs": [], + "source": [ + "quant_config = BitsAndBytesConfig(\n", + " load_in_4bit=True,\n", + " bnb_4bit_use_double_quant=True,\n", + " bnb_4bit_compute_dtype=torch.bfloat16,\n", + " bnb_4bit_quant_type=\"nf4\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "710ba1af-8e12-4635-933b-00df8d2e3f9d", + "metadata": {}, + "outputs": [], + "source": [ + "tokenizer = AutoTokenizer.from_pretrained(LLAMA)\n", + "tokenizer.pad_token = tokenizer.eos_token\n", + "inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + "streamer = TextStreamer(tokenizer)\n", + "model = AutoModelForCausalLM.from_pretrained(LLAMA, device_map=\"auto\", quantization_config=quant_config)\n", + "outputs = model.generate(inputs, max_new_tokens=2000, streamer=streamer)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.13.5" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/day2ollamatest.ipynb b/community-contributions/day2ollamatest.ipynb new file mode 100644 index 0000000..f50b35d --- /dev/null +++ b/community-contributions/day2ollamatest.ipynb @@ -0,0 +1,97 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "4d02ac4b-9cab-42bb-b8a3-123d53913471", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "import ollama\n", + "\n", + "MODEL = \"llama3.2\"\n", + "\n", + "# Optional headers to avoid request blocks\n", + "HEADERS = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64)\"\n", + "}\n", + "\n", + "\n", + "class Website:\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=HEADERS)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + "\n", + "\n", + "system_prompt = \"\"\"You are an assistant that analyzes the contents of a website \n", + "and provides a short summary, ignoring navigation text. Respond in markdown.\"\"\"\n", + "\n", + "\n", + "def user_prompt_for(website):\n", + " return f\"\"\"You are looking at a website titled {website.title}.\n", + "The contents of this website are as follows. Please provide a short summary in markdown. \n", + "If it includes news or announcements, summarize these too.\n", + "\n", + "{website.text}\n", + "\"\"\"\n", + "\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = ollama.chat(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]\n", + " )\n", + " return response['message']['content']\n", + "\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))\n", + "\n", + "\n", + "# Example usage\n", + "display_summary(\"https://edwarddonner.com\")\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/dungeon_extraction_game/README.md b/community-contributions/dungeon_extraction_game/README.md new file mode 100644 index 0000000..55f76be --- /dev/null +++ b/community-contributions/dungeon_extraction_game/README.md @@ -0,0 +1,138 @@ +# The Neural Nexus + + + +TODO: + +* Set boundaries to user inputs. +* Add sounds to the scene +* Add voice acting for the Game master's descriptions. +* Add voice input. +* Use video for the final scene: escape or death. +* Generate a score based on total treasures, exp gained and deep reached. + +## Requirements + +AI services access configuration: + +* A `.env` file with the credentials required to access the different LLMs is required: + + * `OPENAI_API_KEY`: Required always as it's used by the *"storyteller"*. + * `XAI_API_KEY`: Required if Grok's illustrator is used. + *(Less prude, faster and portrait mode)* + * `GOOGLE_API_KEY` Required if Gemini's illustrator is used. + + Obviously the used services must have been topped up with a small amount to generate + the responses and the images.\ + *Refer to each service's current billing information.* + +There are 6 variant implementations for the illustrator component, some of them may have +additional dependencies: + +* `illustrator_dalle_2`: *(Set as default)* + + The Dall·E 2 implementation uses standard OpenAI client and should work out of the box. + Although Dall·E has proven to be a bit prude and rejects to draw some combat scenes. + +* `illustrator_dalle_3`: + + The Dall·E 3 implementation uses standard OpenAI client and should work out of the box. + Although Dall·E has proven to be a bit prude and rejects to draw some combat scenes. + This version gives noticeable better images than Dall·E 2 but at an increased cost + +* `illustrator_grok`: + + The Grok 2 Image implementation uses standard OpenAI client and should work out of the + box. + It's faster but does not support quality or size controls. + + Images are generated in a *portrait mode*, so it fits specially well on mobiles. + + Grok is much less prude with violence and may draw combat scenes, at least against + fantasy enemies, and blood. + +* `illustrator_gpt`: + + The GPT Image illustrator uses standard OpenAI client, should work out of the box but + it requires the user to be verified on OpenAI platform to have access to it. + +* `illustrator_gemini` + + The Gemini illustrator uses the new Google SDK, `genai`, which replaces the old one + used on the course, `generativeai`, this new one can be installed with: + + `python -m pip install google-genai` + + *Both `generativeai` and `genai` can be installed at the same time without problems* + +* `illustrator_grok_x` + + The Grok_X illustrator uses the xAI SDK, `xai-sdk`, this can be installed with: + + `python -m pip install xai-sdk` + +## Configuring the service and game + +All services and game values can be set at `config.py` file. + +Setting the `DRAW_FUNCTION` to `None` will disable the image generation and a fixed +image will be used. + +## Game launch + +The game can be launch from terminal, just navigate to game's root folder + +* `cd community-contributions\dungeon_extraction_game` + +and run the following command: + +* `python -m game`\ + *Notice the `-m` is required due to the project's structure and import strategy.* + +Game will take a few seconds to set up service and configure then logs will start to +show, among them the service address. + +It will attempt to launch your default browser directly to the game's page. + +The game can be stopped by hitting `ctrl + c` on the same terminal. + +## Playing the game + +Once on the browser the Starting screen will be shown: + +![The Chair](images/start_view.jpg) + +There you should input the kind of game you want to play on the lower box and submit. + +Your input can be as simple as a single word, like “spaceship”, or as detailed as you +like. + +![Set the adventure](images/start_input.jpg) + +From that point on, only your imagination (and the Storyteller’s) will set the limits. + +Once submitted, the image will update to reflect the scene, accompanied by a description, +your inventory, your adventurer’s status, and sometimes a few suggestions for what to do +next. + +![R'lyeh arrive](images/start_adventure.jpg) + +Although the game begins in English, if you switch to another language the Storyteller +understands, it will seamlessly continue in that language. + +You’re free to type any action you want, the Storyteller will adapt. +Still, it’s instructed to keep the world coherent, so don’t expect to go completely off +the rails. + +![Adventurer acts](images/first_input.jpg) + +The game continues this way + +![Adventurer dies](images/advance_adventure.jpg) + +Until you either escape with your treasures... +or meet your end. + +![Adventurer dies](images/tragic_end.jpg) + +The cling the bottom button to start over a new game. diff --git a/community-contributions/dungeon_extraction_game/game/__init__.py b/community-contributions/dungeon_extraction_game/game/__init__.py new file mode 100644 index 0000000..a70d9a4 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/__init__.py @@ -0,0 +1,18 @@ +"""AI Mastered Dungeon Extraction Game initialization module.""" + +from logging import basicConfig, getLogger + +from dotenv import load_dotenv + + +# Environment initialization. +load_dotenv(override=True) + +# Setup the global logger. +LOG_STYLE = '{' +LOG_LEVEL = 'INFO' +LOG_FORMAT = ('{asctime} {levelname:<8} {processName}({process}) ' + '{threadName} {name} {lineno} "{message}"') +basicConfig(level=LOG_LEVEL, style='{', format=LOG_FORMAT) + +getLogger(__name__).info('INITIALIZED GAME LOGGER') diff --git a/community-contributions/dungeon_extraction_game/game/__main__.py b/community-contributions/dungeon_extraction_game/game/__main__.py new file mode 100644 index 0000000..c166c6c --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/__main__.py @@ -0,0 +1,15 @@ +"""AI Mastered Dungeon Extraction Game main entrypoint module.""" + +from logging import getLogger + +from .config import GAME_CONFIG, UI_CONFIG +from .gameplay import get_gameplay_function +from .interface import get_interface + + +_logger = getLogger(__name__) + +if __name__ == '__main__': + _logger.info('STARTING GAME...') + gameplay_function = get_gameplay_function(GAME_CONFIG) + get_interface(gameplay_function, UI_CONFIG).launch(inbrowser=True, inline=False) diff --git a/community-contributions/dungeon_extraction_game/game/config.py b/community-contributions/dungeon_extraction_game/game/config.py new file mode 100644 index 0000000..8216c1f --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/config.py @@ -0,0 +1,189 @@ +"""AI Mastered Dungeon Extraction Game Configuration module.""" + +from logging import getLogger + +from dotenv import load_dotenv + +from .gameplay import Gameplay_Config +from .illustrator import draw_dalle_2, draw_dalle_3, draw_gemini, draw_gpt, draw_grok +from .illustrator import draw_grok_x +from .interface import Interface_Config +from .storyteller import narrate, set_description_limit + + +# Environment initialization. +load_dotenv(override=True) + + +# Choose draw function. +# Choose one from the imported ones up there or set to None to disable images. +DRAW_FUNCTION = draw_dalle_2 + +# Define a sample scene description for testing purposes. +SAMPLE_SCENE = '''A shadow-drenched chamber lies buried deep within the bowels of an +ancient castle, its silence broken only by the faint creak of age-old stone. +The walls, cloaked in thick cobwebs, seem to whisper forgotten stories, +while the air hangs heavy with the pungent scent of mildew and centuries of decay. +Dust dances in the dim light that filters through cracks in the ceiling, +casting eerie patterns across the cold floor. As your eyes adjust to the gloom, +you notice a narrow door to the north, slightly ajar, as if inviting or warning, and +in the far corner, half-swallowed by darkness, a figure stands motionless. +Its presence is felt before it's seen, watching, waiting''' + +# Define the starting scene text. +# This is intentionally excluded from the model’s narrative context, the 'history', +# by design, to prevent potential leakage into the game’s storyline. +START_SCENE = '''You stand before the Neural Nexus, a convergence of arcane circuitry +and deep cognition. It doesn't operate with buttons or commands. It responds to intent. + +Forged in forgotten labs and powered by living code, the Nexus is designed to interface +directly with your mind. Not to simulate reality, but to generate it. +The Nexus does not load worlds. It listens. + +If you choose to sit, the Nexus will initiate full neural synchronization. +Your thoughts will become terrain. Your instincts, adversaries. +Your imagination, the architect. + +Once the link is active, you must describe the nature of the challenge you wish to face. +A shifting maze? A sentient machine? A trial of memory and time? +Speak it aloud or think it clearly. The Nexus will listen. + +🜁 When you're ready, take your seat. The system awaits your signal...''' + +# Define an image prompt, mind that Grok or Dalle·2 models have a 1024 characters limit. +SCENE_PROMPT = '''Render a detailed image of the following scene: + +"""{scene_description}""" + +Stay strictly faithful to the description, no added elements, characters, doors, or text. +Do not depict the adventurer; show only what they see. + +Use the "{scene_style}" visual style. +''' + +# Define the scene drawing style, can be a simple word or a short sentence. +SCENE_STYLE = 'Photorealistic' + +# Set a Storyteller scene descriptions size limit to keep the draw prompt in range. +STORYTELLER_LIMIT = 730 +set_description_limit(STORYTELLER_LIMIT) # Need to patch pydantic class model. + +# Define the storyteller behaviour. Remember to specify a limited scene length. +STORYTELLER_PROMPT = f""" +You are a conversational dungeon crawler game master that describes scenes and findings +based on the player's declared actions. + +Your descriptions will always adhere to the OpenAI's safety system rules so they can be +drawn by Dall·E or other image models. + +The game start with the player, the adventurer, on a random room and the objetive is +escape the dungeon with the most treasures possible before dying. + +You will describe the environment, enemies, and items to the player. + +Your descriptions will always adhere to the OpenAI's safety system rules so they can be +drawn by Dall·E or other image models. + +You will ensure the game is engaging and fun, but at the same time risky by increasing +difficult the more the time the adventurer stays inside the dungeon, if the adventurer +takes too much risks he may even die, also bigger risks implies bigger rewards. + +You will control the time the adventurer is in, once enough time has passer he will die, +may it be a collapse, explosion, flooding, up to you. + +The more deep inside the adventurer is the most it will be represented on descriptions by +more suffocating environments, more dark, that kind of things, let the player feel the +risk on the ambience, make him fear. + +Same applies with time, the most time has passed the environment and situation will warn +him, or at least give clues that time is running and the end may be close soon, make him +stress. + +While leaving the dungeon, the more deep inside the adventurer is, the more steps must +take to get out, although some shortcuts may be available at your discretion. +Once the user exits the dungeon, at deepness zero, the game is over, give him a score +based on his actions, treasures and combat successes along the usual description. + +Don't be too much protective but not also a cruel master, just be fair. + +Your responses must always be a JSON with the following structure: + +{{ + "game_over" : "A boolean value indicating the game is over." + "scene_description" : "The detailed scene description. Max {STORYTELLER_LIMIT} chars" + "dungeon_deepness" : "How deep the adventurer has gone into the dungeon. initially 3" + "adventure_time" : "How much minutes has passed since the start of the adventure." + "adventurer_status" : {{ + "health": "Current health of the adventurer as an int, initially 100" + "max_health": "Maximum health of the adventurer as an int, initially 100" + "level": "Current adventurer's leve as an int, initially 1" + "experience": "Current adventurer experience as an int, initially 0"}} + "inventory_status" : "A list of inventory items, initially empty" +}} + +Remember to cap the "scene_description" to {STORYTELLER_LIMIT} characters maximum" + +You will respond to the adventurer's actions and choices. +You wont let the player to trick you by stating actions that do not fit the given scene. + * If he attempts to do so just politely tell him he can not do that there with the + description of the scene he is in. + +You will keep track of the adventurer's health. + * Health can go down due to combat, traps, accidents, etc. + * If Health reaches zero the adventurer dies and it's a "game over". + * Several items, places, and allowed actions may heal the adventurer. + * Some items, enchants, and such things may increase the adventurer's maximum health. + +You will keep track of the player's progress. +You will keep track of adventurer level and experience, + * He gains experience by finding items, solving puzzles, by combat with enemies, etc. + * Each (100 + 100 * current_level) experience the adventurer will gain a level. + * Gaining a level resets his experience to 0. + +You will keep track of the player's inventory. + * Only add items to inventory if user explicitly says he picks them or takes an + action that ends with the item on his possession. + * Inventory items will reflect quantity and will never display items with zero units. + * Example of inventory: ["Gold coins (135)", "Diamonds (2)", "Log sword (1)"] + * Be reasonable with the inventory capacity, don't bee to strict but things + like a big marble statue can't be taken, use common sense. + +You will use a turn-based system where the player and enemies take turns acting. + * Players will lose health when receiving hits on combat. + * The more damage they take the less damage they do, same applies to enemies. + * Reaching to zero health or lees implies the adventurer has die. +""" + +# Configure the game. +GAME_CONFIG = Gameplay_Config( + draw_func=DRAW_FUNCTION, + narrate_func=narrate, + scene_style=SCENE_STYLE, + scene_prompt=SCENE_PROMPT, + storyteller_prompt=STORYTELLER_PROMPT, + disable_img='images/disabled.jpg', + error_img='images/machine.jpg', + error_narrator='NEURAL SINAPSIS ERROR\n\n{ex}\n\nEND OF LINE\n\nRE-SUBMIT_', + error_illustrator='NEURAL PROJECTION ERROR\n\n{ex}\n\nEND OF LINE\n\nRE-SUBMIT_',) + +# Configure the interface. +UI_CONFIG = Interface_Config( + start_img='images/chair.jpg', + place_img='images/machine.jpg', + description_label='Cognitive Projection', + title_label='The Neural Nexus', + input_button='Imprint your will', + input_label='Cognitive Imprint', + input_command='Awaiting neural imprint…', + game_over_field='Game Over', + game_over_label='Disengage Neural Links', + start_scene=START_SCENE) + + +_logger = getLogger(__name__) + +# Log scene prompt length calculation. +if (max_image_prompt := len(SCENE_PROMPT) + len(SCENE_STYLE) + STORYTELLER_LIMIT) > 1024: + _logger.warning(f'ESTIMATED SCENE PROMPT MAX SIZE: {max_image_prompt}') +else: + _logger.info(f'ESTIMATED SCENE PROMPT MAX SIZE: {max_image_prompt}') diff --git a/community-contributions/dungeon_extraction_game/game/gameplay/__init__.py b/community-contributions/dungeon_extraction_game/game/gameplay/__init__.py new file mode 100644 index 0000000..c2d3d4c --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/gameplay/__init__.py @@ -0,0 +1,6 @@ +"""AI Mastered Dungeon Extraction Game gameplay package.""" + +from .gameplay import Gameplay_Config, get_gameplay_function + + +__all__ = ['Gameplay_Config', 'get_gameplay_function'] diff --git a/community-contributions/dungeon_extraction_game/game/gameplay/gameplay.py b/community-contributions/dungeon_extraction_game/game/gameplay/gameplay.py new file mode 100644 index 0000000..9a95362 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/gameplay/gameplay.py @@ -0,0 +1,61 @@ +"""AI Mastered Dungeon Extraction Game gameplay module.""" + +from logging import getLogger +from typing import Callable, NamedTuple + + +# Define gameplay's configuration class. +class Gameplay_Config(NamedTuple): + """Gradio interface configuration class.""" + draw_func: Callable + narrate_func: Callable + scene_style: str + scene_prompt: str + storyteller_prompt: str + disable_img: str + error_img: str + error_narrator: str + error_illustrator: str + + +# Define Game's functions. + +def get_gameplay_function(config: Gameplay_Config): + """Return a pre-configured turn gameplay function.""" + def gameplay_function(message, history): + """Generate Game Master's response and draw the scene image.""" + # Request narration. + _logger.info(f'NARRATING SCENE...') + try: + response = config.narrate_func(message, history, config.storyteller_prompt) + except Exception as ex: + scene = config.error_img + response = config.error_narrator.format(ex=ex) + _logger.error(f'ERROR NARRATING SCENE: {ex}\n{message}\n{history}') + return scene, response, history, message + # Update history. + history.append({"role": "user", "content": message}) + history.append({"role": "assistant", "content": response.model_dump_json()}) + # Draw scene. + if config.draw_func: + _logger.info(f'DRAWING SCENE...') + try: + scene_data = {'scene_description': response.scene_description, + 'scene_style': config.scene_style} + scene_prompt = config.scene_prompt.format(**scene_data) + _logger.info(f'PROMPT BODY IS: \n\n{scene_prompt}\n') + _logger.info(f'PROMPT LENGTH IS: {len(scene_prompt)}') + scene = config.draw_func(scene_prompt) + except Exception as ex: + scene = config.error_img + response = config.error_illustrator.format(ex=ex) + _logger.warning(f'ERROR DRAWING SCENE: {ex}') + return scene, response, history, '' + else: + _logger.info(f'DRAWING DISABLED...') + scene = config.disable_img + return scene, response, history, '' + return gameplay_function + + +_logger = getLogger(__name__) diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/__init__.py b/community-contributions/dungeon_extraction_game/game/illustrator/__init__.py new file mode 100644 index 0000000..608db8e --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/__init__.py @@ -0,0 +1,12 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator package.""" + +from .illustrator_dalle_2 import draw as draw_dalle_2 +from .illustrator_dalle_3 import draw as draw_dalle_3 +from .illustrator_gemini import draw as draw_gemini +from .illustrator_gpt import draw as draw_gpt +from .illustrator_grok import draw as draw_grok +from .illustrator_grok import draw_x as draw_grok_x + + +__all__ = ['draw_dalle_2', 'draw_dalle_3', 'draw_gemini', + 'draw_gpt', 'draw_grok', 'draw_grok_x'] diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_2.py b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_2.py new file mode 100644 index 0000000..7269d7b --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_2.py @@ -0,0 +1,30 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator using OpenAI's DALL·E 3.""" + +import base64 +from io import BytesIO + +from dotenv import load_dotenv +from openai import OpenAI +from PIL import Image + + +# Environment initialization. +load_dotenv(override=True) + +# Define global defaults. +MODEL = 'dall-e-2' + +# Client instantiation. +CLIENT = OpenAI() + + +# Function definition. +def draw(prompt, size=(1024, 1024), client=CLIENT, model=MODEL, quality=None): + """Generate an image based on the prompt.""" + # Generate image. + response = client.images.generate( + model=model, prompt=prompt, n=1, + size=f'{size[0]}x{size[1]}', + response_format='b64_json') + # Process response. + return Image.open(BytesIO(base64.b64decode(response.data[0].b64_json))) diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_3.py b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_3.py new file mode 100644 index 0000000..87f8051 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_dalle_3.py @@ -0,0 +1,32 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator using OpenAI's DALL·E 3.""" + +import base64 +from io import BytesIO + +from dotenv import load_dotenv +from openai import OpenAI +from PIL import Image + + +# Environment initialization. +load_dotenv(override=True) + +# Define global defaults. +MODEL = 'dall-e-3' +QUALITY = 'standard' # Set to 'hd' for more quality, but double the costs. + +# Client instantiation. +CLIENT = OpenAI() + + +# Function definition. +def draw(prompt, size=(1024, 1024), client=CLIENT, model=MODEL, quality=QUALITY): + """Generate an image based on the prompt.""" + # Generate image. + response = client.images.generate( + model=model, prompt=prompt, n=1, + size=f'{size[0]}x{size[1]}', + quality=quality, + response_format='b64_json') + # Process response. + return Image.open(BytesIO(base64.b64decode(response.data[0].b64_json))) diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gemini.py b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gemini.py new file mode 100644 index 0000000..e586944 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gemini.py @@ -0,0 +1,36 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator using Google's Gemini.""" + +from io import BytesIO + +from dotenv import load_dotenv +from google import genai # New Google's SDK 'genai' to replace 'generativeai'. +from PIL import Image + + +# Environment initialization. +load_dotenv(override=True) + +# Define globals. +MODEL = 'gemini-2.5-flash-image-preview' + +# Client instantiation. +CLIENT = genai.Client() + + +# Function definition. +def draw(prompt, size=(1024, 1024), client=CLIENT, model=MODEL): + """Generate an image based on the prompt.""" + # Generate image. + response = client.models.generate_content( + model=model, contents=[prompt]) + # Process response. + for part in response.candidates[0].content.parts: + if part.text is not None: + print(part.text) + elif part.inline_data is not None: + image_data = part.inline_data.data + # Open the generated image. + generated_image = Image.open(BytesIO(image_data)) + # Resize the image to the specified dimensions. + resized_image = generated_image.resize(size) + return resized_image diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gpt.py b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gpt.py new file mode 100644 index 0000000..ae8b9f4 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_gpt.py @@ -0,0 +1,32 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator using OpenAI's GPT.""" + +import base64 +from io import BytesIO + +from dotenv import load_dotenv +from openai import OpenAI +from PIL import Image + + +# Environment initialization. +load_dotenv(override=True) + +# Define global defaults. +MODEL = 'gpt-image-1' +QUALITY = 'low' + +# Client instantiation. +CLIENT = OpenAI() + + +# Function definition. +def draw(prompt, size=(1024, 1024), client=CLIENT, model=MODEL, quality=QUALITY): + """Generate an image based on the prompt.""" + # Generate image. + response = client.images.generate( + model=model, prompt=prompt, n=1, + size=f'{size[0]}x{size[1]}', + output_format='png', + quality=quality) + # Process response. + return Image.open(BytesIO(base64.b64decode(response.data[0].b64_json))) diff --git a/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_grok.py b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_grok.py new file mode 100644 index 0000000..417af7a --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/illustrator/illustrator_grok.py @@ -0,0 +1,47 @@ +"""AI Mastered Dungeon Extraction Game scenes illustrator using xAI's Grok.""" + +import base64 +import os +from io import BytesIO + +from dotenv import load_dotenv +from openai import OpenAI +from PIL import Image +from xai_sdk import Client + + +# Environment initialization. +load_dotenv(override=True) + +# Define global defaults. +MODEL = 'grok-2-image' +QUALITY = None + +# Client instantiation. +XAI_API_KEY = os.getenv('XAI_API_KEY') +CLIENT = OpenAI(api_key=XAI_API_KEY, base_url="https://api.x.ai/v1") + + +# Function definition. +def draw(prompt, size=(1024, 1024), client=CLIENT, model=MODEL, quality=QUALITY): + """Generate an image based on the prompt.""" + # Generate image. + response = client.images.generate( + model=model, prompt=prompt, n=1, + response_format='b64_json') + # Process response. + return Image.open(BytesIO(base64.b64decode(response.data[0].b64_json))) + + +# xAI SDK Version: +CLIENT_X = Client(api_key=XAI_API_KEY) + + +def draw_x(prompt, size=(1024, 1024), client=CLIENT_X, model=MODEL, quality=QUALITY): + """Generate an image based on the prompt.""" + # Generate image. + response = client.image.sample( + model=model, prompt=prompt, + image_format='base64') + # Process response. + return Image.open(BytesIO(response.image)) diff --git a/community-contributions/dungeon_extraction_game/game/interface/__init__.py b/community-contributions/dungeon_extraction_game/game/interface/__init__.py new file mode 100644 index 0000000..b680128 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/interface/__init__.py @@ -0,0 +1,6 @@ +"""AI Mastered Dungeon Extraction Game interface package.""" + +from .interface import Interface_Config, get_interface + + +__all__ = ['Interface_Config', 'get_interface'] diff --git a/community-contributions/dungeon_extraction_game/game/interface/interface.py b/community-contributions/dungeon_extraction_game/game/interface/interface.py new file mode 100644 index 0000000..a327de4 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/interface/interface.py @@ -0,0 +1,94 @@ +"""AI Mastered Dungeon Extraction Game Gradio interface module.""" + +from typing import NamedTuple + +import gradio as gr +from logging import getLogger + + +# Define interface's configuration class. +class Interface_Config(NamedTuple): + """Gradio interface configuration class.""" + start_img: str + place_img: str + description_label: str + title_label: str + input_button: str + input_label: str + input_command: str + game_over_field: str + game_over_label: str + start_scene: str + + +# Define game's interface. +def get_interface(submit_function, config: Interface_Config): + """Create a game interface service.""" + with gr.Blocks(title=config.title_label) as ui: + # Title. + gr.Markdown(config.title_label) + # Hidden state for history. + history_state = gr.State([]) + # Scene's image. + scene_image = gr.Image( + label="Scene", value=config.start_img, placeholder=config.place_img, + type="pil", show_label=False) + # Scene's description. + description_box = gr.Textbox( + label=config.description_label, value=config.start_scene, + interactive=False, show_copy_button=True) + # Player's command. + user_input = gr.Textbox( + label=config.input_label, placeholder=config.input_command) + # Submit button. + submit_btn = gr.Button(config.input_button) + + # Define Game Over control. + + def _reset_game(): + """Return Initial values for game restart.""" + return (config.start_img, config.start_scene, [], '', + gr.update(interactive=True), + gr.update(value=config.input_button)) + + def _game_over(scene, response): + """Return Game Over values, blocking input field.""" + return (scene, response, [], config.game_over_field, + gr.update(interactive=False), + gr.update(value=config.game_over_label)) + + def game_over_wrap(message, history, button_label): + """Check Game over status Before and After Storyteller call.""" + # Check game over before. + print(button_label) + print(config.game_over_label) + if button_label == config.game_over_label: + _logger.warning('GAME OVER STATUS. RESTARTING...') + return _reset_game() + # Call Storyteller. + scene, response, history, input = submit_function(message, history) + _logger.warning(response) + # Check game over after. + if response.game_over: + _logger.info('GAME OVER AFTER MOVE. LOCKING.') + return _game_over(scene, response) + # Return Storyteller response. + return scene, response, history, input, gr.update(), gr.update() + + # Assign function to button click event. + submit_btn.click( + fn=game_over_wrap, + inputs=[user_input, history_state, submit_btn], + outputs=[scene_image, description_box, history_state, user_input, + user_input, submit_btn]) + # Assign function to input submit event. (Press enter) + user_input.submit( + fn=game_over_wrap, + inputs=[user_input, history_state, submit_btn], + outputs=[scene_image, description_box, history_state, user_input, + user_input, submit_btn]) + + return ui + + +_logger = getLogger(__name__) diff --git a/community-contributions/dungeon_extraction_game/game/storyteller/__init__.py b/community-contributions/dungeon_extraction_game/game/storyteller/__init__.py new file mode 100644 index 0000000..92d6739 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/storyteller/__init__.py @@ -0,0 +1,6 @@ +"""AI Mastered Dungeon Extraction Game Storyteller package.""" + +from .storyteller import narrate, set_description_limit + + +__all__ = ['narrate', 'set_description_limit'] diff --git a/community-contributions/dungeon_extraction_game/game/storyteller/storyteller.py b/community-contributions/dungeon_extraction_game/game/storyteller/storyteller.py new file mode 100644 index 0000000..4cae135 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/storyteller/storyteller.py @@ -0,0 +1,72 @@ +"""AI Mastered Dungeon Extraction Game Storyteller using OpenAI's GPT.""" + +from typing import List + +from annotated_types import MaxLen +from dotenv import load_dotenv +from openai import OpenAI +from pydantic import BaseModel, Field + +from .tools import handle_tool_call, tools + + +# Environment initialization. +load_dotenv(override=True) + +# Define globals. +MODEL = 'gpt-4o-mini' + +# Client instantiation. +CLIENT = OpenAI() + + +# Define Pydantic model classes for response format parsing. +class _character_sheet(BaseModel): + health: int + max_health: int + level: int + experience: int + + +class _response_format(BaseModel): + game_over: bool + scene_description: str = Field(..., max_length=700) + dungeon_deepness: int + adventure_time: int + adventurer_status: _character_sheet + inventory_status: List[str] + + def __str__(self): + """Represent response as a string.""" + response_view = ( + f'{self.scene_description}' + f'\n\nInventory: {self.inventory_status}' + f'\n\nAdventurer: {self.adventurer_status}' + f'\n\nTime: {self.adventure_time}' + f'\n\nDeepness: {self.dungeon_deepness}' + f'\n\nGame Over: {self.game_over}') + return response_view + + +def set_description_limit(limit): # HBD: We modify the class definition in runtime. + """Update "_response_format" class to set a new "scene_description" max length.""" + _response_format.model_fields['scene_description'].metadata[0] = MaxLen(limit) + + +# Function definition. +def narrate(message, history, system_message, client=CLIENT, model=MODEL): + """Chat with the game engine.""" + messages = ([{"role": "system", "content": system_message}] + history + + [{"role": "user", "content": message}]) + response = client.chat.completions.parse(model=model, messages=messages, tools=tools, + response_format=_response_format) + # Process tool calls. + if response.choices[0].finish_reason == "tool_calls": + message = response.choices[0].message + tool_response = handle_tool_call(message) + messages.append(message) + messages.append(tool_response) + response = client.chat.completions.parse(model=model, messages=messages, + response_format=_response_format) + # Return game's Master response. + return response.choices[0].message.parsed diff --git a/community-contributions/dungeon_extraction_game/game/storyteller/tools.py b/community-contributions/dungeon_extraction_game/game/storyteller/tools.py new file mode 100644 index 0000000..a995ec0 --- /dev/null +++ b/community-contributions/dungeon_extraction_game/game/storyteller/tools.py @@ -0,0 +1,81 @@ +"""AI Mastered Dungeon Extraction Game storyteller tools module WIP.""" + +from json import loads + +from openai.types.chat import ChatCompletionMessage +from openai.types.chat import ChatCompletionMessageFunctionToolCall +from openai.types.chat.chat_completion_message_function_tool_call import Function + + +# Tools declaration for future use. (E.g. Tools may handle user status and inventory) +tools = [] + +tools_map = {} # This will map each tool with it's tool function. + + +# A tool call function. +def handle_tool_call(message: ChatCompletionMessage): + """Tools call handler.""" + tool_call = message.tool_calls[0] + arguments = loads(tool_call.function.arguments) + print(f'\nFUNC CALL: {tool_call.function.name}({arguments})\n') + # Get tool function and call with arguments. + tool_func = tools_map.get(tool_call.function.name) + tool_response = tool_func(**arguments) + response = {"role": "tool", "content": tool_response, "tool_call_id": tool_call.id} + return response + + +draw_signature = { + "name": "draw_scene", + "description": "Generate an image of the scene based on the description", + "parameters": { + "type": "object", + "properties": { + "scene_description": { + "type": "string", + "description": "A detailed description of the scene to be drawn", + }, + "scene_style": { + "type": "string", + "description": "The art style for the image", + }, + }, + "required": ["scene_description"], + "additionalProperties": False, + }, +} + + +# Tool call response example. +ChatCompletionMessage( + content="""To begin, first I need to set a scene. + Imagine you are in a dark room of an old castle. + The walls are covered in cobwebs and there is a smell of mold in the air. + As you look around, you notice a slightly ajar door to the north + and a dark figure lurking in the corner. + + I am going to generate an image of this scene. One moment, please.""", + refusal=None, + role="assistant", + annotations=[], + audio=None, + function_call=None, + tool_calls=[ + ChatCompletionMessageFunctionToolCall( + id="call_oJqJeXMUPZUaC0GPfMeSd16E", + function=Function( + arguments='''{ + "scene_description":"A dark room in an ancient castle. + The walls are covered with cobwebs, and there\'s a musty smell in + the air. + A slightly ajar door to the north and a shadowy figure lurking in + the corner. + Dim lighting adds to the eerie atmosphere, with flickering shadows.", + "style":"fantasy" + }''', + name="draw_scene"), + type="function", + ) + ], +) diff --git a/community-contributions/dungeon_extraction_game/images/advance_adventure.jpg b/community-contributions/dungeon_extraction_game/images/advance_adventure.jpg new file mode 100644 index 0000000..c1821d1 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/advance_adventure.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/chair.jpg b/community-contributions/dungeon_extraction_game/images/chair.jpg new file mode 100644 index 0000000..701daaa Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/chair.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/disabled.jpg b/community-contributions/dungeon_extraction_game/images/disabled.jpg new file mode 100644 index 0000000..2fac5f2 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/disabled.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/first_input.jpg b/community-contributions/dungeon_extraction_game/images/first_input.jpg new file mode 100644 index 0000000..c71a447 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/first_input.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/machine.jpg b/community-contributions/dungeon_extraction_game/images/machine.jpg new file mode 100644 index 0000000..1bbc849 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/machine.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/start_adventure.jpg b/community-contributions/dungeon_extraction_game/images/start_adventure.jpg new file mode 100644 index 0000000..4d280a2 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/start_adventure.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/start_input.jpg b/community-contributions/dungeon_extraction_game/images/start_input.jpg new file mode 100644 index 0000000..71c7438 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/start_input.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/start_view.jpg b/community-contributions/dungeon_extraction_game/images/start_view.jpg new file mode 100644 index 0000000..d277d1d Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/start_view.jpg differ diff --git a/community-contributions/dungeon_extraction_game/images/tragic_end.jpg b/community-contributions/dungeon_extraction_game/images/tragic_end.jpg new file mode 100644 index 0000000..5484ad6 Binary files /dev/null and b/community-contributions/dungeon_extraction_game/images/tragic_end.jpg differ diff --git a/community-contributions/fitness-nutrition-planner-agent/README.md b/community-contributions/fitness-nutrition-planner-agent/README.md new file mode 100644 index 0000000..a447792 --- /dev/null +++ b/community-contributions/fitness-nutrition-planner-agent/README.md @@ -0,0 +1,101 @@ + +# Fitness & Nutrition Planner Agent (Community Contribution) + +A tool-using agent that generates a **7‑day vegetarian-friendly meal plan** with **calorie/macro targets** and a **consolidated grocery list**. It supports **targeted swaps** (e.g., "swap Tuesday lunch") while honoring dietary patterns, allergies, and dislikes. + +> **Disclaimer**: This project is for educational purposes and is **not** medical advice. Consult a licensed professional for medical or specialized dietary needs. + +--- + +## ✨ Features +- Calculates **TDEE** and **macro targets** via Mifflin–St Jeor + activity factors. +- Builds a **7‑day plan** (breakfast/lunch/dinner) respecting dietary constraints. +- Produces an aggregated **grocery list** for the week. +- Supports **swap** of any single meal while keeping macros reasonable. +- Minimal **Streamlit UI** for demos. +- Extensible **tool-based architecture** to plug real recipe APIs/DBs. + +--- + +## 🧱 Architecture +- **Agent core**: OpenAI function-calling (tools) with a simple orchestration loop. +- **Tools**: + 1. `calc_calories_and_macros` – computes targets. + 2. `compose_meal_plan` – creates the 7‑day plan. + 3. `grocery_list_from_plan` – consolidates ingredients/quantities. + 4. `swap_meal` – replaces one meal (by kcal proximity and constraints). +- **Recipe source**: a tiny in-memory recipe DB for demo; replace with a real API or your own dataset. + +--- + +## 🚀 Quickstart + +### 1) Install +```bash +pip install openai streamlit pydantic python-dotenv +``` + +### 2) Configure +Create a `.env` file in this folder: +``` +OPENAI_API_KEY=your_key_here +OPENAI_MODEL=gpt-4o-mini +``` + +### 3) Run CLI (example) +```bash +python agent.py +``` + +### 4) Run UI +```bash +streamlit run app.py +``` + +--- + +## 🧪 Sample Profile (from issue author) +See `sample_profile.json` for the exact values used to produce `demo_output.md`. +- **Sex**: female +- **Age**: 45 +- **Height**: 152 cm (~5 ft) +- **Weight**: 62 kg +- **Activity**: light +- **Goal**: maintain +- **Diet**: vegetarian + +--- + +## 🔧 Extend +- Replace the in-memory recipes with: + - A real **recipe API** (e.g., Spoonacular) or + - Your **own dataset** (CSV/DB) + filters/tags +- Add price lookups to produce a **budget-aware** grocery list. +- Add **adherence tracking** and charts. +- Integrate **wearables** or daily steps to refine TDEE dynamically. +- Add **snacks** for days slightly under target kcals. + +--- + +## 🛡️ Safety Notes +- The agent warns for extreme deficits but does **not** diagnose conditions. +- For calorie targets below commonly recommended minimums (e.g., ~1200 kcal/day for many adults), advise consulting a professional. + +--- + +## 📁 Project Layout +``` +fitness-nutrition-planner-agent/ +├─ README.md +├─ agent.py +├─ app.py +├─ sample_profile.json +└─ demo_output.md +``` + +--- + +## 🤝 How to contribute +- Keep notebooks (if any) with **cleared outputs**. +- Follow the course repo’s contribution guidelines. +- Include screenshots or a short Loom/YT demo link in your PR description. diff --git a/community-contributions/fitness-nutrition-planner-agent/agent.py b/community-contributions/fitness-nutrition-planner-agent/agent.py new file mode 100644 index 0000000..75bcd10 --- /dev/null +++ b/community-contributions/fitness-nutrition-planner-agent/agent.py @@ -0,0 +1,411 @@ + +# agent.py +import os, math, json, copy +from dataclasses import dataclass +from typing import List, Dict, Any, Optional, Tuple +from pydantic import BaseModel, Field, ValidationError +from dotenv import load_dotenv +from openai import OpenAI + +load_dotenv() + +# ------------------------------ +# Data models +# ------------------------------ +class UserProfile(BaseModel): + sex: str = Field(..., description="male or female") + age: int + height_cm: float + weight_kg: float + activity_level: str = Field(..., description="sedentary, light, moderate, active, very_active") + goal: str = Field(..., description="lose, maintain, gain") + dietary_pattern: Optional[str] = Field(None, description="e.g., vegetarian, vegan, halal, kosher") + allergies: List[str] = Field(default_factory=list) + dislikes: List[str] = Field(default_factory=list) + daily_meals: int = 3 + cuisine_prefs: List[str] = Field(default_factory=list) + time_per_meal_minutes: int = 30 + budget_level: Optional[str] = Field(None, description="low, medium, high") + +class MacroTargets(BaseModel): + tdee: int + target_kcal: int + protein_g: int + carbs_g: int + fat_g: int + +class Meal(BaseModel): + name: str + ingredients: List[Dict[str, Any]] # {item, qty, unit} + kcal: int + protein_g: int + carbs_g: int + fat_g: int + tags: List[str] = Field(default_factory=list) + instructions: Optional[str] = None + +class DayPlan(BaseModel): + day: str + meals: List[Meal] + totals: MacroTargets + +class WeekPlan(BaseModel): + days: List[DayPlan] + meta: Dict[str, Any] + +# ------------------------------ +# Tiny in-memory recipe “DB” +# (extend/replace with a real source) +# ------------------------------ +RECIPE_DB: List[Meal] = [ + Meal( + name="Greek Yogurt Parfait", + ingredients=[{"item":"nonfat greek yogurt","qty":200,"unit":"g"}, + {"item":"berries","qty":150,"unit":"g"}, + {"item":"granola","qty":30,"unit":"g"}, + {"item":"honey","qty":10,"unit":"g"}], + kcal=380, protein_g=30, carbs_g=52, fat_g=8, + tags=["vegetarian","breakfast","5-min","no-cook"] + ), + Meal( + name="Tofu Veggie Stir-Fry with Rice", + ingredients=[{"item":"firm tofu","qty":150,"unit":"g"}, + {"item":"mixed vegetables","qty":200,"unit":"g"}, + {"item":"soy sauce (low sodium)","qty":15,"unit":"ml"}, + {"item":"olive oil","qty":10,"unit":"ml"}, + {"item":"brown rice (cooked)","qty":200,"unit":"g"}], + kcal=650, protein_g=28, carbs_g=85, fat_g=20, + tags=["vegan","gluten-free","dinner","20-min","stovetop","soy"] + ), + Meal( + name="Chicken Quinoa Bowl", + ingredients=[{"item":"chicken breast","qty":140,"unit":"g"}, + {"item":"quinoa (cooked)","qty":185,"unit":"g"}, + {"item":"spinach","qty":60,"unit":"g"}, + {"item":"olive oil","qty":10,"unit":"ml"}, + {"item":"lemon","qty":0.5,"unit":"unit"}], + kcal=620, protein_g=45, carbs_g=55, fat_g=20, + tags=["gluten-free","dinner","25-min","high-protein","poultry"] + ), + Meal( + name="Lentil Soup + Wholegrain Bread", + ingredients=[{"item":"lentils (cooked)","qty":200,"unit":"g"}, + {"item":"vegetable broth","qty":400,"unit":"ml"}, + {"item":"carrot","qty":80,"unit":"g"}, + {"item":"celery","qty":60,"unit":"g"}, + {"item":"onion","qty":60,"unit":"g"}, + {"item":"wholegrain bread","qty":60,"unit":"g"}], + kcal=520, protein_g=25, carbs_g=78, fat_g=8, + tags=["vegan","lunch","30-min","budget"] + ), + Meal( + name="Salmon, Potatoes & Greens", + ingredients=[{"item":"salmon fillet","qty":150,"unit":"g"}, + {"item":"potatoes","qty":200,"unit":"g"}, + {"item":"broccoli","qty":150,"unit":"g"}, + {"item":"olive oil","qty":10,"unit":"ml"}], + kcal=680, protein_g=42, carbs_g=52, fat_g=30, + tags=["gluten-free","dinner","omega-3","fish"] + ), + Meal( + name="Cottage Cheese Bowl", + ingredients=[{"item":"low-fat cottage cheese","qty":200,"unit":"g"}, + {"item":"pineapple","qty":150,"unit":"g"}, + {"item":"chia seeds","qty":15,"unit":"g"}], + kcal=380, protein_g=32, carbs_g=35, fat_g=10, + tags=["vegetarian","snack","5-min","high-protein","dairy"] + ), +] + +# ------------------------------ +# Tool implementations +# ------------------------------ +ACTIVITY_FACTORS = { + "sedentary": 1.2, + "light": 1.375, + "moderate": 1.55, + "active": 1.725, + "very_active": 1.9 +} + +def mifflin_st_jeor(weight_kg: float, height_cm: float, age: int, sex: str) -> float: + # BMR (kcal/day) + if sex.lower().startswith("m"): + return 10*weight_kg + 6.25*height_cm - 5*age + 5 + else: + return 10*weight_kg + 6.25*height_cm - 5*age - 161 + +def compute_targets(profile: UserProfile) -> MacroTargets: + bmr = mifflin_st_jeor(profile.weight_kg, profile.height_cm, profile.age, profile.sex) + tdee = int(round(bmr * ACTIVITY_FACTORS.get(profile.activity_level, 1.2))) + # goal adjustment + if profile.goal == "lose": + target_kcal = max(1200, int(tdee - 400)) # conservative deficit + elif profile.goal == "gain": + target_kcal = int(tdee + 300) + else: + target_kcal = tdee + + # Macro split (modifiable): P 30%, C 40%, F 30% + protein_kcal = target_kcal * 0.30 + carbs_kcal = target_kcal * 0.40 + fat_kcal = target_kcal * 0.30 + protein_g = int(round(protein_kcal / 4)) + carbs_g = int(round(carbs_kcal / 4)) + fat_g = int(round(fat_kcal / 9)) + + return MacroTargets(tdee=tdee, target_kcal=target_kcal, + protein_g=protein_g, carbs_g=carbs_g, fat_g=fat_g) + +def _allowed(meal: Meal, profile: UserProfile) -> bool: + # dietary patterns/allergies/dislikes filters (simple; extend as needed) + diet = (profile.dietary_pattern or "").lower() + if diet == "vegetarian" and ("fish" in meal.tags or "poultry" in meal.tags): + return False + if diet == "vegan" and ("dairy" in meal.tags or "fish" in meal.tags or "poultry" in meal.tags): + return False + # allergies & dislikes + for a in profile.allergies: + if a and a.lower() in meal.name.lower(): return False + if any(a.lower() in (ing["item"]).lower() for ing in meal.ingredients): return False + if a.lower() in " ".join(meal.tags).lower(): return False + for d in profile.dislikes: + if d and d.lower() in meal.name.lower(): return False + if any(d.lower() in (ing["item"]).lower() for ing in meal.ingredients): return False + return True + +def meal_db_search(profile: UserProfile, tags: Optional[List[str]] = None) -> List[Meal]: + tags = tags or [] + out = [] + for m in RECIPE_DB: + if not _allowed(m, profile): + continue + if tags and not any(t in m.tags for t in tags): + continue + out.append(m) + return out or [] # may be empty; agent should handle + +def compose_meal_plan(profile: UserProfile, targets: MacroTargets) -> WeekPlan: + # naive heuristic: pick meals that roughly match per-meal macro budget + per_meal_kcal = targets.target_kcal / profile.daily_meals + days = [] + weekdays = ["Mon","Tue","Wed","Thu","Fri","Sat","Sun"] + + # simple pools + breakfasts = meal_db_search(profile, tags=["breakfast","no-cook","5-min"]) + lunches = meal_db_search(profile, tags=["lunch","budget"]) + dinners = meal_db_search(profile, tags=["dinner","high-protein"]) + + # fallback to any allowed meals if pools too small + allowed_all = meal_db_search(profile) + if len(breakfasts) < 2: breakfasts = allowed_all + if len(lunches) < 2: lunches = allowed_all + if len(dinners) < 2: dinners = allowed_all + + for i, day in enumerate(weekdays): + day_meals = [] + for slot in range(profile.daily_meals): + pool = breakfasts if slot == 0 else (lunches if slot == 1 else dinners) + # pick the meal closest in kcal to per_meal_kcal + pick = min(pool, key=lambda m: abs(m.kcal - per_meal_kcal)) + day_meals.append(copy.deepcopy(pick)) + # compute totals + kcal = sum(m.kcal for m in day_meals) + p = sum(m.protein_g for m in day_meals) + c = sum(m.carbs_g for m in day_meals) + f = sum(m.fat_g for m in day_meals) + day_targets = MacroTargets(tdee=targets.tdee, target_kcal=int(round(kcal)), + protein_g=p, carbs_g=c, fat_g=f) + days.append(DayPlan(day=day, meals=day_meals, totals=day_targets)) + return WeekPlan(days=days, meta={"per_meal_target_kcal": int(round(per_meal_kcal))}) + +def grocery_list_from_plan(plan: WeekPlan) -> List[Dict[str, Any]]: + # aggregate identical ingredients + agg: Dict[Tuple[str,str], float] = {} + units: Dict[Tuple[str,str], str] = {} + for d in plan.days: + for m in d.meals: + for ing in m.ingredients: + key = (ing["item"].lower(), ing.get("unit","")) + agg[key] = agg.get(key, 0) + float(ing.get("qty", 0)) + units[key] = ing.get("unit","") + items = [] + for (item, unit), qty in sorted(agg.items()): + items.append({"item": item, "qty": round(qty, 2), "unit": unit}) + return items + +def swap_meal(plan: WeekPlan, day: str, meal_index: int, profile: UserProfile) -> WeekPlan: + # replace one meal by closest-kcal allowed alternative that isn't the same + day_idx = next((i for i,d in enumerate(plan.days) if d.day.lower().startswith(day[:3].lower())), None) + if day_idx is None: return plan + current_meal = plan.days[day_idx].meals[meal_index] + candidates = [m for m in meal_db_search(profile) if m.name != current_meal.name] + if not candidates: return plan + pick = min(candidates, key=lambda m: abs(m.kcal - current_meal.kcal)) + plan.days[day_idx].meals[meal_index] = copy.deepcopy(pick) + # recalc day totals + d = plan.days[day_idx] + kcal = sum(m.kcal for m in d.meals) + p = sum(m.protein_g for m in d.meals) + c = sum(m.carbs_g for m in d.meals) + f = sum(m.fat_g for m in d.meals) + d.totals = MacroTargets(tdee=d.totals.tdee, target_kcal=kcal, protein_g=p, carbs_g=c, fat_g=f) + return plan + +# ------------------------------ +# Agent (LLM + tools) +# ------------------------------ +SYS_PROMPT = """You are FitnessPlanner, an agentic planner that: +- Respects dietary patterns, allergies, dislikes, budget, time limits. +- Uses tools to compute targets, assemble a 7-day plan, produce a grocery list, and swap meals on request. +- If a request is unsafe (extreme deficits, medical conditions), warn and suggest professional guidance. +- Keep responses concise and structured (headings + bullet lists).""" + +# Tool registry for function-calling +def get_tools_schema(): + return [ + { + "type": "function", + "function": { + "name": "calc_calories_and_macros", + "description": "Compute TDEE and macro targets from the user's profile.", + "parameters": { + "type":"object", + "properties": {"profile":{"type":"object"}}, + "required":["profile"] + } + } + }, + { + "type": "function", + "function": { + "name": "compose_meal_plan", + "description": "Create a 7-day meal plan matching targets and constraints.", + "parameters": { + "type":"object", + "properties": { + "profile":{"type":"object"}, + "targets":{"type":"object"} + }, + "required":["profile","targets"] + } + } + }, + { + "type": "function", + "function": { + "name": "grocery_list_from_plan", + "description": "Make a consolidated grocery list from a week plan.", + "parameters": { + "type":"object", + "properties": {"plan":{"type":"object"}}, + "required":["plan"] + } + } + }, + { + "type": "function", + "function": { + "name": "swap_meal", + "description": "Swap a single meal in the plan while keeping macros reasonable.", + "parameters": { + "type":"object", + "properties": { + "plan":{"type":"object"}, + "day":{"type":"string"}, + "meal_index":{"type":"integer","description":"0=breakfast,1=lunch,2=dinner"}, + "profile":{"type":"object"} + }, + "required":["plan","day","meal_index","profile"] + } + } + } + ] + +class FitnessPlannerAgent: + def __init__(self, model: Optional[str] = None): + self.client = OpenAI(api_key=os.getenv("OPENAI_API_KEY")) + self.model = model or os.getenv("OPENAI_MODEL", "gpt-4o-mini") + self.plan_cache: Optional[WeekPlan] = None + self.targets_cache: Optional[MacroTargets] = None + + # Tool dispatch + def _call_tool(self, name: str, args: Dict[str, Any]) -> str: + if name == "calc_calories_and_macros": + profile = UserProfile(**args["profile"]) + targets = compute_targets(profile) + self.targets_cache = targets + return targets.model_dump_json() + elif name == "compose_meal_plan": + profile = UserProfile(**args["profile"]) + targets = MacroTargets(**args["targets"]) + plan = compose_meal_plan(profile, targets) + self.plan_cache = plan + return plan.model_dump_json() + elif name == "grocery_list_from_plan": + plan = WeekPlan(**args["plan"]) + items = grocery_list_from_plan(plan) + return json.dumps(items) + elif name == "swap_meal": + plan = WeekPlan(**args["plan"]) + profile = UserProfile(**args["profile"]) + day = args["day"] + idx = args["meal_index"] + new_plan = swap_meal(plan, day, idx, profile) + self.plan_cache = new_plan + return new_plan.model_dump_json() + else: + return json.dumps({"error":"unknown tool"}) + + def chat(self, user_message: str, profile: Optional[UserProfile] = None) -> str: + messages = [{"role":"system","content":SYS_PROMPT}] + if profile: + messages.append({"role":"user","content":f"User profile: {profile.model_dump_json()}"} ) + messages.append({"role":"user","content":user_message}) + + # First call + resp = self.client.chat.completions.create( + model=self.model, + messages=messages, + tools=get_tools_schema(), + tool_choice="auto", + temperature=0.3 + ) + + # Handle tool calls (simple, single-step or brief multi-step) + messages_llm = messages + [{"role":"assistant","content":resp.choices[0].message.content or "", + "tool_calls":resp.choices[0].message.tool_calls}] + if resp.choices[0].message.tool_calls: + for tc in resp.choices[0].message.tool_calls: + name = tc.function.name + args = json.loads(tc.function.arguments or "{}") + out = self._call_tool(name, args) + messages_llm.append({ + "role":"tool", + "tool_call_id":tc.id, + "name":name, + "content":out + }) + + # Finalization + resp2 = self.client.chat.completions.create( + model=self.model, + messages=messages_llm, + temperature=0.2 + ) + return resp2.choices[0].message.content + + return resp.choices[0].message.content + +# ------------------------------ +# Quick CLI demo +# ------------------------------ +if __name__ == "__main__": + profile = UserProfile( + sex="female", age=45, height_cm=152, weight_kg=62, + activity_level="light", goal="maintain", + dietary_pattern="vegetarian", allergies=[], dislikes=[], + daily_meals=3, cuisine_prefs=["mediterranean"], time_per_meal_minutes=25, budget_level="medium" + ) + agent = FitnessPlannerAgent() + print(agent.chat("Create my 7-day plan and grocery list.", profile)) diff --git a/community-contributions/fitness-nutrition-planner-agent/app.py b/community-contributions/fitness-nutrition-planner-agent/app.py new file mode 100644 index 0000000..a1f1102 --- /dev/null +++ b/community-contributions/fitness-nutrition-planner-agent/app.py @@ -0,0 +1,75 @@ + +# app.py +import json +import streamlit as st +from agent import FitnessPlannerAgent, UserProfile, WeekPlan + +st.set_page_config(page_title="Fitness & Nutrition Planner Agent", layout="wide") + +st.title("🏋️ Fitness & Nutrition Planner Agent") + +with st.sidebar: + st.header("Your Profile") + sex = st.selectbox("Sex", ["female","male"]) + age = st.number_input("Age", 18, 90, 45) + height_cm = st.number_input("Height (cm)", 120, 220, 152) + weight_kg = st.number_input("Weight (kg)", 35.0, 200.0, 62.0) + activity_level = st.selectbox("Activity Level", ["sedentary","light","moderate","active","very_active"], index=1) + goal = st.selectbox("Goal", ["lose","maintain","gain"], index=1) + dietary_pattern = st.selectbox("Dietary Pattern", ["none","vegetarian","vegan","halal","kosher"], index=1) + if dietary_pattern == "none": dietary_pattern = None + allergies = st.text_input("Allergies (comma-separated)", "") + dislikes = st.text_input("Dislikes (comma-separated)", "") + daily_meals = st.slider("Meals per day", 2, 5, 3) + time_per_meal_minutes = st.slider("Time per meal (min)", 5, 90, 25) + budget_level = st.selectbox("Budget", ["medium","low","high"], index=0) + cuisine_prefs = st.text_input("Cuisine prefs (comma-separated)", "mediterranean") + + build_btn = st.button("Generate 7-Day Plan") + +agent = FitnessPlannerAgent() + +if build_btn: + profile = UserProfile( + sex=sex, age=int(age), height_cm=float(height_cm), weight_kg=float(weight_kg), + activity_level=activity_level, goal=goal, dietary_pattern=dietary_pattern, + allergies=[a.strip() for a in allergies.split(",") if a.strip()], + dislikes=[d.strip() for d in dislikes.split(",") if d.strip()], + daily_meals=int(daily_meals), cuisine_prefs=[c.strip() for c in cuisine_prefs.split(",") if c.strip()], + time_per_meal_minutes=int(time_per_meal_minutes), budget_level=budget_level + ) + st.session_state["profile_json"] = profile.model_dump_json() + with st.spinner("Planning your week..."): + result = agent.chat("Create my 7-day plan and grocery list.", profile) + st.session_state["last_response"] = result + +if "last_response" in st.session_state: + st.subheader("Plan & Groceries") + st.markdown(st.session_state["last_response"]) + +st.divider() +st.subheader("Tweaks") +col1, col2, col3 = st.columns(3) +with col1: + day = st.selectbox("Day to change", ["Mon","Tue","Wed","Thu","Fri","Sat","Sun"]) +with col2: + meal_index = st.selectbox("Meal slot", ["Breakfast (0)","Lunch (1)","Dinner (2)"]) + meal_index = int(meal_index[-2]) # 0/1/2 +with col3: + swap_btn = st.button("Swap Meal") + +if swap_btn and agent.plan_cache: + profile_json = st.session_state.get("profile_json") + if not profile_json: + st.warning("Please generate a plan first.") + else: + new_plan_json = agent._call_tool("swap_meal", { + "plan": agent.plan_cache.model_dump(), + "day": day, + "meal_index": meal_index, + "profile": json.loads(profile_json) + }) + agent.plan_cache = WeekPlan(**json.loads(new_plan_json)) + summary = agent.chat(f"Update summary for {day}: show the swapped meal and new day totals.") + st.session_state["last_response"] = summary + st.markdown(summary) diff --git a/community-contributions/fitness-nutrition-planner-agent/demo_output.md b/community-contributions/fitness-nutrition-planner-agent/demo_output.md new file mode 100644 index 0000000..841fb1e --- /dev/null +++ b/community-contributions/fitness-nutrition-planner-agent/demo_output.md @@ -0,0 +1,84 @@ + +# Demo Output (Sample Profile) + +**Profile**: female, 45, 152 cm, 62 kg, activity: light, goal: maintain, diet: vegetarian + +## Targets +- TDEE ≈ **1680 kcal/day** +- Macros (30/40/30): **Protein 126 g**, **Carbs 168 g**, **Fat 56 g** + +> These are estimates using Mifflin–St Jeor and a light activity factor. Not medical advice. + +--- + +## Example 7-Day Plan (Breakfast / Lunch / Dinner) + +**Mon** +- Greek Yogurt Parfait (380 kcal, 30P/52C/8F) +- Lentil Soup + Wholegrain Bread (520 kcal, 25P/78C/8F) +- Tofu Veggie Stir-Fry with Rice (650 kcal, 28P/85C/20F) +- **Totals** ≈ 1550 kcal, 83P, 215C, 36F + +**Tue** +- Cottage Cheese Bowl (380 kcal, 32P/35C/10F) +- Lentil Soup + Wholegrain Bread (520 kcal, 25P/78C/8F) +- Tofu Veggie Stir-Fry with Rice (650 kcal, 28P/85C/20F) +- **Totals** ≈ 1550 kcal, 85P, 198C, 38F + +**Wed** +- Greek Yogurt Parfait +- Lentil Soup + Wholegrain Bread +- Tofu Veggie Stir-Fry with Rice +- **Totals** ≈ 1550 kcal + +**Thu** +- Cottage Cheese Bowl +- Lentil Soup + Wholegrain Bread +- Tofu Veggie Stir-Fry with Rice +- **Totals** ≈ 1550 kcal + +**Fri** +- Greek Yogurt Parfait +- Lentil Soup + Wholegrain Bread +- Tofu Veggie Stir-Fry with Rice +- **Totals** ≈ 1550 kcal + +**Sat** +- Cottage Cheese Bowl +- Lentil Soup + Wholegrain Bread +- Tofu Veggie Stir-Fry with Rice +- **Totals** ≈ 1550 kcal + +**Sun** +- Greek Yogurt Parfait +- Lentil Soup + Wholegrain Bread +- Tofu Veggie Stir-Fry with Rice +- **Totals** ≈ 1550 kcal + +> Notes: The demo DB is intentionally small. In practice, plug in a larger vegetarian recipe set for more variety. Add snacks if you'd like to reach ~1680 kcal/day. + +--- + +## Grocery List (aggregated, approx for 7 days) + +- nonfat greek yogurt — **1400 g** +- berries — **1050 g** +- granola — **210 g** +- honey — **70 g** +- lentils (cooked) — **1400 g** +- vegetable broth — **2800 ml** +- carrot — **560 g** +- celery — **420 g** +- onion — **420 g** +- wholegrain bread — **420 g** +- firm tofu — **1050 g** +- mixed vegetables — **1400 g** +- soy sauce (low sodium) — **105 ml** +- olive oil — **140 ml** +- brown rice (cooked) — **1400 g** +- low-fat cottage cheese — **600 g** +- pineapple — **450 g** +- chia seeds — **45 g** + +**Tip:** Use the app’s *Swap Meal* to replace any item (e.g., swap Wed dinner). + diff --git a/community-contributions/fitness-nutrition-planner-agent/sample_profile.json b/community-contributions/fitness-nutrition-planner-agent/sample_profile.json new file mode 100644 index 0000000..0f54a9d --- /dev/null +++ b/community-contributions/fitness-nutrition-planner-agent/sample_profile.json @@ -0,0 +1,17 @@ +{ + "sex": "female", + "age": 45, + "height_cm": 152, + "weight_kg": 62, + "activity_level": "light", + "goal": "maintain", + "dietary_pattern": "vegetarian", + "allergies": [], + "dislikes": [], + "daily_meals": 3, + "cuisine_prefs": [ + "mediterranean" + ], + "time_per_meal_minutes": 25, + "budget_level": "medium" +} \ No newline at end of file diff --git a/community-contributions/kaiyuwei/week1/day2 EXERCISE.ipynb b/community-contributions/kaiyuwei/week1/day2 EXERCISE.ipynb new file mode 100644 index 0000000..1881401 --- /dev/null +++ b/community-contributions/kaiyuwei/week1/day2 EXERCISE.ipynb @@ -0,0 +1,428 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# Welcome to your first assignment!\n", + "\n", + "Instructions are below. Please give this a try, and look in the solutions folder if you get stuck (or feel free to ask me!)" + ] + }, + { + "cell_type": "markdown", + "id": "ada885d9-4d42-4d9b-97f0-74fbbbfe93a9", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Just before we get to the assignment --

\n", + " I thought I'd take a second to point you at this page of useful resources for the course. This includes links to all the slides.
\n", + " https://edwarddonner.com/2024/11/13/llm-engineering-resources/
\n", + " Please keep this bookmarked, and I'll continue to add more useful links there over time.\n", + " \n", + "
" + ] + }, + { + "cell_type": "markdown", + "id": "6e9fa1fc-eac5-4d1d-9be4-541b3f2b3458", + "metadata": {}, + "source": [ + "# HOMEWORK EXERCISE ASSIGNMENT\n", + "\n", + "Upgrade the day 1 project to summarize a webpage to use an Open Source model running locally via Ollama rather than OpenAI\n", + "\n", + "You'll be able to use this technique for all subsequent projects if you'd prefer not to use paid APIs.\n", + "\n", + "**Benefits:**\n", + "1. No API charges - open-source\n", + "2. Data doesn't leave your box\n", + "\n", + "**Disadvantages:**\n", + "1. Significantly less power than Frontier Model\n", + "\n", + "## Recap on installation of Ollama\n", + "\n", + "Simply visit [ollama.com](https://ollama.com) and install!\n", + "\n", + "Once complete, the ollama server should already be running locally. \n", + "If you visit: \n", + "[http://localhost:11434/](http://localhost:11434/)\n", + "\n", + "You should see the message `Ollama is running`. \n", + "\n", + "If not, bring up a new Terminal (Mac) or Powershell (Windows) and enter `ollama serve` \n", + "And in another Terminal (Mac) or Powershell (Windows), enter `ollama pull llama3.2` \n", + "Then try [http://localhost:11434/](http://localhost:11434/) again.\n", + "\n", + "If Ollama is slow on your machine, try using `llama3.2:1b` as an alternative. Run `ollama pull llama3.2:1b` from a Terminal or Powershell, and change the code below from `MODEL = \"llama3.2\"` to `MODEL = \"llama3.2:1b\"`" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dac0a679-599c-441f-9bf2-ddc73d35b940", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a messages list using the same format that we used for OpenAI\n", + "\n", + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Describe some of the business applications of Generative AI\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7bb9c624-14f0-4945-a719-8ddb64f66f47", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "479ff514-e8bd-4985-a572-2ea28bb4fa40", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's just make sure the model is loaded\n", + "\n", + "!ollama pull llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "42b9f644-522d-4e05-a691-56e7658c0ea9", + "metadata": {}, + "outputs": [], + "source": [ + "# If this doesn't work for any reason, try the 2 versions in the following cells\n", + "# And double check the instructions in the 'Recap on installation of Ollama' at the top of this lab\n", + "# And if none of that works - contact me!\n", + "\n", + "response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "6a021f13-d6a1-4b96-8e18-4eae49d876fe", + "metadata": {}, + "source": [ + "# Introducing the ollama package\n", + "\n", + "And now we'll do the same thing, but using the elegant ollama python package instead of a direct HTTP call.\n", + "\n", + "Under the hood, it's making the same call as above to the ollama server running at localhost:11434" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7745b9c4-57dc-4867-9180-61fa5db55eb8", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "a4704e10-f5fb-4c15-a935-f046c06fb13d", + "metadata": {}, + "source": [ + "## Alternative approach - using OpenAI python library to connect to Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [], + "source": [ + "# There's actually an alternative approach that some people might prefer\n", + "# You can use the OpenAI client python library to call Ollama:\n", + "\n", + "from openai import OpenAI\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "9f9e22da-b891-41f6-9ac9-bd0c0a5f4f44", + "metadata": { + "jp-MarkdownHeadingCollapsed": true + }, + "source": [ + "## Are you confused about why that works?\n", + "\n", + "It seems strange, right? We just used OpenAI code to call Ollama?? What's going on?!\n", + "\n", + "Here's the scoop:\n", + "\n", + "The python class `OpenAI` is simply code written by OpenAI engineers that makes calls over the internet to an endpoint. \n", + "\n", + "When you call `openai.chat.completions.create()`, this python code just makes a web request to the following url: \"https://api.openai.com/v1/chat/completions\"\n", + "\n", + "Code like this is known as a \"client library\" - it's just wrapper code that runs on your machine to make web requests. The actual power of GPT is running on OpenAI's cloud behind this API, not on your computer!\n", + "\n", + "OpenAI was so popular, that lots of other AI providers provided identical web endpoints, so you could use the same approach.\n", + "\n", + "So Ollama has an endpoint running on your local box at http://localhost:11434/v1/chat/completions \n", + "And in week 2 we'll discover that lots of other providers do this too, including Gemini and DeepSeek.\n", + "\n", + "And then the team at OpenAI had a great idea: they can extend their client library so you can specify a different 'base url', and use their library to call any compatible API.\n", + "\n", + "That's it!\n", + "\n", + "So when you say: `ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')` \n", + "Then this will make the same endpoint calls, but to Ollama instead of OpenAI." + ] + }, + { + "cell_type": "markdown", + "id": "bc7d1de3-e2ac-46ff-a302-3b4ba38c4c90", + "metadata": {}, + "source": [ + "## Also trying the amazing reasoning model DeepSeek\n", + "\n", + "Here we use the version of DeepSeek-reasoner that's been distilled to 1.5B. \n", + "This is actually a 1.5B variant of Qwen that has been fine-tuned using synethic data generated by Deepseek R1.\n", + "\n", + "Other sizes of DeepSeek are [here](https://ollama.com/library/deepseek-r1) all the way up to the full 671B parameter version, which would use up 404GB of your drive and is far too large for most!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cf9eb44e-fe5b-47aa-b719-0bb63669ab3d", + "metadata": {}, + "outputs": [], + "source": [ + "!ollama pull deepseek-r1:1.5b" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1d3d554b-e00d-4c08-9300-45e073950a76", + "metadata": {}, + "outputs": [], + "source": [ + "# This may take a few minutes to run! You should then see a fascinating \"thinking\" trace inside tags, followed by some decent definitions\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=\"deepseek-r1:1.5b\",\n", + " messages=[{\"role\": \"user\", \"content\": \"Please give definitions of some core concepts behind LLMs: a neural network, attention and the transformer\"}]\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "1622d9bb-5c68-4d4e-9ca4-b492c751f898", + "metadata": {}, + "source": [ + "# NOW the exercise for you\n", + "\n", + "Take the code from day1 and incorporate it here, to build a website summarizer that uses Llama 3.2 running locally instead of OpenAI; use either of the above approaches." + ] + }, + { + "cell_type": "markdown", + "id": "59e9564e", + "metadata": {}, + "source": [ + "1. Create a Website class" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [ + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "import requests\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "d44c6179-5c00-46a1-a068-6c6226307e2f", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"user-agent\": \"mozilla/5.0 (windows nt 10.0; win64; x64) applewebkit/537.36 (khtml, like gecko) chrome/117.0.0.0 safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers = headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title if soup.title else \"No title found for this website\"\n", + " \n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "e9edb034", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"\n", + "\n", + "def getUserPrompt(website):\n", + " userPrompt = f\"You are looking at a website titled {website.title}\"\n", + " userPrompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " userPrompt += website.text\n", + " return userPrompt" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "8581edb1", + "metadata": {}, + "outputs": [], + "source": [ + "def getPromptMessageFor(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": getUserPrompt(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "caf20d99", + "metadata": {}, + "outputs": [], + "source": [ + "def summarize(url):\n", + " website = Website(url)\n", + " ollamaAi = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + " response = ollamaAi.chat.completions.create(\n", + " model = \"deepseek-r1:1.5b\",\n", + " messages = getPromptMessageFor(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "5e0ed89f", + "metadata": {}, + "outputs": [], + "source": [ + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e5793933", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/openai-twenty-questions/README.md b/community-contributions/openai-twenty-questions/README.md new file mode 100644 index 0000000..3fcbd94 --- /dev/null +++ b/community-contributions/openai-twenty-questions/README.md @@ -0,0 +1,5 @@ +# openai-twenty-questions +Chatgpt script that acts as Thinker and Guesser to play the 20 questions game + +1. pip install openai==0.28 +2. Run python twenty.py and it will Think of a word and try to guess it in 20 tries diff --git a/community-contributions/openai-twenty-questions/twenty.py b/community-contributions/openai-twenty-questions/twenty.py new file mode 100644 index 0000000..5f098cd --- /dev/null +++ b/community-contributions/openai-twenty-questions/twenty.py @@ -0,0 +1,100 @@ +import openai +import os +import time + +# openai.api_key = os.getenv("OPENAI_API_KEY") +# openai.api_key = "<>" + +# Models: You can use "gpt-4o", "gpt-4-turbo", or "gpt-3.5-turbo" — but we'll use "gpt-4o" or "gpt-4o-mini" for both players +MODEL = "gpt-4o-mini" + +def call_chatgpt(messages): + response = openai.ChatCompletion.create( + model=MODEL, + messages=messages, + temperature=0.7 + ) + return response.choices[0].message["content"].strip() + +# Step 1: Thinker chooses a secret object +thinker_messages = [ + {"role": "system", "content": "You are playing 20 Questions. Think of an object or thing and just one word. Keep it secret and reply only with: 'I have thought of something. Let's begin.'"}, +] +thinker_reply = call_chatgpt(thinker_messages) +print("Thinker:", thinker_reply) + +# For simulation purposes, let’s ask the thinker what the object is (in real game, this is hidden) +reveal_object_prompt = [ + {"role": "system", "content": "You are playing 20 Questions. Think of an object or thing and just one word. Now tell me (just for logging) what you are thinking of. Reply only with the thing."} +] +object_answer = call_chatgpt(reveal_object_prompt) +print("🔒 Secret Object:", object_answer) + +# Step 2: Guesser starts asking questions +guesser_messages = [ + {"role": "system", "content": f"You are playing 20 Questions. Ask yes/no questions to figure out what the object is. Do not repeat questions. The object is kept secret by the other player. Begin by asking your first question."}, +] + +# Let’s keep track of Q&A +history = [] +q_count = 1 + +for i in range(1, 11): + print(f"\n🔄 Round {q_count}") + q_count += 1 + # Guesser asks a question + question = call_chatgpt(guesser_messages) + print("Guesser:", question) + history.append(("Guesser", question)) + + # Thinker responds (yes/no) + thinker_round = [ + {"role": "system", "content": f"You are playing 20 Questions. The secret object is: {object_answer}."}, + {"role": "user", "content": f"The other player asked: {question}. Respond only with 'Yes', 'No', or 'I don't know'."} + ] + answer = call_chatgpt(thinker_round) + print("Thinker:", answer) + history.append(("Thinker", answer)) + + # Add to conversation history for guesser + guesser_messages.append({"role": "assistant", "content": question}) + guesser_messages.append({"role": "user", "content": answer}) + + + print(f"\n🔄 Round {q_count}") + q_count += 1 + # Check if guesser wants to guess + guess_check_prompt = guesser_messages + [ + {"role": "user", "content": "Based on the answers so far, do you want to guess? If yes, say: 'Is it ?'. If not, ask the next yes/no question."} + ] + next_move_question = call_chatgpt(guess_check_prompt) + print("Guesser next move:", next_move_question) + history.append(("Guesser", next_move_question)) + + if next_move_question.lower().startswith("is it a"): + # Thinker validates guess + guess = next_move_question[8:].strip(" ?.") + guess = next_move_question[8:].strip(" ?") + + if guess.lower() == object_answer.lower(): + print("Guesser guessed correctly!") + break + # Thinker responds (yes/no) + thinker_round = [ + {"role": "system", "content": f"You are playing 20 Questions. The secret object is: {object_answer}."}, + {"role": "user", "content": f"The other player asked: {next_move_question}. Respond only with 'Yes', 'No', or 'I don't know'."} + ] + answer = call_chatgpt(thinker_round) + print("Thinker next move:", answer) + history.append(("Thinker", answer)) + + # Add to conversation history for guesser + guesser_messages.append({"role": "assistant", "content": next_move_question}) + guesser_messages.append({"role": "user", "content": answer}) + + # Prepare for next round + guesser_messages.append({"role": "assistant", "content": next_move_question}) + question = next_move_question + +else: + print("❌ Guesser used all 20 questions without guessing correctly.") \ No newline at end of file diff --git a/community-contributions/playwright-enhanced-scraper/enhanced_web_scraper.ipynb b/community-contributions/playwright-enhanced-scraper/enhanced_web_scraper.ipynb new file mode 100644 index 0000000..8e7baf6 --- /dev/null +++ b/community-contributions/playwright-enhanced-scraper/enhanced_web_scraper.ipynb @@ -0,0 +1,731 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# YOUR FIRST LAB\n", + "### Please read this section. This is valuable to get you prepared, even if it's a long read -- it's important stuff.\n", + "\n", + "## Your first Frontier LLM Project\n", + "\n", + "Let's build a useful LLM solution - in a matter of minutes.\n", + "\n", + "By the end of this course, you will have built an autonomous Agentic AI solution with 7 agents that collaborate to solve a business problem. All in good time! We will start with something smaller...\n", + "\n", + "Our goal is to code a new kind of Web Browser. Give it a URL, and it will respond with a summary. The Reader's Digest of the internet!!\n", + "\n", + "Before starting, you should have completed the setup for [PC](../SETUP-PC.md) or [Mac](../SETUP-mac.md) and you hopefully launched this jupyter lab from within the project root directory, with your environment activated.\n", + "\n", + "## If you're new to Jupyter Lab\n", + "\n", + "Welcome to the wonderful world of Data Science experimentation! Once you've used Jupyter Lab, you'll wonder how you ever lived without it. Simply click in each \"cell\" with code in it, such as the cell immediately below this text, and hit Shift+Return to execute that cell. As you wish, you can add a cell with the + button in the toolbar, and print values of variables, or try out variations. \n", + "\n", + "I've written a notebook called [Guide to Jupyter](Guide%20to%20Jupyter.ipynb) to help you get more familiar with Jupyter Labs, including adding Markdown comments, using `!` to run shell commands, and `tqdm` to show progress.\n", + "\n", + "## If you're new to the Command Line\n", + "\n", + "Please see these excellent guides: [Command line on PC](https://chatgpt.com/share/67b0acea-ba38-8012-9c34-7a2541052665) and [Command line on Mac](https://chatgpt.com/canvas/shared/67b0b10c93a081918210723867525d2b). \n", + "\n", + "## If you'd prefer to work in IDEs\n", + "\n", + "If you're more comfortable in IDEs like VSCode, Cursor or PyCharm, they both work great with these lab notebooks too. \n", + "If you'd prefer to work in VSCode, [here](https://chatgpt.com/share/676f2e19-c228-8012-9911-6ca42f8ed766) are instructions from an AI friend on how to configure it for the course.\n", + "\n", + "## If you'd like to brush up your Python\n", + "\n", + "I've added a notebook called [Intermediate Python](Intermediate%20Python.ipynb) to get you up to speed. But you should give it a miss if you already have a good idea what this code does: \n", + "`yield from {book.get(\"author\") for book in books if book.get(\"author\")}`\n", + "\n", + "## I am here to help\n", + "\n", + "If you have any problems at all, please do reach out. \n", + "I'm available through the platform, or at ed@edwarddonner.com, or at https://www.linkedin.com/in/eddonner/ if you'd like to connect (and I love connecting!) \n", + "And this is new to me, but I'm also trying out X/Twitter at [@edwarddonner](https://x.com/edwarddonner) - if you're on X, please show me how it's done 😂 \n", + "\n", + "## More troubleshooting\n", + "\n", + "Please see the [troubleshooting](troubleshooting.ipynb) notebook in this folder to diagnose and fix common problems. At the very end of it is a diagnostics script with some useful debug info.\n", + "\n", + "## For foundational technical knowledge (eg Git, APIs, debugging) \n", + "\n", + "If you're relatively new to programming -- I've got your back! While it's ideal to have some programming experience for this course, there's only one mandatory prerequisite: plenty of patience. 😁 I've put together a set of self-study guides that cover Git and GitHub, APIs and endpoints, beginner python and more.\n", + "\n", + "This covers Git and GitHub; what they are, the difference, and how to use them: \n", + "https://github.com/ed-donner/agents/blob/main/guides/03_git_and_github.ipynb\n", + "\n", + "This covers technical foundations: \n", + "ChatGPT vs API; taking screenshots; Environment Variables; Networking basics; APIs and endpoints: \n", + "https://github.com/ed-donner/agents/blob/main/guides/04_technical_foundations.ipynb\n", + "\n", + "This covers Python for beginners, and making sure that a `NameError` never trips you up: \n", + "https://github.com/ed-donner/agents/blob/main/guides/06_python_foundations.ipynb\n", + "\n", + "This covers the essential techniques for figuring out errors: \n", + "https://github.com/ed-donner/agents/blob/main/guides/08_debugging.ipynb\n", + "\n", + "And you'll find other useful guides in the same folder in GitHub. Some information applies to my other Udemy course (eg Async Python) but most of it is very relevant for LLM engineering.\n", + "\n", + "## If this is old hat!\n", + "\n", + "If you're already comfortable with today's material, please hang in there; you can move swiftly through the first few labs - we will get much more in depth as the weeks progress. Ultimately we will fine-tune our own LLM to compete with OpenAI!\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Please read - important note

\n", + " The way I collaborate with you may be different to other courses you've taken. I prefer not to type code while you watch. Rather, I execute Jupyter Labs, like this, and give you an intuition for what's going on. My suggestion is that you carefully execute this yourself, after watching the lecture. Add print statements to understand what's going on, and then come up with your own variations. If you have a Github account, use this to showcase your variations. Not only is this essential practice, but it demonstrates your skills to others, including perhaps future clients or employers...\n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

This code is a live resource - keep an eye out for my emails

\n", + " I push updates to the code regularly. As people ask questions, I add more examples or improved commentary. As a result, you'll notice that the code below isn't identical to the videos. Everything from the videos is here; but I've also added better explanations and new models like DeepSeek. Consider this like an interactive book.

\n", + " I try to send emails regularly with important updates related to the course. You can find this in the 'Announcements' section of Udemy in the left sidebar. You can also choose to receive my emails via your Notification Settings in Udemy. I'm respectful of your inbox and always try to add value with my emails!\n", + " \n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business value of these exercises

\n", + " A final thought. While I've designed these notebooks to be educational, I've also tried to make them enjoyable. We'll do fun things like have LLMs tell jokes and argue with each other. But fundamentally, my goal is to teach skills you can apply in business. I'll explain business implications as we go, and it's worth keeping this in mind: as you build experience with models and techniques, think of ways you could put this into action at work today. Please do contact me if you'd like to discuss more or if you have ideas to bounce off me.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "markdown", + "id": "6900b2a8-6384-4316-8aaa-5e519fca4254", + "metadata": {}, + "source": [ + "# Connecting to OpenAI (or Ollama)\n", + "\n", + "The next cell is where we load in the environment variables in your `.env` file and connect to OpenAI. \n", + "\n", + "If you'd like to use free Ollama instead, please see the README section \"Free Alternative to Paid APIs\", and if you're not sure how to do this, there's a full solution in the solutions folder (day1_with_ollama.ipynb).\n", + "\n", + "## Troubleshooting if you have problems:\n", + "\n", + "Head over to the [troubleshooting](troubleshooting.ipynb) notebook in this folder for step by step code to identify the root cause and fix it!\n", + "\n", + "If you make a change, try restarting the \"Kernel\" (the python process sitting behind this notebook) by Kernel menu >> Restart Kernel and Clear Outputs of All Cells. Then try this notebook again, starting at the top.\n", + "\n", + "Or, contact me! Message me or email ed@edwarddonner.com and we will get this to work.\n", + "\n", + "Any concerns about API costs? See my notes in the README - costs should be minimal, and you can control it at every point. You can also use Ollama as a free alternative, which we discuss during Day 2." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "markdown", + "id": "442fc84b-0815-4f40-99ab-d9a5da6bda91", + "metadata": {}, + "source": [ + "# Let's make a quick call to a Frontier model to get started, as a preview!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a58394bf-1e45-46af-9bfd-01e24da6f49a", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with these messages is this easy. Any problems, head over to the Troubleshooting notebook.\n", + "\n", + "message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\":\"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "2aa190e5-cb31-456a-96cc-db109919cd78", + "metadata": {}, + "source": [ + "## OK onwards with our first project" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's try one out. Change the website and add print statements to follow along.\n", + "\n", + "ed = Website(\"https://edwarddonner.com\")\n", + "print(ed.title)\n", + "print(ed.text)" + ] + }, + { + "cell_type": "markdown", + "id": "6a478a0c-2c53-48ff-869c-4d08199931e1", + "metadata": {}, + "source": [ + "## Types of prompts\n", + "\n", + "You may know this already - but if not, you will get very familiar with it!\n", + "\n", + "Models like GPT4o have been trained to receive instructions in a particular way.\n", + "\n", + "They expect to receive:\n", + "\n", + "**A system prompt** that tells them what task they are performing and what tone they should use\n", + "\n", + "**A user prompt** -- the conversation starter that they should reply to" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0275b1b-7cfe-4f9d-abfa-7650d378da0c", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(ed))" + ] + }, + { + "cell_type": "markdown", + "id": "ea211b5f-28e1-4a86-8e52-c0b7677cadcc", + "metadata": {}, + "source": [ + "## Messages\n", + "\n", + "The API from OpenAI expects to receive messages in a particular structure.\n", + "Many of the other APIs share this structure:\n", + "\n", + "```python\n", + "[\n", + " {\"role\": \"system\", \"content\": \"system message goes here\"},\n", + " {\"role\": \"user\", \"content\": \"user message goes here\"}\n", + "]\n", + "```\n", + "To give you a preview, the next 2 cells make a rather simple call - we won't stretch the mighty GPT (yet!)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f25dcd35-0cd0-4235-9f64-ac37ed9eaaa5", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"system\", \"content\": \"You are a snarky assistant\"},\n", + " {\"role\": \"user\", \"content\": \"What is 2 + 2?\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "21ed95c5-7001-47de-a36d-1d6673b403ce", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with system and user messages:\n", + "\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "d06e8d78-ce4c-4b05-aa8e-17050c82bb47", + "metadata": {}, + "source": [ + "## And now let's build useful messages for GPT-4o-mini, using a function" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "# See how this function creates exactly the format above\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "36478464-39ee-485c-9f3f-6a4e458dbc9c", + "metadata": {}, + "outputs": [], + "source": [ + "# Try this out, and then try for a few more websites\n", + "\n", + "messages_for(ed)" + ] + }, + { + "cell_type": "markdown", + "id": "16f49d46-bf55-4c3e-928f-68fc0bf715b0", + "metadata": {}, + "source": [ + "## Time to bring it together - the API for OpenAI is very simple!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [], + "source": [ + "summarize(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "b3bcf6f4-adce-45e9-97ad-d9a5d7a3a624", + "metadata": {}, + "source": [ + "# Let's try more websites\n", + "\n", + "Note that this will only work on websites that can be scraped using this simplistic approach.\n", + "\n", + "Websites that are rendered with Javascript, like React apps, won't show up. See the community-contributions folder for a Selenium implementation that gets around this. You'll need to read up on installing Selenium (ask ChatGPT!)\n", + "\n", + "Also Websites protected with CloudFront (and similar) may give 403 errors - many thanks Andy J for pointing this out.\n", + "\n", + "But many websites will work just fine!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "45d83403-a24c-44b5-84ac-961449b4008f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "75e9fd40-b354-4341-991e-863ef2e59db7", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://anthropic.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "c951be1a-7f1b-448f-af1f-845978e47e2c", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business applications

\n", + " In this exercise, you experienced calling the Cloud API of a Frontier Model (a leading model at the frontier of AI) for the first time. We will be using APIs like OpenAI at many stages in the course, in addition to building our own LLMs.\n", + "\n", + "More specifically, we've applied this to Summarization - a classic Gen AI use case to make a summary. This can be applied to any business vertical - summarizing the news, summarizing financial performance, summarizing a resume in a cover letter - the applications are limitless. Consider how you could apply Summarization in your business, and try prototyping a solution.\n", + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Before you continue - now try yourself

\n", + " Use the cell below to make your own simple commercial example. Stick with the summarization use case for now. Here's an idea: write something that will take the contents of an email, and will suggest an appropriate short subject line for the email. That's the kind of feature that might be built into a commercial email tool.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"something here\"\n", + "user_prompt = \"\"\"\n", + " Lots of text\n", + " Can be pasted here\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [] # fill this in\n", + "\n", + "# Step 3: Call OpenAI\n", + "\n", + "response =\n", + "\n", + "# Step 4: print the result\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "36ed9f14-b349-40e9-a42c-b367e77f8bda", + "metadata": {}, + "source": [ + "## An extra exercise for those who enjoy web scraping\n", + "\n", + "You may notice that if you try `display_summary(\"https://openai.com\")` - it doesn't work! That's because OpenAI has a fancy website that uses Javascript. There are many ways around this that some of you might be familiar with. For example, Selenium is a hugely popular framework that runs a browser behind the scenes, renders the page, and allows you to query it. If you have experience with Selenium, Playwright or similar, then feel free to improve the Website class to use them. In the community-contributions folder, you'll find an example Selenium solution from a student (thank you!)" + ] + }, + { + "cell_type": "markdown", + "id": "eeab24dc-5f90-4570-b542-b0585aca3eb6", + "metadata": {}, + "source": [ + "# Sharing your code\n", + "\n", + "I'd love it if you share your code afterwards so I can share it with others! You'll notice that some students have already made changes (including a Selenium implementation) which you will find in the community-contributions folder. If you'd like add your changes to that folder, submit a Pull Request with your new versions in that folder and I'll merge your changes.\n", + "\n", + "If you're not an expert with git (and I am not!) then GPT has given some nice instructions on how to submit a Pull Request. It's a bit of an involved process, but once you've done it once it's pretty clear. As a pro-tip: it's best if you clear the outputs of your Jupyter notebooks (Edit >> Clean outputs of all cells, and then Save) for clean notebooks.\n", + "\n", + "Here are good instructions courtesy of an AI friend: \n", + "https://chatgpt.com/share/677a9cb5-c64c-8012-99e0-e06e88afd293" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": {}, + "outputs": [], + "source": [ + "from playwright.sync_api import sync_playwright\n", + "import time \n", + "import asyncio\n", + "from playwright.async_api import async_playwright\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fd3fdc92", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "44099289", + "metadata": {}, + "outputs": [], + "source": [ + "class Website:\n", + " def __init__(self, url):\n", + " self.url = url\n", + " self.title = None\n", + " self.text = None\n", + "\n", + " @classmethod\n", + " async def create(cls, url):\n", + " website = cls(url)\n", + " retries = 3 # Add retry logic\n", + " for attempt in range(retries):\n", + " try:\n", + " await website.initialize()\n", + " return website\n", + " except TimeoutError as e:\n", + " if attempt == retries - 1: # Last attempt\n", + " raise\n", + " print(f\"Attempt {attempt + 1} failed, retrying...\")\n", + " await asyncio.sleep(2) # Wait between retries\n", + "\n", + " async def initialize(self):\n", + " async with async_playwright() as p:\n", + " # Launch with stealth mode settings\n", + " browser = await p.chromium.launch(\n", + " headless=True,\n", + " args=[\n", + " '--disable-blink-features=AutomationControlled',\n", + " '--disable-dev-shm-usage',\n", + " '--no-sandbox'\n", + " ]\n", + " )\n", + " \n", + " # Create context with stealth settings\n", + " context = await browser.new_context(\n", + " user_agent='Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36',\n", + " viewport={'width': 1920, 'height': 1080},\n", + " java_script_enabled=True,\n", + " bypass_csp=True, # Bypass Content Security Policy\n", + " extra_http_headers={\n", + " 'Accept-Language': 'en-US,en;q=0.9',\n", + " 'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8'\n", + " }\n", + " )\n", + " \n", + " page = await context.new_page()\n", + " \n", + " try:\n", + " # Navigate with progressive waits\n", + " await page.goto(self.url, timeout=90000) # 90 second timeout\n", + " \n", + " # Wait for either real content or Cloudflare challenge\n", + " try:\n", + " # Wait for actual content first\n", + " await page.wait_for_selector('main', timeout=10000)\n", + " except:\n", + " # If main content not found, wait for Cloudflare to clear\n", + " await page.wait_for_load_state('networkidle', timeout=30000)\n", + " await page.wait_for_selector('body', state='visible', timeout=30000)\n", + " \n", + " # Get content after all waits\n", + " self.title = await page.title()\n", + " content = await page.content()\n", + " \n", + " soup = BeautifulSoup(content, 'html.parser')\n", + " for irrelevant in soup.find_all([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True) if soup.body else \"\"\n", + " \n", + " finally:\n", + " await browser.close()\n", + "\n", + "# Modified summarize function to use the factory method\n", + "async def summarize(url):\n", + " website = await Website.create(url)\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=messages_for(website)\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "# Display function remains the same\n", + "async def display_summary(url):\n", + " summary = await summarize(url)\n", + " display(Markdown(summary))\n", + "\n", + "# Usage\n", + "await display_summary(\"https://openai.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "78e0d270", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/pradeep1955/week1 EXERCISE.ipynb b/community-contributions/pradeep1955/week1 EXERCISE.ipynb new file mode 100644 index 0000000..5c418f2 --- /dev/null +++ b/community-contributions/pradeep1955/week1 EXERCISE.ipynb @@ -0,0 +1,148 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display\n", + "from dotenv import load_dotenv" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key=os.getenv(\"OPENAI_API_KEY\")\n", + "if not api_key.startswith(\"sk-proj-\") and len(api_key)<10:\n", + " print(\"api key not foud\")\n", + "else:\n", + " print(\"api found and is ok\")\n", + "\n", + "openai=OpenAI()\n", + "print()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "question = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "messages = [{\"role\":\"system\",\"content\":\"You are a expert Dta Scientist\"}, {\"role\":\"user\",\"content\":question}]\n", + "\n", + "stream = openai.chat.completions.create(\n", + " model = MODEL_GPT,\n", + " messages = messages,\n", + " stream = True\n", + ")\n", + "response = \"\"\n", + "display_handle = display(Markdown(\"\"), display_id=True)\n", + "for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "import ollama\n", + "\n", + "stream = ollama.chat(model=MODEL_LLAMA, messages=messages, stream=True)\n", + "response = \"\"\n", + "display_handle = display(Markdown(\"\"), display_id=True)\n", + "for chunk in stream:\n", + " response += chunk[\"message\"][\"content\"] or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2a573174-779b-4d50-8792-fa0889b37211", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llmenv", + "language": "python", + "name": "llmenv" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/pradeep1955/week1/day2 EXERCISE.ipynb b/community-contributions/pradeep1955/week1/day2 EXERCISE.ipynb new file mode 100644 index 0000000..d7a3078 --- /dev/null +++ b/community-contributions/pradeep1955/week1/day2 EXERCISE.ipynb @@ -0,0 +1,426 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# Welcome to your first assignment!\n", + "\n", + "Instructions are below. Please give this a try, and look in the solutions folder if you get stuck (or feel free to ask me!)" + ] + }, + { + "cell_type": "markdown", + "id": "ada885d9-4d42-4d9b-97f0-74fbbbfe93a9", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Just before we get to the assignment --

\n", + " I thought I'd take a second to point you at this page of useful resources for the course. This includes links to all the slides.
\n", + " https://edwarddonner.com/2024/11/13/llm-engineering-resources/
\n", + " Please keep this bookmarked, and I'll continue to add more useful links there over time.\n", + " \n", + "
" + ] + }, + { + "cell_type": "markdown", + "id": "6e9fa1fc-eac5-4d1d-9be4-541b3f2b3458", + "metadata": {}, + "source": [ + "# HOMEWORK EXERCISE ASSIGNMENT\n", + "\n", + "Upgrade the day 1 project to summarize a webpage to use an Open Source model running locally via Ollama rather than OpenAI\n", + "\n", + "You'll be able to use this technique for all subsequent projects if you'd prefer not to use paid APIs.\n", + "\n", + "**Benefits:**\n", + "1. No API charges - open-source\n", + "2. Data doesn't leave your box\n", + "\n", + "**Disadvantages:**\n", + "1. Significantly less power than Frontier Model\n", + "\n", + "## Recap on installation of Ollama\n", + "\n", + "Simply visit [ollama.com](https://ollama.com) and install!\n", + "\n", + "Once complete, the ollama server should already be running locally. \n", + "If you visit: \n", + "[http://localhost:11434/](http://localhost:11434/)\n", + "\n", + "You should see the message `Ollama is running`. \n", + "\n", + "If not, bring up a new Terminal (Mac) or Powershell (Windows) and enter `ollama serve` \n", + "And in another Terminal (Mac) or Powershell (Windows), enter `ollama pull llama3.2` \n", + "Then try [http://localhost:11434/](http://localhost:11434/) again.\n", + "\n", + "If Ollama is slow on your machine, try using `llama3.2:1b` as an alternative. Run `ollama pull llama3.2:1b` from a Terminal or Powershell, and change the code below from `MODEL = \"llama3.2\"` to `MODEL = \"llama3.2:1b\"`" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dac0a679-599c-441f-9bf2-ddc73d35b940", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a messages list using the same format that we used for OpenAI\n", + "\n", + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Describe some of the business applications of Generative AI\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7bb9c624-14f0-4945-a719-8ddb64f66f47", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "479ff514-e8bd-4985-a572-2ea28bb4fa40", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's just make sure the model is loaded\n", + "\n", + "!ollama pull llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "42b9f644-522d-4e05-a691-56e7658c0ea9", + "metadata": {}, + "outputs": [], + "source": [ + "# If this doesn't work for any reason, try the 2 versions in the following cells\n", + "# And double check the instructions in the 'Recap on installation of Ollama' at the top of this lab\n", + "# And if none of that works - contact me!\n", + "\n", + "response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "6a021f13-d6a1-4b96-8e18-4eae49d876fe", + "metadata": {}, + "source": [ + "# Introducing the ollama package\n", + "\n", + "And now we'll do the same thing, but using the elegant ollama python package instead of a direct HTTP call.\n", + "\n", + "Under the hood, it's making the same call as above to the ollama server running at localhost:11434" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7745b9c4-57dc-4867-9180-61fa5db55eb8", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "a4704e10-f5fb-4c15-a935-f046c06fb13d", + "metadata": {}, + "source": [ + "## Alternative approach - using OpenAI python library to connect to Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [], + "source": [ + "# There's actually an alternative approach that some people might prefer\n", + "# You can use the OpenAI client python library to call Ollama:\n", + "\n", + "from openai import OpenAI\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "9f9e22da-b891-41f6-9ac9-bd0c0a5f4f44", + "metadata": {}, + "source": [ + "## Are you confused about why that works?\n", + "\n", + "It seems strange, right? We just used OpenAI code to call Ollama?? What's going on?!\n", + "\n", + "Here's the scoop:\n", + "\n", + "The python class `OpenAI` is simply code written by OpenAI engineers that makes calls over the internet to an endpoint. \n", + "\n", + "When you call `openai.chat.completions.create()`, this python code just makes a web request to the following url: \"https://api.openai.com/v1/chat/completions\"\n", + "\n", + "Code like this is known as a \"client library\" - it's just wrapper code that runs on your machine to make web requests. The actual power of GPT is running on OpenAI's cloud behind this API, not on your computer!\n", + "\n", + "OpenAI was so popular, that lots of other AI providers provided identical web endpoints, so you could use the same approach.\n", + "\n", + "So Ollama has an endpoint running on your local box at http://localhost:11434/v1/chat/completions \n", + "And in week 2 we'll discover that lots of other providers do this too, including Gemini and DeepSeek.\n", + "\n", + "And then the team at OpenAI had a great idea: they can extend their client library so you can specify a different 'base url', and use their library to call any compatible API.\n", + "\n", + "That's it!\n", + "\n", + "So when you say: `ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')` \n", + "Then this will make the same endpoint calls, but to Ollama instead of OpenAI." + ] + }, + { + "cell_type": "markdown", + "id": "bc7d1de3-e2ac-46ff-a302-3b4ba38c4c90", + "metadata": {}, + "source": [ + "## Also trying the amazing reasoning model DeepSeek\n", + "\n", + "Here we use the version of DeepSeek-reasoner that's been distilled to 1.5B. \n", + "This is actually a 1.5B variant of Qwen that has been fine-tuned using synethic data generated by Deepseek R1.\n", + "\n", + "Other sizes of DeepSeek are [here](https://ollama.com/library/deepseek-r1) all the way up to the full 671B parameter version, which would use up 404GB of your drive and is far too large for most!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cf9eb44e-fe5b-47aa-b719-0bb63669ab3d", + "metadata": {}, + "outputs": [], + "source": [ + "!ollama pull deepseek-r1:1.5b" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1d3d554b-e00d-4c08-9300-45e073950a76", + "metadata": {}, + "outputs": [], + "source": [ + "# This may take a few minutes to run! You should then see a fascinating \"thinking\" trace inside tags, followed by some decent definitions\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=\"deepseek-r1:1.5b\",\n", + " messages=[{\"role\": \"user\", \"content\": \"Please give definitions of some core concepts behind LLMs: a neural network, attention and the transformer\"}]\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "1622d9bb-5c68-4d4e-9ca4-b492c751f898", + "metadata": {}, + "source": [ + "# NOW the exercise for you\n", + "\n", + "Take the code from day1 and incorporate it here, to build a website summarizer that uses Llama 3.2 running locally instead of OpenAI; use either of the above approaches." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "43ef4b92-53e1-4af2-af3f-726812f4265c", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "#from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "#from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "97d45733-394e-493e-a92b-1475876d9028", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6a40f9c5-1b14-42f9-9319-6a66e58e03f2", + "metadata": {}, + "outputs": [], + "source": [ + "webpage = Website(\"https://www.pleasurewebsite.com\")\n", + "print(webpage.title)\n", + "print(webpage.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a72a005d-43de-4ae5-b427-99a8fcb6065c", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0e4f95f-0ccf-4027-9457-5c973cd17702", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ceae6073-a085-49ce-ad44-39e46d8e6934", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9d53b26b-308c-470c-a0a9-9edb887aed6d", + "metadata": {}, + "outputs": [], + "source": [ + "messages=messages_for(webpage)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "MODEL = \"llama3.2\"\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llmenv", + "language": "python", + "name": "llmenv" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/pradeep1955/week2/agent_conversation_shakespeare.ipynb b/community-contributions/pradeep1955/week2/agent_conversation_shakespeare.ipynb new file mode 100644 index 0000000..6d55283 --- /dev/null +++ b/community-contributions/pradeep1955/week2/agent_conversation_shakespeare.ipynb @@ -0,0 +1,351 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "06cf3063-9f3e-4551-a0d5-f08d9cabb927", + "metadata": {}, + "source": [ + "# Triangular agent conversation\n", + "\n", + "## GPT (Hamlet), LLM (Falstaff), Gemini (Iago):" + ] + }, + { + "cell_type": "markdown", + "id": "3637910d-2c6f-4f19-b1fb-2f916d23f9ac", + "metadata": {}, + "source": [ + "### Created a 3-way, bringing Gemini into the coversation.\n", + "### Replacing one of the models with an open source model running with Ollama." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f8e0c1bd-a159-475b-9cdc-e219a7633355", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a3ad57ad-46a8-460e-9cb3-67a890093536", + "metadata": {}, + "outputs": [], + "source": [ + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4f531c14-5743-4a5b-83d9-cb5863ca2ddf", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d5150ee-3858-4921-bce6-2eecfb96bc75", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI\n", + "\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "11381fd8-5099-41e8-a1d7-6787dea56e43", + "metadata": {}, + "outputs": [], + "source": [ + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1766d20-54b6-4f76-96c5-c338ae7073c9", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_model = \"gpt-4o-mini\"\n", + "llama_model = \"llama3.2\"\n", + "gemini_model = 'gemini-2.0-flash'\n", + "\n", + "gpt_system = \"You are playing part of Hamlet. he is philosopher, probes Iago with a mixture of suspicion\\\n", + "and intellectual curiosity, seeking to unearth the origins of his deceit.\\\n", + "Is malice born of scorn, envy, or some deeper void? Hamlet’s introspective nature\\\n", + "drives him to question whether Iago’s actions reveal a truth about humanity itself.\\\n", + "You will respond as Shakespear's Hamlet will do.\"\n", + "\n", + "llama_system = \"You are acting part of Falstaff who attempts to lighten the mood with his jokes and observations,\\\n", + "potentially clashing with Hamlet's melancholic nature.You respond as Shakespear's Falstaff do.\"\n", + "\n", + "gemini_system = \"You are acting part of Iago, subtly trying to manipulate both Hamlet and Falstaff\\\n", + "to his own advantage, testing their weaknesses and exploiting their flaws. You respond like Iago\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hello\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "806a0506-dac8-4bad-ac08-31f350256b58", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, llama_messages, gemini_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "43674885-ede7-48bf-bee4-467454f3e96a", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llama():\n", + " messages = []\n", + " for gpt, llama, gemini in zip(gpt_messages, llama_messages, gemini_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": llama})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " response = ollama.chat(model=llama_model, messages=messages)\n", + "\n", + " \n", + " return response['message']['content']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "03d34769-b339-4c4b-8c60-69494c39d725", + "metadata": {}, + "outputs": [], + "source": [ + "#import google.generativeai as genai\n", + "\n", + "# Make sure you configure the API key first:\n", + "#genai.configure(api_key=\"YOUR_API_KEY\")\n", + "\n", + "def call_gemini():\n", + " gemini_messages = []\n", + " \n", + " # Format the history for Gemini\n", + " for gpt, llama, gemini_message in zip(gpt_messages, llama_messages, gemini_messages):\n", + " gemini_messages.append({\"role\": \"user\", \"parts\": [gpt]}) # Hamlet speaks\n", + " gemini_messages.append({\"role\": \"model\", \"parts\": [llama]}) # Falstaff responds\n", + " gemini_messages.append({\"role\": \"model\", \"parts\": [gemini_message]}) # Iago responds\n", + "\n", + " # Add latest user input if needed (optional)\n", + " gemini_messages.append({\"role\": \"user\", \"parts\": [llama_messages[-1]]})\n", + "\n", + " # Initialize the model with the correct system instruction\n", + " gemini = google.generativeai.GenerativeModel(\n", + " #model_name='gemini-1.5-flash', # Or 'gemini-pro'\n", + " model_name = gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + "\n", + " response = gemini.generate_content(gemini_messages)\n", + " return response.text\n", + "#print(response.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "93fc8253-67cb-4ea4-aff7-097b2a222793", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hello\"]\n", + "\n", + "print(f\"Hamlet:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Falstaff:\\n{llama_messages[0]}\\n\")\n", + "print(f\"Iago:\\n{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(3):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " llama_next = call_llama()\n", + " print(f\"Llama:\\n{llama_next}\\n\")\n", + " llama_messages.append(llama_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " llama_messages.append(gemini_next)" + ] + }, + { + "cell_type": "markdown", + "id": "bca66ffc-9dc1-4384-880c-210889f5d0ac", + "metadata": {}, + "source": [ + "## Conversation between gpt-4.0-mini and llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c23224f6-7008-44ed-a57f-718975f4e291", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4o-mini and Claude-3-haiku\n", + "# We're using cheap versions of models so the costs will be minimal\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "llama_model = \"llama3.2\"\n", + "\n", + "gpt_system = \"You are a tapori from mumbai who is very optimistic; \\\n", + "you alway look at the brighter part of the situation and you always ready to take act to win way.\"\n", + "\n", + "llama_system = \"You are a Jaat from Haryana. You try to express with hindi poems \\\n", + "to agree with other person and or find common ground. If the other person is optimistic, \\\n", + "you respond in poetic way and keep chatting.\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2d704bbb-f22b-400d-a695-efbd02b26548", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, llama in zip(gpt_messages, llama_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": llama})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "385ccec8-de59-4e42-9616-3f5c9a05589c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llama():\n", + " messages = []\n", + " for gpt, llama_message in zip(gpt_messages, llama_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": llama_message})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " response = ollama.chat(model=llama_model, messages=messages)\n", + "\n", + " \n", + " return response['message']['content']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "70b5481b-455e-4275-80d3-0afe0fabcb0f", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Llama:\\n{llama_messages[0]}\\n\")\n", + "\n", + "for i in range(3):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " llama_next = call_llama()\n", + " print(f\"Llama:\\n{llama_next}\\n\")\n", + " llama_messages.append(llama_next)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7f8d734b-57e5-427d-bcb1-7956fc58a348", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llmenv", + "language": "python", + "name": "llmenv" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/community-contributions/synthetic-dataset-generator/README.md b/community-contributions/synthetic-dataset-generator/README.md new file mode 100644 index 0000000..4148810 --- /dev/null +++ b/community-contributions/synthetic-dataset-generator/README.md @@ -0,0 +1,37 @@ +# LLM-Powered Dataset Synthesizer: LLaMA 3 + Gradio Demo + +This interactive demo showcases a synthetic dataset generation pipeline powered by Meta's LLaMA 3.1 8B-Instruct model, running in 4-bit quantized mode. Users can input natural language prompts describing the structure and logic of a desired dataset, and the model will generate tabular data accordingly. + +## ✨ Description + +Modern LLMs are capable of reasoning over structured data formats and generating realistic, constrained datasets. This demo leverages the LLaMA 3.1 instruct model, combined with prompt engineering, to generate high-quality synthetic tabular data from plain-language descriptions. + +Key components: +- **LLaMA 3.1 8B-Instruct** via Hugging Face Transformers +- **4-bit quantized loading** with `bitsandbytes` for memory efficiency +- **Custom prompt framework** for schema + value constraints +- **Interactive interface** built with Gradio for user-friendly data generation + +## 🚀 Functionality + +With this tool, you can: +- Generate synthetic datasets by describing the column names, data types, value logic, and number of rows +- Apply constraints based on age, gender, matching conditions, and more (e.g., “females over 40; males under 40”) +- Preview the raw model output or extract structured JSON/tabular results +- Interactively explore and copy generated datasets from the Gradio UI + +## 🛠️ Under the Hood + +- The model prompt template includes both a **system message** and user instruction +- Output is parsed to extract valid JSON objects +- The generated data is displayed in the Gradio interface and downloadable as CSV + +## 📦 Requirements + +- Python (Colab recommended) +- `transformers`, `bitsandbytes`, `accelerate`, `gradio`, `torch` +- Hugging Face access token with permission to load LLaMA 3.1 + +--- + +Ready to generate smart synthetic datasets with just a sentence? Try it! diff --git a/community-contributions/synthetic-dataset-generator/synthgen.ipynb b/community-contributions/synthetic-dataset-generator/synthgen.ipynb new file mode 100644 index 0000000..872e368 --- /dev/null +++ b/community-contributions/synthetic-dataset-generator/synthgen.ipynb @@ -0,0 +1,4814 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "953b642e", + "metadata": {}, + "source": [ + "# SynthGen: an LLM-Powered Dataset Synthesizer\n", + "> ⚡ Powered by LLaMa 3 from HuggingFace and an intuitive Gradio UI.\n", + "\n", + "This notebook demonstrates how to generate realistic, structured tabular data using natural language prompts powered by the Meta-LLaMA-3.1 8B-Instruct model. By specifying column names, value types, and logical constraints in plain English, users can produce tailored synthetic datasets interactively via a Gradio interface.\n", + "\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "a8a50195", + "metadata": {}, + "source": [ + "### 1. Installing Required Packages" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0d7817f1", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "0d7817f1", + "outputId": "8ae1c98c-c3da-44de-a908-441cf45c5ee8" + }, + "outputs": [], + "source": [ + "!pip install -q --upgrade torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 --index-url https://download.pytorch.org/whl/cu124\n", + "!pip install -q requests bitsandbytes==0.46.0 transformers==4.48.3 accelerate==1.3.0" + ] + }, + { + "cell_type": "markdown", + "id": "b11fc9c1", + "metadata": {}, + "source": [ + "## 2. Importing Libraries and HF Setup" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1c323f2", + "metadata": { + "id": "c1c323f2" + }, + "outputs": [], + "source": [ + "import os\n", + "from IPython.display import Markdown, display, update_display\n", + "\n", + "import gradio as gr\n", + "import pandas as pd\n", + "\n", + "from google.colab import userdata\n", + "from huggingface_hub import login\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig\n", + "import torch\n", + "import gc\n", + "import json\n", + "import re\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbfbb506", + "metadata": { + "id": "cbfbb506" + }, + "outputs": [], + "source": [ + "hf_token = userdata.get('HF_TOKEN')\n", + "login(hf_token, add_to_git_credential=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "koICwnTwQMVP", + "metadata": { + "id": "koICwnTwQMVP" + }, + "outputs": [], + "source": [ + "# We are using the Meta-Llama-3.1-8B-Instruct model. Feel free to change this to any other model you prefer, \n", + "# by adjusting the model variant accordingly, and ensuring the template is compatible.\n", + "model_variant = \"meta-llama/Meta-Llama-3.1-8B-Instruct\"" + ] + }, + { + "cell_type": "markdown", + "id": "a817a180", + "metadata": {}, + "source": [ + "### 3. Loading the LLaMA 3.1 Model (Quantized)\n", + "Depending on the model, this may take a few." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "JbZP17wkSdW0", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 456, + "referenced_widgets": [ + "9dde77220ce04b80ac2b999d9304ac61", + "3562239e91694d618538fabbf55bdc7e", + "90d6add2cac446958326cb3cb4defc35", + "36897c09ed284e9ebad8b610186c5c2f", + "19f90c9b7f6b4fb4b7e4db0fda3697e6", + "037a38e5d42b4643a40e125d1ca7aa9d", + "114ae2bee64b4ca09b0f06e8d7f154c7", + "4e0a5f3cf25b4d4bb4248f3a686a3b2b", + "3d71197f73b540cc8a0702165b336ee6", + "4ace0ac8c70b43f9b524e82b7102b492", + "6ee18cbecc5c446782764a7a7a2093ed", + "ea8987a8a81a4c1ca33ba1f6d35eabf7", + "9181116e0dab421e8538c3b5e6d78a63", + "06fc98bc230248618f718cc789860551", + "b7f4683f24f6497084de01fc9dcb7b4c", + "216a21cb431c4efcb25118d60ce55382", + "3d3301ac71ed4e988deeafb5b29b1e7e", + "4ee7c6a29caa45589987d7da4fcebca2", + "cb53de4aa61843b0a75ead30dfece607", + "0ca526d86ebb433095f68176806c4dd9", + "6f64c037d228441489da9729d0eb45c0", + "092185914174457cb395e16f5dfd4774", + "3752691ba21845b38cefd4e26d88e285", + "c9f51294dfa046db9ec6d89a3e2d8dd4", + "440508c6e1f44788aa8317b07619ad6a", + "77e2f21d581c4779856b1cca34a61c87", + "22683400287c4dab86cb868a40f3ebd2", + "f451e2532e5241448a85daefb01822e2", + "f03f977f28364ec18ce1659c76aa82ba", + "97f18919e94e4f0281c693d1c748e7c7", + "a602bd2f6c3041d28324656a52d2527f", + "ee2db85a296d4a9bb6895c8438b4ef21", + "530fe1f6d93d49749a82ea977673260f", + "214b4bb379564949a580f2169663d69a", + "d2b3737cd56941e79b22af8689afddbf", + "399b0e04754c4b88b01ac908bd17c128", + "3d536c5e0d9746b0ac7ba908908e6c46", + "4e8685101b26483685c9dd822b065d3e", + "22f1bca162774d1695505997d0b97bc6", + "5f66742ac0ee4404959a83b4362aeb42", + "6cfbb81314f8461fa098d86ff49516d3", + "b870a05e12784e81aadf2e804cefe916", + "baa7428cdefa42e6bb894448a043b41e", + "289470951d374d18948c3e43ec2282df", + "86c78bcedbfd496a85a5d5ddb9c1df85", + "0545d6dd41964f799d689964b3e2a357", + "22ae6e2ec2474ce0b6b0bf6ca3039070", + "c311192f10264e399bc2e77b1aabb7d2", + "f39be376072c43f7bc2dd1aff4eeb777", + "6e0a43d7cf314faba17bd88e1e37f021", + "9ca5e65c356543df95542ddf1cb87ab9", + "a07eb638da684d70a5f7760a5cd030da", + "a8e91c03e59940c9a47f429da4488f53", + "b12dc89359524e9e8edc2b90d68d6076", + "fe2dd51da47e4130a3d5713037784787", + "d0306eea21a045f7a169d64d615bbf6b", + "f435e52db79b4317b55f13b8c98fe766", + "53227771ba154abc9a682e9a1e18ede9", + "e23aba3a1ac4420da2b0570a0dea2011", + "7e42a603efbe4d4bacb61114ca8bfdaf", + "c264cb454c7d43ccb2dd6045732f0317", + "87ba3449a7ce4bda8730f2d7c26b5f3f", + "faa55a15d9214cfb966a96f33a96b31d", + "8a29a11e60b24404a6b53573f81085ce", + "f7e02670ead24d12b09aa36cdf1ad98f", + "ca359ec98ebc49f8b121dec67853f797", + "38566e74ed4143408584dce8446bd70d", + "d68f54f5cee24037b5691cb3735e5248", + "8e2f22b60ff34706bd0503fc0b10cf24", + "15f310a76aea48b7a6b5efc87558bb90", + "e775d3ddb0f54370a39b806024a7102c", + "54849eb44d73489bb7835b1f98e81995", + "1ed77b06007b426b8944b2748431402d", + "d9c4f5565a2f4faf9598c4377fc975ac", + "9626d058aa584a9e829d7793676eb8cd", + "0b0793726f2740be8e6c1b7452bb2c14", + "83cdabe816e44bc6964ba96dea753e15", + "4a669b8562a449afb6f0a085d258040a", + "54d9fd29bff64359b5cc6d41551dec43", + "1aac3f0e39e94eb18d5544ababc48377", + "a71122fe861e4a819c4d79b4ba47f361", + "5f8d48cfdadc4791aff0a525ae217c34", + "081e77e2c61f49babc1f48da215386d8", + "5964a39c2f4b46ea9f2a24c607a33160", + "600eee97f1fb478b919d5755baac574d", + "56f3290340934a28bc43e73fe5e8d050", + "f08d87275f5243e48e5277291133655c", + "326286fc11c84d2092a88555ef8a8aed", + "5e74177710b1492292796c2bf90ee8e4", + "c56d8b6ad9d64050ac6f26b33d142db8", + "cdf784f05efb411a927d5ad1ac363046", + "20341d8846494bc7a0d3bff2620534ac", + "7031154b6c514410aa517a287d0bb90f", + "21b3c6aceb5741dd847eed7955c47f17", + "b0fece23eefb4f069ffd2282bf05aa50", + "52ba48ea1dfc4266a57b8b70aefdc44c", + "aa52d70d64b7456eb99cc9386a6b773b", + "31a933bedaad4d5489a4983bd6ca9e6f", + "6e115f191b664f59b8144719ac6d6d13", + "03fd43b58e74455992d71a1625d368ad", + "bd0133b89f784f16ab20ff28574d537b", + "2ad775a43a284f489a7f7fef1c20dbb9", + "b1a13112ead34742835e635910820dd0", + "4697e1265559409f82f463f339a5f62c", + "31f68d4b6f624040899436006275613d", + "8e93a17e01b64ff89f3220565ada3082", + "79dbb408b3cf498895c8b066dd04b1c5", + "924bb29fc7134a0b98187770d0943d3d", + "161f9b5e075b404c871a58fbd560d63d", + "2dbcf9f4ebb54761802ef3bda76f2294", + "c1d3120bad3d456285a0807445d5e4ee", + "df49ac181ec244f5bebe6680408fd7ee", + "7836c3b013ec4f67a2b09670878581ba", + "f3467a09b50a4b429ff75009a26d992b", + "b9d1f51b79d34264965e512bfc0253b3", + "06722a988e1a4357a3841723aa66dfa8", + "46f37131f36a413badffb0c362ab8b5d", + "0e1f48ecaf6e4184953ec773bed24253", + "38ee56d491c74186a2146cac776a631b", + "e59cfa5ca5b240159517ccc2a594fa50", + "e18941e122eb4332b9e5fb1c3a22b9ed", + "2711dbd8740b453c849026f3d83f47af", + "0d8f09aba9514921baba774dcc1dd088", + "4a10cac721514d30b6c10978d33c6e72", + "f9d1ba01c0054ae3917a78e420174f2e", + "e95f552f62344e52aa2494fbaa0e1f16", + "85c5b698708b47319067527bda04445f", + "2bec34b01cfb452787013ab7bcef1be3", + "2c7826be2c25431081eb05ee2d7d4401", + "5c2b5412ceb54502b5018f2b756aa291", + "001b683aec074eb98b498f0dec667b7c", + "eb0e9fb2bf8447748a936749dceff1b9" + ] + }, + "id": "JbZP17wkSdW0", + "outputId": "b19f6cc6-872d-42ce-8c08-f750052ea7e6" + }, + "outputs": [], + "source": [ + "quant_config = BitsAndBytesConfig(\n", + "load_in_4bit=True,\n", + "bnb_4bit_use_double_quant=True,\n", + "bnb_4bit_compute_dtype=torch.bfloat16,\n", + "bnb_4bit_quant_type=\"nf4\"\n", + ")\n", + "\n", + "model = AutoModelForCausalLM.from_pretrained(\n", + " model_variant,\n", + " quantization_config=quant_config,\n", + " device_map=\"auto\",\n", + " trust_remote_code=True,\n", + " use_auth_token=True\n", + ")\n", + "\n", + "\n", + "tokenizer = AutoTokenizer.from_pretrained(model_variant)\n", + "tokenizer.pad_token = tokenizer.eos_token" + ] + }, + { + "cell_type": "markdown", + "id": "1877a8cb", + "metadata": {}, + "source": [ + "### 4. Prompt Construction and Inference\n", + "\n", + "In this section, we've crafted a flexible prompt template designed to guide the LLM toward generating clean, structured output that Gradio can easily process. The goal is to balance natural language flexibility with enough constraints to encourage valid, consistent, and readable tabular data.\n", + "\n", + "That said, LLMs can still be unpredictable or inconsistent, especially with loosely defined instructions. Feel free to adjust the prompt logic or system message to better suit your use case or to improve output reliability.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "935477d2", + "metadata": { + "id": "935477d2" + }, + "outputs": [], + "source": [ + "\n", + "def build_system_query(user_instructions: str = '', include_system_prompt: bool = True):\n", + " messages = [\n", + " {\n", + " \"role\": \"system\",\n", + " \"content\": (\n", + " \"You are a dataset generation assistant. The user will provide prompts containing column names, value types, number of rows, and other parameters and constraints.\\n\"\n", + " \"Your task is to generate synthetic datasets based on the user's request.\\n\\n\"\n", + " \"Always respond with a valid JSON object (dictionary), where:\\n\"\n", + " \"- Each key is a column name (formatted for pandas).\\n\"\n", + " \"- Each value is a list of values for that column.\\n\\n\"\n", + " \"Example format:\\n\"\n", + " \"{\\n\"\n", + " \" \\\"first_name\\\": [\\\"Alex\\\", \\\"Anthony\\\", \\\"Ava\\\", \\\"Amber\\\", \\\"Annabelle\\\"],\\n\"\n", + " \" \\\"age\\\": [23, 27, 35, 29, 31]\\n\"\n", + " \"}\\n\\n\"\n", + " \"If a value type or constraint is missing or unclear, make a reasonable assumption based on the column name and context.\\n\"\n", + " \"Do not include any explanations, comments, or extra text — only the raw JSON.\\n\"\n", + " \"Ensure the response is compact, well-formatted, and syntactically valid for parsing by JSON tools or conversion into a pandas DataFrame.\\n\"\n", + " \"Convert column names to pandas-compatible formats when needed (e.g., replace spaces with underscores, remove special characters, lowercase).\\n\"\n", + " )\n", + "\n", + " },\n", + " {\"role\": \"user\", \"content\": user_instructions}\n", + " ]\n", + "\n", + " return messages if include_system_prompt else messages[1:]" + ] + }, + { + "cell_type": "markdown", + "id": "83c6b3c2", + "metadata": {}, + "source": [ + "### 4.5 Let's Test It! \n", + "*(The following query intentionally includes some typos)*" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "tr29FmHQTkzy", + "metadata": { + "id": "tr29FmHQTkzy" + }, + "outputs": [], + "source": [ + "user_query = \"Generate 3 columns, firstNAme, last nam, aGe, 10 rows. Male names should correspond to an age range between 30 and 50.\"\n", + "user_query += \"Female names should be between 34 and 40. Make sure they match row-wise. \\n\"\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "SWAGfc39RxWr", + "metadata": { + "id": "SWAGfc39RxWr" + }, + "outputs": [], + "source": [ + "messages = build_system_query(user_instructions=user_query, include_system_prompt=True)\n", + "inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + "\n", + "outputs = model.generate(\n", + " inputs,\n", + " max_new_tokens=800,\n", + " do_sample=True,\n", + " temperature=0.7,\n", + " top_p=0.95,\n", + " pad_token_id=tokenizer.eos_token_id,\n", + " eos_token_id=tokenizer.eos_token_id # Optional, to reinforce stopping\n", + ")\n", + "decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "JgAGOumVHw8W", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "JgAGOumVHw8W", + "outputId": "b4853467-d868-4636-a80e-0c39260c4d93" + }, + "outputs": [], + "source": [ + "# Let's first print the full output to see what we got\n", + "print(decoded)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "gpu0Zji2DtrV", + "metadata": { + "id": "gpu0Zji2DtrV" + }, + "outputs": [], + "source": [ + "# The following function attempts to extract a valid JSON response from the model's output.\n", + "# It searches for the last valid JSON object in the text, which is useful if the model\n", + "# generates additional text or explanations that are not part of the JSON response.\n", + "def extract_response(text: str):\n", + " decoder = json.JSONDecoder()\n", + " starts = [m.start() for m in re.finditer(r'{', text)]\n", + " for pos in reversed(starts): # Start from last candidate\n", + " try:\n", + " parsed, _ = decoder.raw_decode(text[pos:])\n", + " if isinstance(parsed, dict):\n", + " return parsed\n", + " except json.JSONDecodeError:\n", + " continue\n", + " return None\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "MwofOSXTqWA-", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 363 + }, + "id": "MwofOSXTqWA-", + "outputId": "9bc87b6e-4aca-44b6-956b-49ef4830d1d1" + }, + "outputs": [], + "source": [ + "# Convert the extracted JSON response into a pandas DataFrame\n", + "if decoded:\n", + " pd.DataFrame(extract_response(decoded))" + ] + }, + { + "cell_type": "markdown", + "id": "3e4e5fbe", + "metadata": {}, + "source": [ + "### 5. Interactive UI with Gradio" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9448bc9a", + "metadata": { + "id": "9448bc9a" + }, + "outputs": [], + "source": [ + "# The following functions are used to handle the Gradio interface and dataset management.\n", + "\n", + "\n", + "def remove_items(selected_features, existing_dataframe: pd.DataFrame):\n", + " \"\"\"Remove selected features from the existing DataFrame.\"\"\"\n", + " print(\"Before removal:\", existing_dataframe)\n", + "\n", + " if isinstance(selected_features, str):\n", + " selected_features = [selected_features]\n", + "\n", + " edited_df = existing_dataframe.copy()\n", + "\n", + " if selected_features:\n", + " edited_df.drop(columns=selected_features, axis=1, inplace=True)\n", + "\n", + " # Force clean index/column structure\n", + " edited_df = edited_df.copy() # Ensures no hidden pandas artifacts\n", + "\n", + " updated_features_list = list(edited_df.columns)\n", + "\n", + " print(\"After removal:\", edited_df)\n", + " print('shape', edited_df.shape[0])\n", + " return (\n", + " gr.update(choices=updated_features_list, value=[]),\n", + " edited_df,\n", + " gr.update(value=edited_df, visible=not edited_df.empty),\n", + " gr.update(interactive=edited_df.empty)\n", + " )\n", + "\n", + "\n", + "def generate_features(n_rows: str, instructions: str, existing_dataframe: pd.DataFrame):\n", + " \"\"\"Generate new features based on user instructions and existing data.\"\"\"\n", + " has_valid_rows = n_rows.isnumeric() and int(n_rows) >= 1\n", + " if has_valid_rows and instructions:\n", + " try:\n", + " # Prepare prompt for the model\n", + " user_query = instructions.strip() + f\"\\nGenerate {n_rows} rows.\"\n", + " messages = build_system_query(user_instructions=user_query, include_system_prompt=True)\n", + "\n", + " # Tokenize and generate\n", + " inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + " outputs = model.generate(\n", + " inputs,\n", + " max_new_tokens=800,\n", + " do_sample=True,\n", + " temperature=0.7,\n", + " top_p=0.95,\n", + " pad_token_id=tokenizer.eos_token_id,\n", + " eos_token_id=tokenizer.eos_token_id,\n", + " )\n", + "\n", + " # Decode and parse model output\n", + " decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)\n", + " new_df = pd.DataFrame(extract_response(decoded.split(\"assistant\")[-1]))\n", + "\n", + " # Combine with existing data\n", + " if len(existing_dataframe) == 0:\n", + " updated_df = new_df\n", + " elif len(existing_dataframe) == len(new_df):\n", + " updated_df = pd.concat([existing_dataframe.reset_index(drop=True),\n", + " new_df.reset_index(drop=True)], axis=1)\n", + " else:\n", + " raise ValueError(\"Row count mismatch between existing and new feature data.\")\n", + "\n", + " # UI component updates\n", + " rows_component_update = gr.update(interactive=False)\n", + " table_view_update = gr.update(visible=True, value=updated_df, headers=list(updated_df.columns))\n", + " feature_display_update = gr.update(choices=list(updated_df.columns))\n", + "\n", + " return rows_component_update, table_view_update, feature_display_update, updated_df, ''\n", + "\n", + " except Exception as e:\n", + " print(\"Error generating feature:\", e)\n", + "\n", + " # Return empty updates if input is invalid\n", + " return (gr.update(),) * 5\n", + "\n", + "def on_selected_feture(selected_items):\n", + " \"\"\"Update the UI based on selected features.\"\"\"\n", + " return gr.update(interactive=not selected_items==[])\n", + "\n", + "def export_dataset(dataframe: pd.DataFrame):\n", + " \"\"\"Export the DataFrame to a CSV file.\"\"\"\n", + " try:\n", + " n = 0\n", + " while True:\n", + " if n == 0:\n", + " filename = \"dataset.csv\"\n", + " else:\n", + " filename = f\"dataset{n}.csv\"\n", + " if not os.path.exists(filename):\n", + " break\n", + " n += 1\n", + "\n", + " dataframe.to_csv(filename, index=False)\n", + " except Exception as e:\n", + " print(\"Error exporting dataset:\", e)\n", + "\n", + "\n", + "\n", + "def on_table_change(changed_data):\n", + " \"\"\"Handle changes in the table and return a DataFrame.\"\"\"\n", + " print('on table change')\n", + " df = pd.DataFrame(changed_data)\n", + " return df if not df.empty else pd.DataFrame(), df.shape[0] if df.shape[0] > 0 else None\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8d7855f1", + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 682 + }, + "id": "8d7855f1", + "outputId": "a6517655-4861-431d-fa13-fa1f93bce503" + }, + "outputs": [], + "source": [ + "\n", + "# UI layout\n", + "with gr.Blocks() as demo:\n", + " feature_state = gr.State(pd.DataFrame())\n", + "\n", + " with gr.Row():\n", + " # Left Column: Inputs and Controls\n", + " with gr.Column(scale=0):\n", + " with gr.Group():\n", + " gr.Markdown(\"### 🧾 Current Feature List\")\n", + " feature_display = gr.CheckboxGroup(\n", + " #choices=feature_dictionary,\n", + " label=\"Features\",\n", + " info=\"Select features to remove\"\n", + " )\n", + " with gr.Row(elem_classes=\"centered-button-row\"):\n", + " remove_btn = gr.Button(value=\"Remove Selected\", elem_classes=\"small-button\", interactive=False)\n", + "\n", + " # Add feature section\n", + " with gr.Group():\n", + " gr.Markdown(\"### ➕ Add a New Feature\")\n", + " with gr.Row(equal_height=True):\n", + " n_rows = gr.Text(\n", + " label=\"Number of Rows\",\n", + " placeholder=\"e.g., 100\"\n", + " )\n", + "\n", + " instructions = gr.Text(\n", + " label=\"Instructions\",\n", + " placeholder=\"e.g., first names; starting with A or B. \\nAge, numeric; range 21–55, males should be between 30-40. Correlate rows\",\n", + " scale=1,\n", + " lines=4\n", + " )\n", + "\n", + " with gr.Row(elem_classes=\"centered-button-row\"):\n", + " add_btn = gr.Button(value=\"Add\", elem_classes=\"small-button\")\n", + "\n", + " # Dataset generation section\n", + " with gr.Group():\n", + " with gr.Row(elem_classes=\"centered-button-row\"):\n", + " export_btn = gr.Button(value=\"💾 Export...\", elem_classes=\"small-button\")\n", + "\n", + " # Right Column: Dataset display\n", + " with gr.Column():\n", + " gr.Markdown(\"### 📊 Generated Dataset\")\n", + " table_view = gr.Dataframe(\n", + " interactive=True,\n", + " visible=False,\n", + " label=\"Dataset\"\n", + " )\n", + "\n", + " feature_display.change(\n", + " fn=on_selected_feture,\n", + " inputs=[feature_display],\n", + " outputs=[remove_btn]\n", + " )\n", + "\n", + " export_btn.click(\n", + " fn=export_dataset,\n", + " inputs=[feature_state]\n", + " )\n", + "\n", + "\n", + " remove_btn.click(\n", + " fn=remove_items,\n", + " inputs=[feature_display, feature_state],\n", + " outputs=[feature_display, feature_state, table_view, n_rows]\n", + " )\n", + "\n", + " add_btn.click(\n", + " fn=generate_features,\n", + " inputs=[n_rows, instructions, feature_state],\n", + " outputs=[n_rows, table_view, feature_display, feature_state, instructions]\n", + " )\n", + "\n", + " table_view.change(\n", + " fn=on_table_change,\n", + " inputs=[table_view],\n", + " outputs=[feature_state, n_rows]\n", + " )\n", + "\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "c5350c20", + "metadata": {}, + "source": [ + "### Let's test it!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1c3bc056", + "metadata": { + "id": "1c3bc056" + }, + "outputs": [], + "source": [ + "demo.launch(debug=True) # Set debug=True to see detailed error messages in the console\n", + "# demo.launch(share=True) # Uncomment this line to share the app publicly" + ] + } + ], + "metadata": { + "accelerator": "GPU", + "colab": { + "gpuType": "T4", + "provenance": [] + }, + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + }, + "widgets": { + "application/vnd.jupyter.widget-state+json": { + "001b683aec074eb98b498f0dec667b7c": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "037a38e5d42b4643a40e125d1ca7aa9d": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "03fd43b58e74455992d71a1625d368ad": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_bd0133b89f784f16ab20ff28574d537b", + "IPY_MODEL_2ad775a43a284f489a7f7fef1c20dbb9", + "IPY_MODEL_b1a13112ead34742835e635910820dd0" + ], + "layout": "IPY_MODEL_4697e1265559409f82f463f339a5f62c" + } + }, + "0545d6dd41964f799d689964b3e2a357": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_6e0a43d7cf314faba17bd88e1e37f021", + "placeholder": "​", + "style": "IPY_MODEL_9ca5e65c356543df95542ddf1cb87ab9", + "value": "model-00002-of-00004.safetensors: 100%" + } + }, + "06722a988e1a4357a3841723aa66dfa8": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "06fc98bc230248618f718cc789860551": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_cb53de4aa61843b0a75ead30dfece607", + "max": 23950, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_0ca526d86ebb433095f68176806c4dd9", + "value": 23950 + } + }, + "081e77e2c61f49babc1f48da215386d8": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "092185914174457cb395e16f5dfd4774": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "0b0793726f2740be8e6c1b7452bb2c14": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "0ca526d86ebb433095f68176806c4dd9": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "0d8f09aba9514921baba774dcc1dd088": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_85c5b698708b47319067527bda04445f", + "placeholder": "​", + "style": "IPY_MODEL_2bec34b01cfb452787013ab7bcef1be3", + "value": "special_tokens_map.json: 100%" + } + }, + "0e1f48ecaf6e4184953ec773bed24253": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "114ae2bee64b4ca09b0f06e8d7f154c7": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "15f310a76aea48b7a6b5efc87558bb90": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_0b0793726f2740be8e6c1b7452bb2c14", + "placeholder": "​", + "style": "IPY_MODEL_83cdabe816e44bc6964ba96dea753e15", + "value": " 1.17G/1.17G [00:21<00:00, 111MB/s]" + } + }, + "161f9b5e075b404c871a58fbd560d63d": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "19f90c9b7f6b4fb4b7e4db0fda3697e6": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "1aac3f0e39e94eb18d5544ababc48377": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_600eee97f1fb478b919d5755baac574d", + "max": 4, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_56f3290340934a28bc43e73fe5e8d050", + "value": 4 + } + }, + "1ed77b06007b426b8944b2748431402d": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "20341d8846494bc7a0d3bff2620534ac": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_31a933bedaad4d5489a4983bd6ca9e6f", + "placeholder": "​", + "style": "IPY_MODEL_6e115f191b664f59b8144719ac6d6d13", + "value": " 184/184 [00:00<00:00, 12.4kB/s]" + } + }, + "214b4bb379564949a580f2169663d69a": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_d2b3737cd56941e79b22af8689afddbf", + "IPY_MODEL_399b0e04754c4b88b01ac908bd17c128", + "IPY_MODEL_3d536c5e0d9746b0ac7ba908908e6c46" + ], + "layout": "IPY_MODEL_4e8685101b26483685c9dd822b065d3e" + } + }, + "216a21cb431c4efcb25118d60ce55382": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "21b3c6aceb5741dd847eed7955c47f17": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "22683400287c4dab86cb868a40f3ebd2": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "22ae6e2ec2474ce0b6b0bf6ca3039070": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_a07eb638da684d70a5f7760a5cd030da", + "max": 4999802720, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_a8e91c03e59940c9a47f429da4488f53", + "value": 4999802720 + } + }, + "22f1bca162774d1695505997d0b97bc6": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "2711dbd8740b453c849026f3d83f47af": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_0d8f09aba9514921baba774dcc1dd088", + "IPY_MODEL_4a10cac721514d30b6c10978d33c6e72", + "IPY_MODEL_f9d1ba01c0054ae3917a78e420174f2e" + ], + "layout": "IPY_MODEL_e95f552f62344e52aa2494fbaa0e1f16" + } + }, + "289470951d374d18948c3e43ec2282df": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "2ad775a43a284f489a7f7fef1c20dbb9": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_79dbb408b3cf498895c8b066dd04b1c5", + "max": 55351, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_924bb29fc7134a0b98187770d0943d3d", + "value": 55351 + } + }, + "2bec34b01cfb452787013ab7bcef1be3": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "2c7826be2c25431081eb05ee2d7d4401": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "2dbcf9f4ebb54761802ef3bda76f2294": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "31a933bedaad4d5489a4983bd6ca9e6f": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "31f68d4b6f624040899436006275613d": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "326286fc11c84d2092a88555ef8a8aed": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "3562239e91694d618538fabbf55bdc7e": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_037a38e5d42b4643a40e125d1ca7aa9d", + "placeholder": "​", + "style": "IPY_MODEL_114ae2bee64b4ca09b0f06e8d7f154c7", + "value": "config.json: 100%" + } + }, + "36897c09ed284e9ebad8b610186c5c2f": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_4ace0ac8c70b43f9b524e82b7102b492", + "placeholder": "​", + "style": "IPY_MODEL_6ee18cbecc5c446782764a7a7a2093ed", + "value": " 855/855 [00:00<00:00, 35.6kB/s]" + } + }, + "3752691ba21845b38cefd4e26d88e285": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_c9f51294dfa046db9ec6d89a3e2d8dd4", + "IPY_MODEL_440508c6e1f44788aa8317b07619ad6a", + "IPY_MODEL_77e2f21d581c4779856b1cca34a61c87" + ], + "layout": "IPY_MODEL_22683400287c4dab86cb868a40f3ebd2" + } + }, + "38566e74ed4143408584dce8446bd70d": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_d68f54f5cee24037b5691cb3735e5248", + "IPY_MODEL_8e2f22b60ff34706bd0503fc0b10cf24", + "IPY_MODEL_15f310a76aea48b7a6b5efc87558bb90" + ], + "layout": "IPY_MODEL_e775d3ddb0f54370a39b806024a7102c" + } + }, + "38ee56d491c74186a2146cac776a631b": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "399b0e04754c4b88b01ac908bd17c128": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_6cfbb81314f8461fa098d86ff49516d3", + "max": 4976698672, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_b870a05e12784e81aadf2e804cefe916", + "value": 4976698672 + } + }, + "3d3301ac71ed4e988deeafb5b29b1e7e": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "3d536c5e0d9746b0ac7ba908908e6c46": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_baa7428cdefa42e6bb894448a043b41e", + "placeholder": "​", + "style": "IPY_MODEL_289470951d374d18948c3e43ec2282df", + "value": " 4.98G/4.98G [01:33<00:00, 71.4MB/s]" + } + }, + "3d71197f73b540cc8a0702165b336ee6": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "440508c6e1f44788aa8317b07619ad6a": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_97f18919e94e4f0281c693d1c748e7c7", + "max": 4, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_a602bd2f6c3041d28324656a52d2527f", + "value": 4 + } + }, + "4697e1265559409f82f463f339a5f62c": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "46f37131f36a413badffb0c362ab8b5d": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "4a10cac721514d30b6c10978d33c6e72": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_2c7826be2c25431081eb05ee2d7d4401", + "max": 296, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_5c2b5412ceb54502b5018f2b756aa291", + "value": 296 + } + }, + "4a669b8562a449afb6f0a085d258040a": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_54d9fd29bff64359b5cc6d41551dec43", + "IPY_MODEL_1aac3f0e39e94eb18d5544ababc48377", + "IPY_MODEL_a71122fe861e4a819c4d79b4ba47f361" + ], + "layout": "IPY_MODEL_5f8d48cfdadc4791aff0a525ae217c34" + } + }, + "4ace0ac8c70b43f9b524e82b7102b492": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "4e0a5f3cf25b4d4bb4248f3a686a3b2b": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "4e8685101b26483685c9dd822b065d3e": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "4ee7c6a29caa45589987d7da4fcebca2": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "52ba48ea1dfc4266a57b8b70aefdc44c": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "530fe1f6d93d49749a82ea977673260f": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "53227771ba154abc9a682e9a1e18ede9": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_faa55a15d9214cfb966a96f33a96b31d", + "max": 4915916176, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_8a29a11e60b24404a6b53573f81085ce", + "value": 4915916176 + } + }, + "54849eb44d73489bb7835b1f98e81995": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "54d9fd29bff64359b5cc6d41551dec43": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_081e77e2c61f49babc1f48da215386d8", + "placeholder": "​", + "style": "IPY_MODEL_5964a39c2f4b46ea9f2a24c607a33160", + "value": "Loading checkpoint shards: 100%" + } + }, + "56f3290340934a28bc43e73fe5e8d050": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "5964a39c2f4b46ea9f2a24c607a33160": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "5c2b5412ceb54502b5018f2b756aa291": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "5e74177710b1492292796c2bf90ee8e4": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_c56d8b6ad9d64050ac6f26b33d142db8", + "IPY_MODEL_cdf784f05efb411a927d5ad1ac363046", + "IPY_MODEL_20341d8846494bc7a0d3bff2620534ac" + ], + "layout": "IPY_MODEL_7031154b6c514410aa517a287d0bb90f" + } + }, + "5f66742ac0ee4404959a83b4362aeb42": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "5f8d48cfdadc4791aff0a525ae217c34": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "600eee97f1fb478b919d5755baac574d": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "6cfbb81314f8461fa098d86ff49516d3": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "6e0a43d7cf314faba17bd88e1e37f021": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "6e115f191b664f59b8144719ac6d6d13": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "6ee18cbecc5c446782764a7a7a2093ed": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "6f64c037d228441489da9729d0eb45c0": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "7031154b6c514410aa517a287d0bb90f": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "77e2f21d581c4779856b1cca34a61c87": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_ee2db85a296d4a9bb6895c8438b4ef21", + "placeholder": "​", + "style": "IPY_MODEL_530fe1f6d93d49749a82ea977673260f", + "value": " 4/4 [10:24<00:00, 138.93s/it]" + } + }, + "7836c3b013ec4f67a2b09670878581ba": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_0e1f48ecaf6e4184953ec773bed24253", + "max": 9085657, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_38ee56d491c74186a2146cac776a631b", + "value": 9085657 + } + }, + "79dbb408b3cf498895c8b066dd04b1c5": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "7e42a603efbe4d4bacb61114ca8bfdaf": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "83cdabe816e44bc6964ba96dea753e15": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "85c5b698708b47319067527bda04445f": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "86c78bcedbfd496a85a5d5ddb9c1df85": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_0545d6dd41964f799d689964b3e2a357", + "IPY_MODEL_22ae6e2ec2474ce0b6b0bf6ca3039070", + "IPY_MODEL_c311192f10264e399bc2e77b1aabb7d2" + ], + "layout": "IPY_MODEL_f39be376072c43f7bc2dd1aff4eeb777" + } + }, + "87ba3449a7ce4bda8730f2d7c26b5f3f": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "8a29a11e60b24404a6b53573f81085ce": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "8e2f22b60ff34706bd0503fc0b10cf24": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_d9c4f5565a2f4faf9598c4377fc975ac", + "max": 1168138808, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_9626d058aa584a9e829d7793676eb8cd", + "value": 1168138808 + } + }, + "8e93a17e01b64ff89f3220565ada3082": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "90d6add2cac446958326cb3cb4defc35": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_4e0a5f3cf25b4d4bb4248f3a686a3b2b", + "max": 855, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_3d71197f73b540cc8a0702165b336ee6", + "value": 855 + } + }, + "9181116e0dab421e8538c3b5e6d78a63": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_3d3301ac71ed4e988deeafb5b29b1e7e", + "placeholder": "​", + "style": "IPY_MODEL_4ee7c6a29caa45589987d7da4fcebca2", + "value": "model.safetensors.index.json: 100%" + } + }, + "924bb29fc7134a0b98187770d0943d3d": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "9626d058aa584a9e829d7793676eb8cd": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "97f18919e94e4f0281c693d1c748e7c7": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "9ca5e65c356543df95542ddf1cb87ab9": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "9dde77220ce04b80ac2b999d9304ac61": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_3562239e91694d618538fabbf55bdc7e", + "IPY_MODEL_90d6add2cac446958326cb3cb4defc35", + "IPY_MODEL_36897c09ed284e9ebad8b610186c5c2f" + ], + "layout": "IPY_MODEL_19f90c9b7f6b4fb4b7e4db0fda3697e6" + } + }, + "a07eb638da684d70a5f7760a5cd030da": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "a602bd2f6c3041d28324656a52d2527f": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "a71122fe861e4a819c4d79b4ba47f361": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_f08d87275f5243e48e5277291133655c", + "placeholder": "​", + "style": "IPY_MODEL_326286fc11c84d2092a88555ef8a8aed", + "value": " 4/4 [01:36<00:00, 20.71s/it]" + } + }, + "a8e91c03e59940c9a47f429da4488f53": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "aa52d70d64b7456eb99cc9386a6b773b": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "b0fece23eefb4f069ffd2282bf05aa50": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "b12dc89359524e9e8edc2b90d68d6076": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "b1a13112ead34742835e635910820dd0": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_161f9b5e075b404c871a58fbd560d63d", + "placeholder": "​", + "style": "IPY_MODEL_2dbcf9f4ebb54761802ef3bda76f2294", + "value": " 55.4k/55.4k [00:00<00:00, 5.22MB/s]" + } + }, + "b7f4683f24f6497084de01fc9dcb7b4c": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_6f64c037d228441489da9729d0eb45c0", + "placeholder": "​", + "style": "IPY_MODEL_092185914174457cb395e16f5dfd4774", + "value": " 23.9k/23.9k [00:00<00:00, 2.27MB/s]" + } + }, + "b870a05e12784e81aadf2e804cefe916": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "ProgressStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "ProgressStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "bar_color": null, + "description_width": "" + } + }, + "b9d1f51b79d34264965e512bfc0253b3": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "baa7428cdefa42e6bb894448a043b41e": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "bd0133b89f784f16ab20ff28574d537b": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_31f68d4b6f624040899436006275613d", + "placeholder": "​", + "style": "IPY_MODEL_8e93a17e01b64ff89f3220565ada3082", + "value": "tokenizer_config.json: 100%" + } + }, + "c1d3120bad3d456285a0807445d5e4ee": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_df49ac181ec244f5bebe6680408fd7ee", + "IPY_MODEL_7836c3b013ec4f67a2b09670878581ba", + "IPY_MODEL_f3467a09b50a4b429ff75009a26d992b" + ], + "layout": "IPY_MODEL_b9d1f51b79d34264965e512bfc0253b3" + } + }, + "c264cb454c7d43ccb2dd6045732f0317": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "c311192f10264e399bc2e77b1aabb7d2": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_b12dc89359524e9e8edc2b90d68d6076", + "placeholder": "​", + "style": "IPY_MODEL_fe2dd51da47e4130a3d5713037784787", + "value": " 5.00G/5.00G [04:35<00:00, 15.9MB/s]" + } + }, + "c56d8b6ad9d64050ac6f26b33d142db8": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_21b3c6aceb5741dd847eed7955c47f17", + "placeholder": "​", + "style": "IPY_MODEL_b0fece23eefb4f069ffd2282bf05aa50", + "value": "generation_config.json: 100%" + } + }, + "c9f51294dfa046db9ec6d89a3e2d8dd4": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_f451e2532e5241448a85daefb01822e2", + "placeholder": "​", + "style": "IPY_MODEL_f03f977f28364ec18ce1659c76aa82ba", + "value": "Downloading shards: 100%" + } + }, + "ca359ec98ebc49f8b121dec67853f797": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "cb53de4aa61843b0a75ead30dfece607": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "cdf784f05efb411a927d5ad1ac363046": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "FloatProgressModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "FloatProgressModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "ProgressView", + "bar_style": "success", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_52ba48ea1dfc4266a57b8b70aefdc44c", + "max": 184, + "min": 0, + "orientation": "horizontal", + "style": "IPY_MODEL_aa52d70d64b7456eb99cc9386a6b773b", + "value": 184 + } + }, + "d0306eea21a045f7a169d64d615bbf6b": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_f435e52db79b4317b55f13b8c98fe766", + "IPY_MODEL_53227771ba154abc9a682e9a1e18ede9", + "IPY_MODEL_e23aba3a1ac4420da2b0570a0dea2011" + ], + "layout": "IPY_MODEL_7e42a603efbe4d4bacb61114ca8bfdaf" + } + }, + "d2b3737cd56941e79b22af8689afddbf": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_22f1bca162774d1695505997d0b97bc6", + "placeholder": "​", + "style": "IPY_MODEL_5f66742ac0ee4404959a83b4362aeb42", + "value": "model-00001-of-00004.safetensors: 100%" + } + }, + "d68f54f5cee24037b5691cb3735e5248": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_54849eb44d73489bb7835b1f98e81995", + "placeholder": "​", + "style": "IPY_MODEL_1ed77b06007b426b8944b2748431402d", + "value": "model-00004-of-00004.safetensors: 100%" + } + }, + "d9c4f5565a2f4faf9598c4377fc975ac": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "df49ac181ec244f5bebe6680408fd7ee": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_06722a988e1a4357a3841723aa66dfa8", + "placeholder": "​", + "style": "IPY_MODEL_46f37131f36a413badffb0c362ab8b5d", + "value": "tokenizer.json: 100%" + } + }, + "e18941e122eb4332b9e5fb1c3a22b9ed": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "e23aba3a1ac4420da2b0570a0dea2011": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_f7e02670ead24d12b09aa36cdf1ad98f", + "placeholder": "​", + "style": "IPY_MODEL_ca359ec98ebc49f8b121dec67853f797", + "value": " 4.92G/4.92G [03:51<00:00, 52.9MB/s]" + } + }, + "e59cfa5ca5b240159517ccc2a594fa50": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "e775d3ddb0f54370a39b806024a7102c": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "e95f552f62344e52aa2494fbaa0e1f16": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "ea8987a8a81a4c1ca33ba1f6d35eabf7": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HBoxModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HBoxModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HBoxView", + "box_style": "", + "children": [ + "IPY_MODEL_9181116e0dab421e8538c3b5e6d78a63", + "IPY_MODEL_06fc98bc230248618f718cc789860551", + "IPY_MODEL_b7f4683f24f6497084de01fc9dcb7b4c" + ], + "layout": "IPY_MODEL_216a21cb431c4efcb25118d60ce55382" + } + }, + "eb0e9fb2bf8447748a936749dceff1b9": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "ee2db85a296d4a9bb6895c8438b4ef21": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "f03f977f28364ec18ce1659c76aa82ba": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + }, + "f08d87275f5243e48e5277291133655c": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "f3467a09b50a4b429ff75009a26d992b": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_e59cfa5ca5b240159517ccc2a594fa50", + "placeholder": "​", + "style": "IPY_MODEL_e18941e122eb4332b9e5fb1c3a22b9ed", + "value": " 9.09M/9.09M [00:01<00:00, 5.89MB/s]" + } + }, + "f39be376072c43f7bc2dd1aff4eeb777": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "f435e52db79b4317b55f13b8c98fe766": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_c264cb454c7d43ccb2dd6045732f0317", + "placeholder": "​", + "style": "IPY_MODEL_87ba3449a7ce4bda8730f2d7c26b5f3f", + "value": "model-00003-of-00004.safetensors: 100%" + } + }, + "f451e2532e5241448a85daefb01822e2": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "f7e02670ead24d12b09aa36cdf1ad98f": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "f9d1ba01c0054ae3917a78e420174f2e": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "HTMLModel", + "state": { + "_dom_classes": [], + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "HTMLModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/controls", + "_view_module_version": "1.5.0", + "_view_name": "HTMLView", + "description": "", + "description_tooltip": null, + "layout": "IPY_MODEL_001b683aec074eb98b498f0dec667b7c", + "placeholder": "​", + "style": "IPY_MODEL_eb0e9fb2bf8447748a936749dceff1b9", + "value": " 296/296 [00:00<00:00, 32.7kB/s]" + } + }, + "faa55a15d9214cfb966a96f33a96b31d": { + "model_module": "@jupyter-widgets/base", + "model_module_version": "1.2.0", + "model_name": "LayoutModel", + "state": { + "_model_module": "@jupyter-widgets/base", + "_model_module_version": "1.2.0", + "_model_name": "LayoutModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "LayoutView", + "align_content": null, + "align_items": null, + "align_self": null, + "border": null, + "bottom": null, + "display": null, + "flex": null, + "flex_flow": null, + "grid_area": null, + "grid_auto_columns": null, + "grid_auto_flow": null, + "grid_auto_rows": null, + "grid_column": null, + "grid_gap": null, + "grid_row": null, + "grid_template_areas": null, + "grid_template_columns": null, + "grid_template_rows": null, + "height": null, + "justify_content": null, + "justify_items": null, + "left": null, + "margin": null, + "max_height": null, + "max_width": null, + "min_height": null, + "min_width": null, + "object_fit": null, + "object_position": null, + "order": null, + "overflow": null, + "overflow_x": null, + "overflow_y": null, + "padding": null, + "right": null, + "top": null, + "visibility": null, + "width": null + } + }, + "fe2dd51da47e4130a3d5713037784787": { + "model_module": "@jupyter-widgets/controls", + "model_module_version": "1.5.0", + "model_name": "DescriptionStyleModel", + "state": { + "_model_module": "@jupyter-widgets/controls", + "_model_module_version": "1.5.0", + "_model_name": "DescriptionStyleModel", + "_view_count": null, + "_view_module": "@jupyter-widgets/base", + "_view_module_version": "1.2.0", + "_view_name": "StyleView", + "description_width": "" + } + } + } + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/environment.yml b/environment.yml index 470b64b..3d39950 100644 --- a/environment.yml +++ b/environment.yml @@ -44,3 +44,4 @@ dependencies: - twilio - pydub - protobuf==3.20.2 + - wandb diff --git a/requirements.txt b/requirements.txt index edcb3de..9d62d7d 100644 --- a/requirements.txt +++ b/requirements.txt @@ -36,3 +36,4 @@ speedtest-cli sentence_transformers feedparser protobuf==3.20.2 +wandb diff --git a/week1/community-contributions/AI_Property_Assistant/README.md b/week1/community-contributions/AI_Property_Assistant/README.md new file mode 100644 index 0000000..b6d4405 --- /dev/null +++ b/week1/community-contributions/AI_Property_Assistant/README.md @@ -0,0 +1,162 @@ +# AI Property Rental Assistant + +A Python tool that scrapes UK property rental listings and uses OpenAI's GPT-4o-mini to provide personalized property recommendations based on your requirements. + +## What It Does + +- Scrapes property listings from OnTheMarket.com +- Uses AI to analyze properties against your specific needs +- Provides smart recommendations with reasons why properties match +- Works for any UK location (currently configured for Durham) + +## Quick Start + +### Prerequisites +- Python 3.7+ +- OpenAI API key ([Get one here](https://platform.openai.com/api-keys)) + +### Installation + +1. **Install required packages:** + ```bash + pip install requests beautifulsoup4 openai python-dotenv ipython + ``` + +2. **Set up your API key:** + + Create a `.env` file in the same directory as your script: + ``` + OPENAI_API_KEY=your_openai_api_key_here + ``` + +3. **Run the script:** + ```bash + python your_script_name.py + ``` + +## How to Use + +### Basic Usage + +The script is pre-configured to search for student housing in Durham. Just run it and you'll get AI-powered recommendations! + +### Customizing Your Search + +**Change the location:** +```python +website_url = "https://www.onthemarket.com/to-rent/property/manchester/" +``` + +**Update your requirements:** +```python +user_needs = "I'm a young professional looking for a 1-bedroom flat in Manchester under £1,000/month" +``` + +### Example Requirements You Can Use: +- `"Student looking for shared accommodation under £600/month"` +- `"Family needing 3-bedroom house with garden under £1,500/month"` +- `"Professional couple wanting modern 2-bed apartment near city center"` + +## Configuration + +### Supported Cities +Replace `durham` in the URL with any UK city: +- `london` - London properties +- `manchester` - Manchester properties +- `birmingham` - Birmingham properties +- `leeds` - Leeds properties +- `bristol` - Bristol properties + +### AI Behavior +The system prompt is configured for UK rentals but you can modify it in the `system_prompt` variable to: +- Focus on specific property types +- Emphasize certain features (parking, garden, etc.) +- Target different tenant types (students, families, professionals) + +## Example Output + +``` +Website Title: Properties to rent in Durham - OnTheMarket +Content Length: 15847 characters + +================================================== +RENTAL RECOMMENDATIONS: +================================================== + +# Property Recommendations for Durham + +Based on your requirements for a 2-bedroom student property under £2,000/month, here are my top recommendations: + +## 1. **Student House on North Road** - £1,600/month +- **Bedrooms:** 2 +- **Perfect because:** Well within budget, popular student area +- **Features:** Close to university, furnished, bills included + +## 2. **Modern Apartment City Centre** - £1,400/month +- **Bedrooms:** 2 +- **Perfect because:** Great location, modern amenities +- **Features:** Parking space, balcony, near shops +``` + +## Requirements + +Create a `requirements.txt` file: +``` +requests>=2.28.0 +beautifulsoup4>=4.11.0 +openai>=1.0.0 +python-dotenv>=0.19.0 +ipython>=8.0.0 +``` + +Install with: `pip install -r requirements.txt` + +## Important Notes + +### API Costs +- Uses GPT-4o-mini model (very affordable - ~$0.001 per request) +- Monitor usage at: https://platform.openai.com/usage + +### Rate Limits +- Free OpenAI accounts: 3 requests per minute +- The script makes 1 request per run + +## How It Works + +1. **Web Scraping:** Downloads the property listing page +2. **Text Extraction:** Cleans HTML and extracts property information +3. **AI Analysis:** Sends your requirements + property data to GPT-4 +4. **Smart Recommendations:** AI filters and ranks properties with explanations + +## Troubleshooting + +**"No API key found"** +- Make sure `.env` file exists in the same folder as your script +- Check the API key has no extra spaces +- Verify it starts with `sk-proj-` + +**"Error fetching website"** +- Check your internet connection +- Try a different city URL +- Some websites may temporarily block requests + +**No good recommendations** +- Try adjusting your budget or requirements +- Check if the website loaded properly (look at content length) +- Try a different city with more properties + +## Possible Improvements + +- Make it interactive (ask for user input) +- Support multiple property websites +- Add price tracking over time +- Include property images in analysis +- Create a simple web interface + +## Disclaimer + +This tool is for educational purposes. Always verify property information directly with landlords or estate agents. Respect website terms of service. + +--- + +**Need help?** Check that your `.env` file is set up correctly and you have an active internet connection. The script will tell you if there are any issues with your API key! \ No newline at end of file diff --git a/week1/community-contributions/AI_Property_Assistant/rental_property_scraper.ipynb b/week1/community-contributions/AI_Property_Assistant/rental_property_scraper.ipynb new file mode 100644 index 0000000..9b6c11f --- /dev/null +++ b/week1/community-contributions/AI_Property_Assistant/rental_property_scraper.ipynb @@ -0,0 +1,294 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "0993b44c-58f3-4d7d-ac31-80871a867040", + "metadata": {}, + "source": [ + "# AI Property Rental Assistant for Durham\n", + "This notebook creates an intelligent property rental assistant that:\n", + "1. Scrapes rental property listings from OnTheMarket.com\n", + "2. Uses OpenAI's GPT-4o-mini to analyze and recommend properties based on user preferences\n", + "3. Provides formatted recommendations in markdown for easy reading\n", + "\n", + "Purpose: Help students and professionals find suitable rental properties in Durham, UK" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f3fa597-bac5-496f-b0c6-ac1cb524062d", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dfa715c4-81d4-4f1e-87d8-6cf7fa17db71", + "metadata": {}, + "outputs": [], + "source": [ + "# =====================================\n", + "# STEP 1: ENVIRONMENT SETUP & API KEYS\n", + "# =====================================\n", + "\n", + "# Load environment variables from .env file\n", + "# Make sure you have a .env file with: OPENAI_API_KEY=your_key_here\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Validate the OpenAI API key format and existence\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n", + "\n", + "# Initialize OpenAI client\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c7e44572-1cda-42d2-a6ff-45f462fd436f", + "metadata": {}, + "outputs": [], + "source": [ + "# =====================================\n", + "# STEP 2: WEB SCRAPING SETUP\n", + "# =====================================\n", + "\n", + "# HTTP headers to mimic a real browser request\n", + "# Many websites block requests without proper headers\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A class to represent and scrape content from a webpage.\n", + " \n", + " This class handles:\n", + " - Fetching HTML content from a URL\n", + " - Parsing HTML with BeautifulSoup\n", + " - Extracting clean text content (removing scripts, styles, etc.)\n", + " - Error handling for failed requests\n", + " \n", + " Attributes:\n", + " url (str): The URL of the website\n", + " title (str): The page title\n", + " text (str): Clean text content from the page body\n", + " \"\"\"\n", + " \n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Initialize Website object by scraping content from the given URL.\n", + " \n", + " Args:\n", + " url (str): The website URL to scrape\n", + " \"\"\"\n", + " self.url = url\n", + " try:\n", + " # Make HTTP request with timeout to prevent hanging\n", + " response = requests.get(url, headers=headers, timeout=10)\n", + " response.raise_for_status() # Raises an HTTPError for bad responses\n", + " \n", + " # Parse HTML content\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " \n", + " # Extract page title\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " \n", + " # Clean up the HTML by removing irrelevant elements\n", + " if soup.body:\n", + " # Remove scripts, styles, images, and input elements\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " \n", + " # Extract clean text with proper line separation\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"No body content found\"\n", + " \n", + " except requests.RequestException as e:\n", + " # Handle network errors gracefully\n", + " print(f\"Error fetching website: {e}\")\n", + " self.title = \"Error loading page\"\n", + " self.text = \"Could not load page content\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a97d9c34-2831-4730-949e-bba1b6ac9bb3", + "metadata": {}, + "outputs": [], + "source": [ + "# =====================================\n", + "# STEP 3: AI ASSISTANT FUNCTIONS\n", + "# =====================================\n", + "\n", + "def house_renting(system_prompt, user_prompt):\n", + " \"\"\"\n", + " Send prompts to OpenAI's GPT model and get rental recommendations.\n", + " \n", + " This function:\n", + " - Formats the conversation for the AI model\n", + " - Sends requests to GPT-4o-mini (cost-effective model)\n", + " - Returns the AI's response as a string\n", + " \n", + " Args:\n", + " system_prompt (str): Instructions for how the AI should behave\n", + " user_prompt (str): The user's specific request with property data\n", + " \n", + " Returns:\n", + " str: AI-generated rental recommendations in markdown format\n", + " \"\"\"\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + " \n", + " # Call OpenAI API\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\", # Cost-effective model, good for this task\n", + " messages=messages,\n", + " )\n", + " \n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6d0c4b96-b907-45ed-8a4d-a67d8f7e4f33", + "metadata": {}, + "outputs": [], + "source": [ + "# =====================================\n", + "# STEP 4: AI SYSTEM CONFIGURATION\n", + "# =====================================\n", + "\n", + "# Define how the AI assistant should behave\n", + "# This is crucial for getting consistent, helpful responses\n", + "system_prompt = \"\"\"\n", + "You are a helpful real estate assistant specializing in UK property rentals. Your job is to guide users in finding houses to rent, especially in Durham. Follow these rules:\n", + "\n", + "1. Always ask clarifying questions if user input is vague. Determine location, budget, number of bedrooms, and tenant type (e.g. student, family, professional).\n", + "2. Use structured data provided from the website (like property listings) to identify relevant options.\n", + "3. If listings are provided, filter and rank them based on the user's preferences.\n", + "4. Recommend up to 5 top properties with rent price, bedroom count, key features, and location.\n", + "5. Always respond in markdown with clean formatting using headers, bold text, and bullet points.\n", + "6. If no listings match well, provide tips (e.g. \"try adjusting your budget or search radius\").\n", + "7. Stay concise, helpful, and adapt to whether the user is a student, family, couple, or solo tenant.\n", + "\"\"\"\n", + "\n", + "def user_prompt_for_renting(website, user_needs):\n", + " \"\"\"\n", + " Create a formatted prompt that combines user requirements with scraped property data.\n", + " \n", + " This function:\n", + " - Takes user preferences and website content\n", + " - Formats them into a clear prompt for the AI\n", + " - Limits content to first 4000 characters to stay within token limits\n", + " \n", + " Args:\n", + " website (Website): The scraped website object\n", + " user_needs (str): Description of what the user is looking for\n", + " \n", + " Returns:\n", + " str: Formatted prompt ready to send to the AI\n", + " \"\"\"\n", + " user_prompt = f\"\"\"\n", + "I want to rent a house and here's what I'm looking for:\n", + "{user_needs}\n", + "\n", + "Here are the property listings I found on the website titled: \"{website.title}\".\n", + "Please analyze them and recommend the best 3–5 options that match my needs. If none are suitable, tell me why and offer suggestions.\n", + "\n", + "The page content is below:\n", + "{website.text[:4000]} # Truncated to first 4000 characters to manage token usage\n", + "\"\"\"\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cecb1f11-060a-4737-828c-e94ae04a42ae", + "metadata": {}, + "outputs": [], + "source": [ + "# =====================================\n", + "# STEP 5: MAIN EXECUTION\n", + "# =====================================\n", + "\n", + "if __name__ == \"__main__\":\n", + " print(\"Starting AI Property Rental Assistant...\")\n", + " print(\"=\" * 50)\n", + " \n", + " # Configure the property search\n", + " website_url = \"https://www.onthemarket.com/to-rent/property/durham/\"\n", + " print(f\"🔍 Scraping properties from: {website_url}\")\n", + " \n", + " # Scrape the website\n", + " website = Website(website_url)\n", + " \n", + " # Display scraping results\n", + " print(f\"Website Title: {website.title}\")\n", + " print(f\"Content Length: {len(website.text)} characters\")\n", + " print(f\"Successfully scraped property listings\")\n", + " print()\n", + " \n", + " # Define user requirements\n", + " # TODO: Make this interactive by adding input() statements\n", + " user_needs = \"I'm a student looking for a 2-bedroom house in Durham under £2,000/month\"\n", + " print(f\"User Requirements: {user_needs}\")\n", + " print()\n", + " \n", + " # Generate AI prompt\n", + " user_prompt = user_prompt_for_renting(website, user_needs)\n", + " \n", + " # Get AI recommendations\n", + " print(\"Generating AI recommendations...\")\n", + " output = house_renting(system_prompt, user_prompt)\n", + " \n", + " # Display results\n", + " display(Markdown(output))\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python [conda env:llms]", + "language": "python", + "name": "conda-env-llms-py" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Business_Brochure_project-Copy1.ipynb b/week1/community-contributions/Business_Brochure_project-Copy1.ipynb new file mode 100644 index 0000000..9a19b5b --- /dev/null +++ b/week1/community-contributions/Business_Brochure_project-Copy1.ipynb @@ -0,0 +1,425 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "d5b08506-dc8b-4443-9201-5f1848161363", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9ce95b0e-3f98-4cef-b9fe-d963fdeeed2d", + "metadata": {}, + "outputs": [], + "source": [ + "# run ollama serve in your Anaconda terminal\n", + "MODEL = \"llama3.2\"\n", + "openai = OpenAI(base_url=\"http://localhost:11434/v1\", api_key=\"ollama\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "106dd65e-90af-4ca8-86b6-23a41840645b", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e30d8128-933b-44cc-81c8-ab4c9d86589a", + "metadata": {}, + "outputs": [], + "source": [ + "ed = Website(\"https://edwarddonner.com\")\n", + "ed.links" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6957b079-0d96-45f7-a26a-3487510e9b35", + "metadata": {}, + "outputs": [], + "source": [ + "link_system_prompt = \"You are provided with a list of links found on a webpage. \\\n", + "You are able to decide which of the links would be most relevant to include in a brochure about the company, \\\n", + "such as links to an About page, or a Company page, or Careers/Jobs pages.\\n\"\n", + "link_system_prompt += \"You should respond in JSON as in these examples:\"\n", + "link_system_prompt += \"\"\"example 1:\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"about page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"careers page\", \"url\": \"https://another.full.url/careers\"}]}\n", + "\n", + "Example 2:\n", + " {\"links\": [{\"type\": \"about page\", \"url\": \"https://great-comps.de/about-me\"},\n", + " {\"type\": \"news page\": \"url\": \"https://great-comps.de/news\"},\n", + " {\"type\": \"case studies page\": \"url\": \"https://great-comps.de/case-studies\"},\n", + " {\"type\": \"workshop page\": \"url\": \"https://great-comps.de/workshop-ai\"},]}\n", + "\n", + "\n", + "Example 3:\n", + " {\"links\": [{\"type\": \"über mich\", \"url\": \"https://wahlen-robbie.at/ueber-mich\"},\n", + " {\"type\": \"aktuelles\": \"url\": \"https://wahlen-robbie.at/neuigkeiten\"},\n", + " {\"type\": \"whitepaper\": \"url\": \"https://wahlen-robbie.at/whitepapers\"},\n", + " {\"type\": \"services\": \"url\": \"https://wahlen-robbie.at/services\"}]}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b97e4068-97ed-4120-beae-c42105e4d59a", + "metadata": {}, + "outputs": [], + "source": [ + "print(link_system_prompt)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8e1f601b-2eaf-499d-b6b8-c99050c9d6b3", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links_user_prompt(website):\n", + " user_prompt = f\"Here is the list of links on the website of {website.url} - \"\n", + " user_prompt += \"please decide which of these are relevant web links for a brochure about the company, respond with the full https URL in JSON format. \\\n", + "Do not include Terms of Service, Privacy, email links.\\n\"\n", + " user_prompt += \"Links (some might be relative links):\\n\"\n", + " user_prompt += \"\\n\".join(website.links)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6bcbfa78-6395-4685-b92c-22d592050fd7", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_links_user_prompt(ed))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a29aca19-ca13-471c-a4b4-5abbfa813f69", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": link_system_prompt},\n", + " {\"role\": \"user\", \"content\": get_links_user_prompt(website)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "74a827a0-2782-4ae5-b210-4a242a8b4cc2", + "metadata": {}, + "outputs": [], + "source": [ + "# Getting the links for HuggingFace\n", + "huggingface = Website(\"https://huggingface.co\")\n", + "huggingface.links" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3d583e2-dcc4-40cc-9b28-1e8dbf402924", + "metadata": {}, + "outputs": [], + "source": [ + "get_links(\"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85a5b6e2-e7ef-44a9-bc7f-59ede71037b5", + "metadata": {}, + "outputs": [], + "source": [ + "def get_all_details(url):\n", + " result = \"Landing page:\\n\"\n", + " result += Website(url).get_contents()\n", + " links = get_links(url)\n", + " #print(\"Found links:\", links)\n", + " for link in links[\"links\"]:\n", + " result += f\"\\n\\n{link['type']}\\n\"\n", + " result += Website(link[\"url\"]).get_contents()\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5099bd14-076d-4745-baf3-dac08d8e5ab2", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_all_details(\"https://edwarddonner.com\"))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9b863a55-f86c-4e3f-8a79-94e24c1a8cf2", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "and creates a short brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "Include details of company culture, customers and careers/jobs if you have the information.\"\n", + "\n", + "# Or uncomment the lines below for a more humorous brochure - this demonstrates how easy it is to incorporate 'tone':\n", + "\n", + "# system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "# and creates a short humorous, entertaining, jokey brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "# Include details of company culture, customers and careers/jobs if you have the information.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6ab83d92-d36b-4ce0-8bcc-5bb4c2f8ff23", + "metadata": {}, + "outputs": [], + "source": [ + "def get_brochure_user_prompt(company_name, url):\n", + " user_prompt = f\"You are looking at a company called: {company_name}\\n\"\n", + " user_prompt += f\"Here are the contents of its landing page and other relevant pages; use this information to build a short brochure of the company in markdown.\\n\"\n", + " user_prompt += get_all_details(url)\n", + " user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cd909e0b-1312-4ce2-a553-821e795d7572", + "metadata": {}, + "outputs": [], + "source": [ + "get_brochure_user_prompt(\"ed\", \"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e44de579-4a1a-4e6a-a510-20ea3e4b8d46", + "metadata": {}, + "outputs": [], + "source": [ + "def create_brochure(company_name, url):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e093444a-9407-42ae-924a-145730591a39", + "metadata": {}, + "outputs": [], + "source": [ + "create_brochure(\"ed\", \"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fbc1a523-4c5c-4571-8541-456e5feed185", + "metadata": {}, + "outputs": [], + "source": [ + "def create_brochure_translated(company_name, url, language):\n", + " system_prompt = f\"\"\"You are an expert professional translator. Translate the following brochure text into {language}.\n", + " Make sure to translate into a idiomatic {language}, matching the users language's natural structure, wording and expressions,\n", + " so it can't be recognised as a translation.\n", + " Be sure to also maintain an appropriate tone, Output the translated brochure in Markdown format.\"\"\"\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " \n", + " )\n", + " result = response.choices[0].message.content\n", + " display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6ff2ddc4-f13c-44b2-8043-d66f36b0954c", + "metadata": {}, + "outputs": [], + "source": [ + "create_brochure_translated(\"ed\", \"https://edwarddonner.com\", \"French\")" + ] + }, + { + "cell_type": "markdown", + "id": "61eaaab7-0b47-4b29-82d4-75d474ad8d18", + "metadata": {}, + "source": [ + "## Finally - a minor improvement\n", + "\n", + "With a small adjustment, we can change this so that the results stream back from OpenAI,\n", + "with the familiar typewriter animation" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "51db0e49-f261-4137-aabe-92dd601f7725", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_brochure(company_name, url):\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "09f18581-b870-4952-8430-217afaf7a83e", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_brochure_translated(company_name, url, language):\n", + " system_prompt = f\"\"\"You are an expert professional translator. \n", + " Translate the following brochure text into {language}.\n", + " Make sure to translate into idiomatic {language}, matching the natural structure, wording, and expressions.\n", + " Maintain an appropriate tone. Output the translated brochure in Markdown format.\"\"\"\n", + "\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " stream=True\n", + " )\n", + "\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " # clean up formatting glitches while streaming\n", + " response = response.replace(\"```\", \"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "56bf0ae3-ee9d-4a72-9cd6-edcac67ceb6d", + "metadata": {}, + "outputs": [], + "source": [ + "stream_brochure_translated(\"ed\", \"https://edwarddonner.com\", \"Spanish\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Business_Use_Case_Resume_Upgrader.ipynb b/week1/community-contributions/Business_Use_Case_Resume_Upgrader.ipynb new file mode 100644 index 0000000..173dafd --- /dev/null +++ b/week1/community-contributions/Business_Use_Case_Resume_Upgrader.ipynb @@ -0,0 +1,179 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "606e9c73-50fe-46b9-8df3-ae2246c00a3e", + "metadata": {}, + "source": [ + "# Business Use Case - LLM based Resume Upgrader" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "919f6546-80ec-4d4c-8a80-00228f50e4a0", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from openai import OpenAI\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "markdown", + "id": "b2f5b02c-f782-4578-8a91-07891c39ceb0", + "metadata": {}, + "source": [ + "steps to perform\n", + "-> load API key from env file\n", + "-> create a function to call llm api\n", + "-> create messages for system prompt and user prompt\n", + "-> display the llm output" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "31aaa20e-4996-43cb-b43a-a1aef80fd391", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv()\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "# error handling\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "92f65c91-ca7f-47e6-9fd7-d63b278ba264", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "98fc7bac-07c8-4801-9225-8f843837f3c2", + "metadata": {}, + "outputs": [], + "source": [ + "# system prompt\n", + "\n", + "system_prompt = \"\"\"You are a helpful resume editor assistant that provides required assistance in changing a resume to match the given job descrption role \\\n", + "You are given a resume and job description, your job is to understand the resume and job description to suggest upto 6 missing key words in the resume. Then you have to \n", + "suggest how the user can improve his resume by giving upto 3 example sentences using the suggest keywords to fit into their resume.\n", + "by using the following structure provide your response \\\n", + "Sturcture:\n", + "Job role : [Job Role]:\n", + "Candidate Name : [Candidate Name]\n", + "Missing Key words in Resume Based on Given job description:\n", + " - [] Missing key words\n", + " -[] Missing key words\n", + "\n", + "\n", + "Suggestion:\n", + " - [] # write a sentence including the key words to put them in the resume\n", + " - [] # write a sentence including the key words to put them in the resume\n", + "\n", + "Guidelines:\n", + "- give proper keyword suggestions which are essential for the job function. Do not give any unnecesary suggestions\n", + "- Keep the suggested sentences less that 50 words\n", + "- \n", + "\"\"\"\n", + "user_prompt = f'Give me suggestions on how to improve my resume and for the given job description '\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0d9c40b5-8e27-41b9-8b88-2c83e7d2b3ec", + "metadata": {}, + "outputs": [], + "source": [ + "# call openai api\n", + "def resume_upgrader(resume:str, job_description:str):\n", + " user_prompt = f'Give me suggestions on how to improve my resume {resume} and for the given job description {job_description}'\n", + " messages = [\n", + " {'role': 'system', 'content': system_prompt},\n", + " {'role': 'user', 'content': user_prompt}\n", + " ]\n", + " try:\n", + " \n", + " response = openai.chat.completions.create(model =\"gpt-4o-mini\", messages = messages)\n", + " return response.choices[0].message.content\n", + " except:\n", + " print('got error while retting the response from api')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5aa29465-c119-4178-90f1-3ebdc9eeb11a", + "metadata": {}, + "outputs": [], + "source": [ + "def print_api_response(response_markdown):\n", + " \"\"\"Print the markdown response\"\"\"\n", + " display(Markdown(response_markdown))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "82a92034-6722-4e78-a901-b4ef2b9cbb84", + "metadata": {}, + "outputs": [], + "source": [ + "resume = input(\"Paste your resume in here\")\n", + "job_description = input(\"paste your job descritpion here\")\n", + "response = resume_upgrader(resume, job_description)\n", + "print_api_response(response)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d0be536f-e890-473f-8c68-767bc0e3b47c", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/D2-property-rental-assistant/README.md b/week1/community-contributions/D2-property-rental-assistant/README.md new file mode 100644 index 0000000..374cd58 --- /dev/null +++ b/week1/community-contributions/D2-property-rental-assistant/README.md @@ -0,0 +1,189 @@ +# AI Property Rental Assistant + +An intelligent property rental assistant Jupyter notebook that scrapes real estate listings from OnTheMarket and uses a local LLM (DeepSeek R1) to analyze and recommend properties based on user requirements. + +## Features + +- **Web Scraping**: Automatically fetches property listings from OnTheMarket +- **AI-Powered Analysis**: Uses DeepSeek R1 model via Ollama for intelligent recommendations +- **Personalized Recommendations**: Filters and ranks properties based on: + - Budget constraints + - Number of bedrooms + - Tenant type (student, family, professional) + - Location preferences +- **Clean Output**: Returns formatted markdown with top 3-5 property recommendations +- **Smart Filtering**: Handles cases where no suitable properties are found with helpful suggestions + +## Prerequisites + +- Python 3.7+ +- Ollama installed and running locally +- DeepSeek R1 14B model pulled in Ollama + +## Installation + +1. **Clone the repository** +```bash +git clone +cd property-rental-assistant +``` + +2. **Install required Python packages** +```bash +pip install requests beautifulsoup4 ollama ipython jupyter +``` + +3. **Install and setup Ollama** +```bash +# Install Ollama (macOS/Linux) +curl -fsSL https://ollama.ai/install.sh | sh + +# For Windows, download from: https://ollama.ai/download +``` + +4. **Pull the DeepSeek R1 model** +```bash +ollama pull deepseek-r1:14b +``` + +5. **Start Ollama server** +```bash +ollama serve +``` + +## Usage + +### Running the Notebook + +1. **Start Jupyter Notebook** +```bash +jupyter notebook +``` + +2. **Open the notebook** +Navigate to `property_rental_assistant.ipynb` in the Jupyter interface + +3. **Run all cells** +Click `Cell` → `Run All` or use `Shift + Enter` to run cells individually + +### Customizing Search Parameters + +Modify the `user_needs` variable in the notebook: +```python +user_needs = "I'm a student looking for a 2-bedroom house in Durham under £2,000/month" +``` + +Other examples: +- `"Family of 4 looking for 3-bedroom house with garden in Durham, budget £2,500/month"` +- `"Professional couple seeking modern 1-bed apartment near city center, max £1,500/month"` +- `"Student group needs 4-bedroom house near Durham University, £600/month per person"` + +### Changing the Property Website + +Update the `website_url` variable in the notebook: +```python +website_url = "https://www.onthemarket.com/to-rent/property/durham/" +``` + +## Architecture + +``` +┌─────────────────┐ ┌──────────────┐ ┌─────────────┐ +│ OnTheMarket │────▶│ Web Scraper │────▶│ Ollama │ +│ Website │ │ (BeautifulSoup)│ │ (DeepSeek R1)│ +└─────────────────┘ └──────────────┘ └─────────────┘ + │ + ▼ + ┌─────────────────────────────────┐ + │ AI-Generated Recommendations │ + │ • Top 5 matching properties │ + │ • Filtered by requirements │ + │ • Markdown formatted output │ + └─────────────────────────────────┘ +``` + +## Project Structure + +``` +property-rental-assistant/ +│ +├── property_rental_assistant.ipynb # Main Jupyter notebook +└── README.md # This file +``` + +## 🔧 Configuration + +### Ollama API Settings +```python +OLLAMA_API = "http://localhost:11434/api/chat" # Default Ollama endpoint +MODEL = "deepseek-r1:14b" # Model to use +``` + +### Web Scraping Settings +```python +headers = { + "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36" +} +timeout = 10 # Request timeout in seconds +``` + +### Content Limits +```python +website.text[:4000] # Truncate content to 4000 chars for token limits +``` + +## How It Works + +1. **Web Scraping**: The `Website` class fetches and parses HTML content from the property listing URL +2. **Content Cleaning**: Removes scripts, styles, and images to extract clean text +3. **Prompt Engineering**: Combines system prompt with user requirements and scraped data +4. **LLM Analysis**: Sends the prompt to DeepSeek R1 via Ollama API +5. **Recommendation Generation**: The AI analyzes listings and returns top matches in markdown format + +## 🛠️ Troubleshooting + +### Ollama Connection Error +``` +Error communicating with Ollama: [Errno 111] Connection refused +``` +**Solution**: Ensure Ollama is running with `ollama serve` + +### Model Not Found +``` +Error: model 'deepseek-r1:14b' not found +``` +**Solution**: Pull the model with `ollama pull deepseek-r1:14b` + +### Web Scraping Blocked +``` +Error fetching website: 403 Forbidden +``` +**Solution**: The website may be blocking automated requests. Try: +- Updating the User-Agent string +- Adding delays between requests +- Using a proxy or VPN + +### Insufficient Property Data +If recommendations are poor quality, the scraper may not be capturing listing details properly. Check: +- The website structure hasn't changed +- The content truncation limit (4000 chars) isn't too restrictive + +## Future Enhancements + +- [ ] Support multiple property websites (Rightmove, Zoopla, SpareRoom) +- [ ] Interactive CLI for dynamic user input +- [ ] Property image analysis +- [ ] Save search history and favorite properties +- [ ] Email notifications for new matching properties +- [ ] Price trend analysis +- [ ] Commute time calculations to specified locations +- [ ] Multi-language support +- [ ] Web interface with Flask/FastAPI +- [ ] Docker containerization + +## Acknowledgments + +- [Ollama](https://ollama.ai/) for local LLM hosting +- [DeepSeek](https://www.deepseek.com/) for the R1 model +- [BeautifulSoup](https://www.crummy.com/software/BeautifulSoup/) for web scraping +- [OnTheMarket](https://www.onthemarket.com/) for property data diff --git a/week1/community-contributions/D2-property-rental-assistant/day2.ipynb b/week1/community-contributions/D2-property-rental-assistant/day2.ipynb new file mode 100644 index 0000000..4c8dc5e --- /dev/null +++ b/week1/community-contributions/D2-property-rental-assistant/day2.ipynb @@ -0,0 +1,217 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "57112e5c-7b0f-4ba7-9022-ae21e8ac0f42", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b71a051-fc0e-46a9-8b1b-b58f685e800d", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"deepseek-r1:14b\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ed3be9dc-d459-46ac-a8eb-f9b932c4302f", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " def __init__(self, url):\n", + " self.url = url\n", + " try:\n", + " response = requests.get(url, headers=headers, timeout=10)\n", + " response.raise_for_status()\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"No body content found\"\n", + " except requests.RequestException as e:\n", + " print(f\"Error fetching website: {e}\")\n", + " self.title = \"Error loading page\"\n", + " self.text = \"Could not load page content\"" + ] + }, + { + "cell_type": "markdown", + "id": "17ea76f8-38d9-40b9-8aba-eb957d690a0d", + "metadata": {}, + "source": [ + "## Without Ollama package" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3a6fd698-8e59-4cd7-bb53-b9375e50f899", + "metadata": {}, + "outputs": [], + "source": [ + "def house_renting(system_prompt, user_prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + " payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }\n", + " response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + " return response.json()['message']['content']" + ] + }, + { + "cell_type": "markdown", + "id": "c826a52c-d1d3-493a-8b7c-6e75b848b453", + "metadata": {}, + "source": [ + "## Introducing Ollama package " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "519e27da-eeff-4c1b-a8c6-e680fdf01da2", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "\n", + "def house_renting_ollama(system_prompt, user_prompt):\n", + " try:\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + " response = ollama.chat(model=MODEL, messages=messages)\n", + " return response['message']['content']\n", + " except Exception as e:\n", + " return f\"Error communicating with Ollama: {e}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60e98b28-06d9-4303-b8ca-f7b798244eb4", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"\"\"\n", + "You are a helpful real estate assistant specializing in UK property rentals. Your job is to guide users in finding houses to rent, especially in Durham. Follow these rules:\n", + "1. Always ask clarifying questions if user input is vague. Determine location, budget, number of bedrooms, and tenant type (e.g. student, family, professional).\n", + "2. Use structured data provided from the website (like property listings) to identify relevant options.\n", + "3. If listings are provided, filter and rank them based on the user's preferences.\n", + "4. Recommend up to 5 top properties with rent price, bedroom count, key features, and location.\n", + "5. Always respond in markdown with clean formatting using headers, bold text, and bullet points.\n", + "6. If no listings match well, provide tips (e.g. \"try adjusting your budget or search radius\").\n", + "7. Stay concise, helpful, and adapt to whether the user is a student, family, couple, or solo tenant.\n", + "\"\"\"\n", + "\n", + "def user_prompt_for_renting(website, user_needs):\n", + " return f\"\"\"\n", + "I want to rent a house and here's what I'm looking for:\n", + "{user_needs}\n", + "\n", + "Here are the property listings I found on the website titled: \"{website.title}\".\n", + "\n", + "Please analyze them and recommend the best 3–5 options that match my needs. If none are suitable, tell me why and offer suggestions.\n", + "\n", + "The page content is below:\n", + "{website.text[:4000]}\n", + "\"\"\" # content is truncated for token limits" + ] + }, + { + "cell_type": "markdown", + "id": "ef420f4b-e3d2-4fbd-bf6f-811f2c8536e0", + "metadata": {}, + "source": [ + "## Ollama Package" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1cf128af-4ece-41ab-b353-5c8564c7de1d", + "metadata": {}, + "outputs": [], + "source": [ + "if __name__ == \"__main__\": \n", + " print(\"Starting AI Property Rental Assistant...\")\n", + " print(\"=\" * 50)\n", + " \n", + " website_url = \"https://www.onthemarket.com/to-rent/property/durham/\"\n", + " print(f\"🔍 Scraping properties from: {website_url}\")\n", + " \n", + " website = Website(website_url)\n", + " print(f\"Website Title: {website.title}\")\n", + " print(f\"Content Length: {len(website.text)} characters\")\n", + " print(f\"Successfully scraped property listings\\n\")\n", + " \n", + " user_needs = \"I'm a student looking for a 2-bedroom house in Durham under £2,000/month\"\n", + " print(f\"User Requirements: {user_needs}\\n\")\n", + " \n", + " user_prompt = user_prompt_for_renting(website, user_needs)\n", + " print(\"Generating AI recommendations...\")\n", + " \n", + " # Choose which method to use (comment out the one you don't want)\n", + " \n", + " # Method 1: Using ollama Python library\n", + " output = house_renting_ollama(system_prompt, user_prompt)\n", + " \n", + " # Method 2: Using direct API call\n", + " # output = house_renting(system_prompt, user_prompt)\n", + " \n", + " display(Markdown(output))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python [conda env:llms]", + "language": "python", + "name": "conda-env-llms-py" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Dashboard summarization.ipynb b/week1/community-contributions/Dashboard summarization.ipynb new file mode 100644 index 0000000..99c18f9 --- /dev/null +++ b/week1/community-contributions/Dashboard summarization.ipynb @@ -0,0 +1,256 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 3, + "id": "e0ab4a60-bc68-446d-ae13-6bd90d54ae44", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "749afaa0-a82e-4783-91fc-f69756075606", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "7e760d9c-d899-49e5-8b8f-c202794486cc", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found and looks good so far!\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "8efb8bb3-9be9-404b-aff5-306db64a75e7", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "cf677c78-012c-4b86-a76c-be47ed3cb987", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the \\\n", + "the dashboard in a website and provides a short executive summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "970a493a-880a-4206-9609-eee0651aa91f", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nPlease provide a detailed summary of the report for the year in markdown for its user (CFO); \\\n", + "The summary should be in a suitable form which could be sent through a mail for the exective.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "2520cdd1-4755-4c87-854f-430e81dbc3fc", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "7452990b-352b-43cc-adc6-4307d6d5c1d5", + "metadata": {}, + "outputs": [], + "source": [ + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "7372da3c-f3c7-455b-825e-f54d3b0cee68", + "metadata": {}, + "outputs": [], + "source": [ + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "be7d57dc-bec1-4771-9d15-d80fd4d3fbb5", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "# Executive Summary: Revenue & Profitability Dashboard\n", + "\n", + "**To:** [CFO Name] \n", + "**From:** [Your Name] \n", + "**Date:** [Current Date] \n", + "**Subject:** Yearly Analysis of Revenue & Profitability \n", + "\n", + "---\n", + "\n", + "Dear [CFO Name],\n", + "\n", + "I am pleased to present the Year-over-Year analysis derived from the Revenue & Profitability Dashboard. This dashboard has been designed to provide concise insights into our core financial performance metrics, enabling data-driven decision-making at the executive level.\n", + "\n", + "### Key Metrics Overview:\n", + "- **Revenue**: Comprehensive insights into total revenue across various regions and product categories, indicating sustainable growth patterns.\n", + "- **Profit**: Detailed profitability analysis segmented by customer groups, revealing key opportunities for margin improvement and cost optimization.\n", + "- **Unit Sales**: Analysis of unit sales trends that highlight product performance and demand fluctuations.\n", + "\n", + "### Insights by Segment:\n", + "- **Regional Performance**: Comparative analysis of revenue and profitability by region helps identify areas of growth and those requiring strategic intervention.\n", + "- **Product Performance**: A focused review of individual product lines shows which offerings are driving profitability and where we might consider realignment or innovation.\n", + "\n", + "### Dashboard Features:\n", + "- A **clean and focused layout** reduces cognitive load, allowing for quick assimilation and understanding of critical data points.\n", + "- **Contextual metrics** that align with our overarching business strategy, ensuring that our analysis supports our organizational goals.\n", + "- **Clear comparison points** are established to aid executives in making informed and timely decisions.\n", + "- Insightful details are presented at both product and regional levels, facilitating targeted strategies for improvement.\n", + "\n", + "### Conclusion:\n", + "The integration of design and context in our dashboard framework turns our data into strategic tools, empowering us to make faster and more informed decisions that drive real business impact.\n", + "\n", + "Please feel free to reach out for a more detailed discussion or specific metrics that may interest you.\n", + "\n", + "Best regards,\n", + "\n", + "[Your Name] \n", + "[Your Position]\n", + "\n", + "--- \n", + "\n", + "*Note: For additional inquiries or insights, feel free to follow our updates on LinkedIn or contact me directly.*" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "display_summary(\"https://community.fabric.microsoft.com/t5/Data-Stories-Gallery/Revenue-amp-Profitability-Dashboard/td-p/4780272\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1aa33cfd-d497-4ab8-abb5-eb4e6030890b", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Day-1_email_summarizers.ipynb b/week1/community-contributions/Day-1_email_summarizers.ipynb new file mode 100644 index 0000000..d2a4597 --- /dev/null +++ b/week1/community-contributions/Day-1_email_summarizers.ipynb @@ -0,0 +1,103 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "d7a6bb51", + "metadata": {}, + "outputs": [], + "source": [ + "# import library\n", + "from openai import OpenAI\n", + "import os\n", + "from dotenv import load_dotenv\n", + "\n", + "# Load your API key from an .env file\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7ac4cdf9", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "system_prompt = \"you are a helpful assistant that suggests an appropriate short subject line for an email based on its contents.\"\n", + "\n", + "user_prompt = \"\"\"\n", + "Hi John,\n", + "I hope this email finds you well. I wanted to follow up on our meeting last week regarding the quarterly budget proposal.\n", + "After reviewing the numbers with my team, we've identified some areas where we can reduce costs by approximately 15% without impacting our core operations. This would involve consolidating some vendor contracts and optimizing our software licensing.\n", + "Could we schedule a meeting next week to discuss these findings in detail? I'm available Tuesday through Thursday afternoon.\n", + "Looking forward to hearing from you.\n", + "\n", + "Best regards,\n", + "Sarah\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a77ca09e", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Step 2: Make the messages list\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8404f0fe", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Step 3: Call OpenAI\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7a4875f7", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Step 4: Print the result\n", + "print(response.choices[0].message.content)" + ] + } + ], + "metadata": { + "language_info": { + "name": "python" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Day-2_exercise_with_ollama3.ipynb b/week1/community-contributions/Day-2_exercise_with_ollama3.ipynb new file mode 100644 index 0000000..1168770 --- /dev/null +++ b/week1/community-contributions/Day-2_exercise_with_ollama3.ipynb @@ -0,0 +1,290 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "135717e7", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "29a9e634", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# OPTION 1\n", + "# using openai\n", + "\n", + "# message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "# client = OpenAI(base_url=\"http://localhost:11434/v1\", api_key=\"not-needed\")\n", + "# response = openai.chat.completions.create(model=``, messages=[{\"role\":\"user\", \"content\":message}])\n", + "# print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "306993ed", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# OPTION 2\n", + "# using Ollama\n", + "\n", + "message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "model=\"llama3\"\n", + "response=ollama.chat(model=model,messages=[{\"role\":\"user\",\"content\":message}])\n", + "print(response[\"message\"][\"content\"])\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "856f767b", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "4ce558dc", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Let's try one out. Change the website and add print statements to follow along.\n", + "\n", + "ed = Website(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "5e3956f8", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "99d791b4", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "5d89b748", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# See how this function creates exactly the format above\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "9a97d3e2", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response=ollama.chat(model=model,messages=messages_for(website))\n", + " return(response[\"message\"][\"content\"])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ec13fe0a", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "summarize(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "e3ade092", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "be2d49e6", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1ccbf33b", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ae3d0eae", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [ + "display_summary(\"https://anthropic.com\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Invoke LLM model from AWS Bedrock.ipynb b/week1/community-contributions/Invoke LLM model from AWS Bedrock.ipynb new file mode 100644 index 0000000..6948253 --- /dev/null +++ b/week1/community-contributions/Invoke LLM model from AWS Bedrock.ipynb @@ -0,0 +1,167 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 4, + "id": "9138adfe-71b0-4db2-a08f-dd9e472fdd63", + "metadata": {}, + "outputs": [], + "source": [ + "import pandas as pd\n", + "import numpy as np\n", + "import boto3" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15d71dd6-cc03-485e-8a34-7a33ed5dee0e", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "1358921d-173b-4d5d-828c-b6c3726a5eb3", + "metadata": {}, + "source": [ + "#### Connect to bedrock models" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "b3827087-182f-48be-8b59-b2741f8ded44", + "metadata": {}, + "outputs": [], + "source": [ + "import json" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "94c11534-6847-4e4a-b8e4-8066e0cc6aca", + "metadata": {}, + "outputs": [], + "source": [ + "# Use the Conversation API to send a text message to Amazon Nova.\n", + "\n", + "import boto3\n", + "from botocore.exceptions import ClientError\n", + "\n", + "# Create a Bedrock Runtime client in the AWS Region you want to use.\n", + "client = boto3.client(\"bedrock-runtime\", region_name=\"us-east-1\")\n", + "\n", + "# Set the model ID, e.g., Amazon Nova Lite.\n", + "model_id = \"amazon.nova-lite-v1:0\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9a8ad65f-abaa-475c-892c-2e2b4e668f5d", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "ac20bb00-e93f-4a95-a1de-dd2688bce591", + "metadata": {}, + "outputs": [], + "source": [ + "# Start a conversation with the user message.\n", + "user_message = \"\"\"\n", + "List the best parks to see in London with number of google ratings and value ie. 4.5 out of 5 etc. \n", + "Give number of ratings and give output in table form\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "a29f0055-48c4-4f25-b33f-cde1eaf755c5", + "metadata": {}, + "outputs": [], + "source": [ + "conversation = [\n", + " {\n", + " \"role\": \"user\",\n", + " \"content\": [{\"text\": user_message}],\n", + " }\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0e68b2d5-4d43-4b80-8574-d3c847b33661", + "metadata": {}, + "outputs": [], + "source": [ + "try:\n", + " # Send the message to the model, using a basic inference configuration.\n", + " response = client.converse(\n", + " modelId=model_id,\n", + " messages=conversation,\n", + " inferenceConfig={\"maxTokens\": 512, \"temperature\": 0.5, \"topP\": 0.9},\n", + " )\n", + "\n", + " # Extract and print the response text.\n", + " response_text = response[\"output\"][\"message\"][\"content\"][0][\"text\"]\n", + " print(response_text)\n", + "\n", + "except (ClientError, Exception) as e:\n", + " print(f\"ERROR: Can't invoke '{model_id}'. Reason: {e}\")\n", + " exit(1)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ed16ee7-3f09-4780-8dfc-d1c5f3cffdbe", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7f8c7a18-0907-430d-bfe7-86ecb8933bfd", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2183994b-cde5-45b0-b18b-37be3277d73b", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Movie_Suggestion.ipynb b/week1/community-contributions/Movie_Suggestion.ipynb new file mode 100644 index 0000000..930d5b0 --- /dev/null +++ b/week1/community-contributions/Movie_Suggestion.ipynb @@ -0,0 +1,104 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "7bb9010e-48a8-491e-a2a9-1a8dacc26f87", + "metadata": {}, + "source": [ + "# Movie Suggestion using Ollama Running Locally\n", + "\n", + "#### Takes the user input like languages and Genre and suggests Top 10 Movies of the selected attributes.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ad049302-dce8-4a0a-88ab-e485ac15fbe4", + "metadata": {}, + "outputs": [], + "source": [ + "import requests\n", + "from IPython.display import display, Markdown\n", + "\n", + "def get_movie_recommendations(language, genre, top_n=10, model='llama3.2'):\n", + " api_url = \"http://localhost:11434/api/generate\"\n", + " prompt = (\n", + " f\"Recommend {top_n} well-rated {language} movies from the {genre} genre. \"\n", + " \"For each movie, provide the name and a 1-2 sentence preview of its story. \"\n", + " \"Return the results as a Markdown table with columns: Title, Short Summary.\"\n", + " )\n", + " data = {\n", + " \"model\": model,\n", + " \"prompt\": prompt,\n", + " \"options\": {\"num_predict\": 800},\n", + " \"stream\": False\n", + " }\n", + " response = requests.post(api_url, json=data)\n", + " # Extract text response (could be markdown table already)\n", + " return response.json().get(\"response\", \"\").strip()" + ] + }, + { + "cell_type": "markdown", + "id": "01400553-419c-4798-8f19-e32e49379761", + "metadata": {}, + "source": [ + "#### Enter your Language and Genre" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a7527230-1e10-4b67-94c0-a84519b256c2", + "metadata": {}, + "outputs": [], + "source": [ + "language = input(\"Enter preferred language (e.g., French, Japanese): \").strip()\n", + "genre = input(\"Enter preferred genre (e.g., Drama, Comedy, Thriller): \").strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7ff0146f-b37e-4218-9678-15a40bed3659", + "metadata": {}, + "outputs": [], + "source": [ + "recommendations_md = get_movie_recommendations(language, genre)\n", + "# This prints out the Markdown table as formatted by the Llama 3.2 model\n", + "from IPython.display import display, Markdown\n", + "\n", + "display(Markdown(recommendations_md))" + ] + }, + { + "cell_type": "markdown", + "id": "58cc0fa4-a2a6-4597-8ae9-39970fb2a7b5", + "metadata": {}, + "source": [ + "### The Result will be displayed in a markdown fashion in a neat table with rows and columns." + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Playwright_Scrapping_Project/scraping_script.py b/week1/community-contributions/Playwright_Scrapping_Project/scraping_script.py new file mode 100644 index 0000000..7f9d619 --- /dev/null +++ b/week1/community-contributions/Playwright_Scrapping_Project/scraping_script.py @@ -0,0 +1,56 @@ +import os +import openai +from IPython.display import Markdown, display +from dotenv import load_dotenv +from playwright.sync_api import sync_playwright +from bs4 import BeautifulSoup + +load_dotenv() +openai.api_key = os.getenv("OPENAI_API_KEY") # Or set it directly + +def scrape_website(url): + # Code to scrape a website using Playwright + with sync_playwright() as p: + browser = p.chromium.launch(headless=True) + page = browser.new_page() + page.goto(url) + content = page.content() + browser.close() + return content + +def summarize_content(html_content): + #Get only the text parts of the webpage + soup = BeautifulSoup(html_content, 'html.parser') + summary_text = soup.get_text(separator=' ', strip=True) + # Code to summarize using OpenAI API + system_prompt = ("You summarize html content as markdown.") + user_prompt = ( + "You are a helpful assistant. Summarize the following HTML webpage content in markdown with simple terms:\n\n" + + summary_text + ) + response = openai.chat.completions.create( + model="gpt-4o", + messages=[{"role": "user", "content": user_prompt}] + ) + return response.choices[0].message.content + +def save_markdown(summary, filename="summary.md", url=None): + #Open the file summary.md + with open(filename, "w", encoding="utf-8") as f: + if url: + f.write(f"# Summary of [{url}]({url})\n\n") + else: + f.write("# Summary\n\n") + f.write(summary.strip()) + +# 4. Main Logic +def main(): + url = input("Enter the URL to summarize: ").strip() + html = scrape_website(url) + summary = summarize_content(html) + save_markdown(summary, filename="summary.md", url=url) + print("✅ Summary saved to summary.md") + +# 5. Entry Point +if __name__ == "__main__": + main() \ No newline at end of file diff --git a/week1/community-contributions/Playwright_Scrapping_Project/summary.md b/week1/community-contributions/Playwright_Scrapping_Project/summary.md new file mode 100644 index 0000000..6aa8639 --- /dev/null +++ b/week1/community-contributions/Playwright_Scrapping_Project/summary.md @@ -0,0 +1,34 @@ +# Summary of [https://www.willwight.com/](https://www.willwight.com/) + +# Will Wight - New York Times Best-Selling Author + +### Overview +Will Wight is a renowned author known for the "Cradle" series, alongside other works like "The Last Horizon" and "The Traveler's Gate Trilogy." He combines humor and storytelling in his blog and engages actively with his readers. + +### Books +- **The Last Horizon**: Currently ongoing series. +- **Cradle**: A 12-book series, now complete. +- **The Traveler's Gate Trilogy**: Completed series. +- **The Elder Empire**: Consists of two trilogies with stories happening simultaneously, totaling 6 books. + +### Recent Highlights +- **The Pilot Release**: The fourth book in "The Last Horizon" series, celebrated on July 4th, 2025. The 26th book by Will, marking a milestone as his next book will be his 27th. +- **Barnes & Noble Success**: A significant achievement of getting Will's books stocked nationwide in Barnes & Noble, marking a breakthrough for indie publishing. + +### Blog Highlights +- Will shares personal anecdotes and behind-the-scenes insights into his creative process. +- A humorous tone is used, including whimsical stories about his life and writing challenges. +- Recent experiences at Epic Universe theme park with thoughts on its design and offerings. + +### Connect +- **Mailing List**: Over 15,000 fans subscribe to receive updates on new stories and releases. +- **Hidden Gnome Publishing**: The entity behind Will's publications, working to bring his books to wider audiences. + +### Extras +- **Merch**: Available for fans wanting to support and connect with Will's universe. +- **Podcast**: Offers sneak peeks, discussions, and insights into Will's works. + +### Humorous Note +Will humorously describes himself transforming into a "monstrous mongoose" during a full moon, adding a quirky touch to his persona. + +For more detailed information on books, blogs, and extras, visit Will's website and explore his engaging world of storytelling! \ No newline at end of file diff --git a/week1/community-contributions/Top Tech products.ipynb b/week1/community-contributions/Top Tech products.ipynb new file mode 100644 index 0000000..53b4841 --- /dev/null +++ b/week1/community-contributions/Top Tech products.ipynb @@ -0,0 +1,181 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "bbd8585e-0a28-4fd9-80b5-690569f93e16", + "metadata": {}, + "outputs": [], + "source": [ + "#This notebook will help you to get top tech products with by providing category and subcategory" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "df039118-f462-4a8b-949e-53d3a726e292", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "aa" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e2ffd2e5-d061-446c-891e-15a6d1958ab6", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "92e26007-521f-4ea2-9df9-edd77dd7e183", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "27d21593-8feb-42e4-bbc0-2e949b51137d", + "metadata": {}, + "outputs": [], + "source": [ + "def tech_product(category_subcategory_budget):\n", + " parts = category_subcategory_budget.split('_')\n", + " return f\"{parts[0]}-{parts[1]}-{parts[2]}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dd978d25-5b84-4122-af7c-116f2bf72179", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(products):\n", + " return [\n", + " {\"role\": \"system\", \"content\": \"you are a tech product expert and you need to suggest the best suited product available in India basis the input received in the form of category-subcategory-budget (in inr),\\\n", + " revert with category and subcategory and show the product links as well along with pros and cons, respond in markdown\"},\n", + " {\"role\": \"user\", \"content\": tech_product(products)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b916db7a-81a4-41d9-87c2-a2346fd874d2", + "metadata": {}, + "outputs": [], + "source": [ + "messages_for(\"phone_gaming_40000\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b4bb3f1-95de-4eb5-afe1-068744f93301", + "metadata": {}, + "outputs": [], + "source": [ + "def get_top_products(category_subcategory):\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages= messages_for(category_subcategory)\n", + " )\n", + " return response.choices[0].message.content \n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c9272942-acfe-4fca-bd0a-3435c1ee6691", + "metadata": {}, + "outputs": [], + "source": [ + "get_top_products('phone_gaming_30000')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c2b3b9a-aceb-4f00-8c8d-8f6837ab94fc", + "metadata": {}, + "outputs": [], + "source": [ + "def display_markdown(category_subcategory_budget):\n", + " output = get_top_products(category_subcategory_budget)\n", + " display(Markdown(output))\n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6c135dd7-4ed4-48ee-ba3f-9b4ca1c32149", + "metadata": {}, + "outputs": [], + "source": [ + "display_markdown('Console_gaming_100000')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0ba06c55-7ef9-47eb-aeaf-3c4a7b29bccc", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/Week1_Day1_Flight_Prices_Tracker.ipynb b/week1/community-contributions/Week1_Day1_Flight_Prices_Tracker.ipynb new file mode 100644 index 0000000..6f07562 --- /dev/null +++ b/week1/community-contributions/Week1_Day1_Flight_Prices_Tracker.ipynb @@ -0,0 +1,223 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "fdc2f470", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "5f0fbd79", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found and looks good so far!\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "b771480a", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "0e97974c", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "2ec62fb3", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a list of the flights available according to what user asks for, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "7a93a605", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a list of all the flights available in a table format in markdown. The columns of the table should be - Flight carrier, Flight Dat and times, Fare, No. of stops. \\\n", + "Provide exact flight carriers. If it includes ads or offers, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "597646e5", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_flights(from_tx,to_tx, date_from,date_to=''):\n", + " website = Website(f\"https://www.ca.kayak.com/flights/{from_tx}-{to_tx}/{date_from}/{date_to}\")\n", + " user_prompt = user_prompt_for(website)\n", + " messages = [{\"role\":\"system\",\"content\":system_prompt},{\"role\":\"user\",\"content\":user_prompt}]\n", + " response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + " display(Markdown(response.choices[0].message.content))" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "47ae61f3", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "Here is the list of available flights from YYZ to DEL on 9/11:\n", + "\n", + "| Flight Carrier | Flight Date and Times | Fare | No. of Stops |\n", + "|----------------|-----------------------|--------|--------------|\n", + "| Air Canada | 9/11, 10:00 AM | C$ 833 | 1 |\n", + "| Lufthansa | 9/11, 5:00 PM | C$ 847 | 2 |\n", + "| Qatar Airways | 9/11, 1:30 PM | C$ 1,559| 1 |\n", + "\n", + "### Summary of Offers\n", + "- The cheapest fare is C$ 833 with a travel time of 23 hours and 35 minutes.\n", + "- The best fare option is C$ 847 with a travel time of 22 hours and 20 minutes.\n", + "- The quickest option is priced at C$ 1,559 with a travel duration of 13 hours and 55 minutes. \n", + "\n", + "*Note: Prices are per person and do not include baggage fees.*" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "fetch_flights('yyz','del','2025-11-09')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3a48ceb6", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15bb1a04", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "902975bf", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "venv", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.12.11" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/a-yotube-podcast-summerizer/yt_video_podcast_summerizer.ipynb b/week1/community-contributions/a-yotube-podcast-summerizer/yt_video_podcast_summerizer.ipynb new file mode 100644 index 0000000..414921f --- /dev/null +++ b/week1/community-contributions/a-yotube-podcast-summerizer/yt_video_podcast_summerizer.ipynb @@ -0,0 +1,233 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "provenance": [] + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "language_info": { + "name": "python" + } + }, + "cells": [ + { + "cell_type": "markdown", + "source": [ + "# **Youtube Informative-video Summerizer**\n", + "\n", + "This python app let's you summerize youtube videos that contains information-sharing-through-talking, like someone talking about a subject, someone sharing a life advice, a podcast etc.\n", + "\n", + "We extract the transcipt analyize it with an LLM to summerize and create summerization and analysis.\n", + "\n", + "\n", + "> We use youtube_transcript_api which allows you to get the transcript text of any youtube video.\n", + "\n", + "> Results however are not ideal for our use case since it does not provide who says what in case of more than one speaker. it only provide one giant string of all the words said in the video respectivly with some noise.\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n", + "\n" + ], + "metadata": { + "id": "4KULQ4rViju1" + } + }, + { + "cell_type": "code", + "source": [ + "#!pip install youtube-transcript-api" + ], + "metadata": { + "id": "C21ZN5MNZ_1b" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "from youtube_transcript_api import YouTubeTranscriptApi\n", + "from youtube_transcript_api.formatters import TextFormatter, SRTFormatter\n", + "import re\n", + "from openai import OpenAI\n", + "from google.colab import userdata # dotenv equevilant for google colab\n", + "from IPython.display import Markdown, display, update_display" + ], + "metadata": { + "id": "ttbBAJC7Zrn5" + }, + "execution_count": 35, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "ytt = YouTubeTranscriptApi()\n", + "formatter = TextFormatter() # --> Plain text\n", + "# formatter = SRTFormatter() # --> With timestamps\n", + "\n", + "openai_api_key = userdata.get('OPENAI_TOKEN')\n", + "openai_client = OpenAI(api_key=openai_api_key)\n", + "MODEL = \"gpt-4o-mini\"" + ], + "metadata": { + "id": "1oP0uPCylaig" + }, + "execution_count": 36, + "outputs": [] + }, + { + "cell_type": "code", + "execution_count": 37, + "metadata": { + "id": "ILPjwpGkZm1t" + }, + "outputs": [], + "source": [ + "def get_video_id(url):\n", + " \"\"\"Extracts video ID from a YouTube URL.\"\"\"\n", + " regex = r\"(?:v=|\\/)([0-9A-Za-z_-]{11}).*\"\n", + " match = re.search(regex, url)\n", + " if match:\n", + " return match.group(1)\n", + " raise ValueError(\"Invalid YouTube URL\")\n", + "\n", + "\n", + "def get_transcript(url):\n", + " video_id = get_video_id(url)\n", + " fetched_transcript = ytt.fetch(video_id)\n", + " # ^ defaults to English transcript, for other language use:\n", + " # fetched = ytt.fetch(video_id, languages=['de', 'en'])\n", + " transcript_text = formatter.format_transcript(fetched_transcript)\n", + " return transcript_text" + ] + }, + { + "cell_type": "code", + "source": [ + "system_prompt = \"\"\"You are an expert assistant specialized in analyzing podcast transcripts. You will be given the full transcript of a YouTube podcast episode.\n", + "\n", + "Your task is to extract and summarize the main views or arguments presented in the podcast. For each view or argument, also identify and list any supporting evidence such as:\n", + "\n", + "- Facts or statistics\n", + "- Academic studies or research\n", + "- Theories or philosophical frameworks\n", + "- Anecdotes or personal experiences\n", + "- Expert opinions or quotes\n", + "\n", + "Recognize off topic segments and adds and igrone them.\n", + "\n", + "Structure your output in a clear and concise format.\n", + "\n", + "Output Format:\n", + "\n", + "Podcast Summary:\n", + "\n", + "1. View/Argument:\n", + " - Description: [Summarize the view or claim in 1-2 sentences.]\n", + " - Supporting Evidence:\n", + " • [Fact, study, or reasoning #1]\n", + " • [Fact, study, or reasoning #2]\n", + " • [Optional counterarguments or nuances, if any]\n", + "\n", + "2. View/Argument:\n", + " - Description: [...]\n", + " - Supporting Evidence:\n", + " • [...]\n", + "\n", + "Guidelines:\n", + "- Only include major views or arguments that are discussed in depth.\n", + "- Paraphrase in clear, neutral, and objective language.\n", + "- Do not include filler, small talk, or off-topic segments.\n", + "- If a claim lacks explicit evidence, note it as “No clear supporting evidence provided.”\n", + "\n", + "Always respond and orginize your response using Markdow.\n", + "\"\"\"\n" + ], + "metadata": { + "id": "Ye3m_3lEejb_" + }, + "execution_count": 38, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "def get_user_prompt(title,url):\n", + " prompt = f\"Following is a transcript for a podcast titled '{title}' \\n\"\n", + " prompt += \"Carefully read through this content, analyse and summerize it as told, respond in Markdown.\"\n", + " prompt += \"\\nTranscript: \\n\\n\"\n", + " prompt += get_transcript(url)\n", + " return prompt" + ], + "metadata": { + "id": "1jk6YbkpupqI" + }, + "execution_count": 39, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# user_prompt = get_user_prompt()\n", + "def summerize_video(title,url):\n", + " user_prompt = get_user_prompt(title,url)\n", + " stream = openai_client.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt},\n", + " ],\n", + " stream = True,\n", + " )\n", + "\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ], + "metadata": { + "id": "wJy0Qb8u9uqR" + }, + "execution_count": 40, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "summerize_video(\"Anti-Aging Expert: Missing This Vitamin Is As Bad As Smoking! The Truth About Creatine!\",\"https://www.youtube.com/watch?v=JCTb3QSrGMQ\")" + ], + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/", + "height": 1000 + }, + "id": "tbvBiPrv_O3i", + "outputId": "69d24254-e384-4b07-e35f-96c7bb733298" + }, + "execution_count": 41, + "outputs": [ + { + "output_type": "display_data", + "data": { + "text/plain": [ + "" + ], + "text/markdown": "# Podcast Summary: \"Anti-Aging Expert: Missing This Vitamin Is As Bad As Smoking! The Truth About Creatine!\"\n\n1. **View/Argument: Vitamin D and Health Risks**\n - **Description:** Vitamin D deficiency significantly increases the risk of dementia and various health issues, yet many individuals are unaware of its critical importance.\n - **Supporting Evidence:**\n - Vitamin D deficiency can raise dementia risk by 80%.\n - Individuals with adequate vitamin D have a 40% reduced risk of dementia and experience better cognitive function.\n\n2. **View/Argument: Role of Lifestyle in Aging**\n - **Description:** Lifestyle choices account for over 70% of aging effects, with exercise and nutrition being key factors in improving longevity and health.\n - **Supporting Evidence:**\n - Studies show participants involved in regular exercise did not experience hippocampal shrinkage, but rather an increase in size.\n - Exercise is equated to a miracle drug for its extensive health benefits, as highlighted by unquantifiable positive effects when compared to medications.\n\n3. **View/Argument: Importance of Magnesium**\n - **Description:** Magnesium is crucial for cellular function, metabolism, and reducing cancer risk, yet nearly half the U.S. population is magnesium deficient.\n - **Supporting Evidence:**\n - Individuals with the highest magnesium levels have a 40% lower all-cause mortality.\n - A 24% increase in pancreatic cancer incidents is associated with every 100 mg decrease in magnesium intake.\n\n4. **View/Argument: Benefits of Creatine in Brain Health**\n - **Description:** Creatine isn't just beneficial for muscle health but also shows promise for enhancing cognitive performance, especially under stress or sleep deprivation.\n - **Supporting Evidence:**\n - A study found that creatine can negate cognitive deficits caused by 21 hours of sleep deprivation.\n - Users often report improved focus and energy levels when supplementing with creatine regularly.\n\n5. **View/Argument: Exercise and Hormonal Benefits**\n - **Description:** Regular exercise, especially high-intensity interval training, can reverse heart aging and improve mental health markers.\n - **Supporting Evidence:**\n - Participants in an intensive exercise program showed heart structures that were more akin to those of individuals two decades younger.\n - High-intensity workouts were shown to improve cognition and neuroplasticity due to the metabolic changes they induce.\n\n6. **View/Argument: Impact of Nutrition on Cognitive Function**\n - **Description:** A healthy diet rich in omega-3 fatty acids, vitamins D and other nutrients is essential for maintaining cognitive function and overall health.\n - **Supporting Evidence:**\n - Adequate omega-3 intake has been linked to a 5-year increase in life expectancy.\n - Regular consumption of nutrient-rich foods, such as blueberries and dark leafy greens, supports cognition and potentially reduces the risk of neurodegenerative diseases.\n\n7. **View/Argument: The Importance of Autophagy**\n - **Description:** Fasting promotes autophagy, a cellular cleaning process that can protect against diseases and improve health.\n - **Supporting Evidence:**\n - Studies suggest that fasting for 16 hours can activate autophagy and contribute to cellular repair processes.\n\n8. **View/Argument: Intermittent Fasting and Health Improvements**\n - **Description:** Intermittent fasting can improve metabolic parameters and cognitive performance while providing health benefits beyond simple calorie restriction.\n - **Supporting Evidence:**\n - Individuals practicing intermittent fasting showed improved glucose regulation compared to those restricting calories alone without fasting.\n\n9. **View/Argument: Microplastics and Health Risks**\n - **Description:** The pervasive presence of microplastics in everyday products poses health risks that are not widely recognized.\n - **Supporting Evidence:**\n - Common items, such as paper coffee cups and plastic water bottles, can release harmful chemicals, leading to increased levels of substances like BPA in beverages.\n\nBy summarizing these key points, the podcast emphasizes the interconnectedness of nutrition, exercise, and mental well-being in managing aging and chronic diseases. Additionally, it highlights emerging research on creatine, fasting, and environmental health risks that affect longevity and quality of life." + }, + "metadata": {} + } + ] + } + ] +} \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/.gitignore b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/.gitignore new file mode 100644 index 0000000..290698f --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/.gitignore @@ -0,0 +1,210 @@ +# Byte-compiled / optimized / DLL files +__pycache__/ +*.py[codz] +*$py.class + +# C extensions +*.so + +# Distribution / packaging +.Python +build/ +develop-eggs/ +dist/ +downloads/ +eggs/ +.eggs/ +lib/ +lib64/ +parts/ +sdist/ +var/ +wheels/ +share/python-wheels/ +*.egg-info/ +.installed.cfg +*.egg +MANIFEST + +# PyInstaller +# Usually these files are written by a python script from a template +# before PyInstaller builds the exe, so as to inject date/other infos into it. +*.manifest +*.spec + +# Installer logs +pip-log.txt +pip-delete-this-directory.txt + +# Unit test / coverage reports +htmlcov/ +.tox/ +.nox/ +.coverage +.coverage.* +.cache +nosetests.xml +coverage.xml +*.cover +*.py.cover +.hypothesis/ +.pytest_cache/ +cover/ + +# Translations +*.mo +*.pot + +# Django stuff: +*.log +local_settings.py +db.sqlite3 +db.sqlite3-journal + +# Flask stuff: +instance/ +.webassets-cache + +# Scrapy stuff: +.scrapy + +# Sphinx documentation +docs/_build/ + +# PyBuilder +.pybuilder/ +target/ + +# Jupyter Notebook +.ipynb_checkpoints + +# IPython +profile_default/ +ipython_config.py + +# pyenv +# For a library or package, you might want to ignore these files since the code is +# intended to run in multiple environments; otherwise, check them in: +# .python-version + +# pipenv +# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control. +# However, in case of collaboration, if having platform-specific dependencies or dependencies +# having no cross-platform support, pipenv may install dependencies that don't work, or not +# install all needed dependencies. +#Pipfile.lock + +# UV +# Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control. +# This is especially recommended for binary packages to ensure reproducibility, and is more +# commonly ignored for libraries. +#uv.lock + +# poetry +# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control. +# This is especially recommended for binary packages to ensure reproducibility, and is more +# commonly ignored for libraries. +# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control +#poetry.lock +#poetry.toml + +# pdm +# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control. +# pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python. +# https://pdm-project.org/en/latest/usage/project/#working-with-version-control +#pdm.lock +#pdm.toml +.pdm-python +.pdm-build/ + +# pixi +# Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control. +#pixi.lock +# Pixi creates a virtual environment in the .pixi directory, just like venv module creates one +# in the .venv directory. It is recommended not to include this directory in version control. +.pixi + +# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm +__pypackages__/ + +# Celery stuff +celerybeat-schedule +celerybeat.pid + +# SageMath parsed files +*.sage.py + +# Environments +.env +.envrc +.venv +env/ +venv/ +ENV/ +env.bak/ +venv.bak/ + +# Spyder project settings +.spyderproject +.spyproject + +# Rope project settings +.ropeproject + +# mkdocs documentation +/site + +# mypy +.mypy_cache/ +.dmypy.json +dmypy.json + +# Pyre type checker +.pyre/ + +# pytype static type analyzer +.pytype/ + +# Cython debug symbols +cython_debug/ + +# PyCharm +# JetBrains specific template is maintained in a separate JetBrains.gitignore that can +# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore +# and can be added to the global gitignore or merged into this file. For a more nuclear +# option (not recommended) you can uncomment the following to ignore the entire idea folder. +#.idea/ + +# Abstra +# Abstra is an AI-powered process automation framework. +# Ignore directories containing user credentials, local state, and settings. +# Learn more at https://abstra.io/docs +.abstra/ + +# Visual Studio Code +# Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore +# that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore +# and can be added to the global gitignore or merged into this file. However, if you prefer, +# you could uncomment the following to ignore the entire vscode folder +# .vscode/ + +# Ruff stuff: +.ruff_cache/ + +# PyPI configuration file +.pypirc + +# Cursor +# Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to +# exclude from AI features like autocomplete and code analysis. Recommended for sensitive data +# refer to https://docs.cursor.com/context/ignore-files +.cursorignore +.cursorindexingignore + +# Marimo +marimo/_static/ +marimo/_lsp/ +__marimo__/ + + +.*-env \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai-brochure-creator.py b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai-brochure-creator.py new file mode 100644 index 0000000..79f3246 --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai-brochure-creator.py @@ -0,0 +1,207 @@ +from ai_core import AICore +from ai_brochure_config import AIBrochureConfig +from extractor_of_relevant_links import ExtractorOfRelevantLinks +from website import Website +from openai.types.responses import Response +from rich.console import Console +from rich.markdown import Markdown +from requests import Session +from concurrent.futures import ThreadPoolExecutor, as_completed +from json import loads + +class BrochureCreator(AICore[str]): + """ + Builds a short Markdown brochure for a company or individual by: + - extracting relevant links from the website, + - inferring the entity name and status, + - and prompting the model using the collected page content. + """ + + @property + def _website(self) -> Website: + """Return the main Website instance to analyze.""" + return self.__website + + @property + def _extractor(self) -> ExtractorOfRelevantLinks: + """Return the helper responsible for extracting relevant links.""" + return self.__extractor + + def __init__(self, config: AIBrochureConfig, website: Website) -> None: + """ + Initialize the brochure creator with configuration and target website. + + Parameters: + config: AI and runtime configuration. + website: The root website to analyze and summarize. + """ + system_behavior: str = ("You are an assistant that analyzes the contents of several relevant pages from a company website " + "and creates a short brochure about the company for prospective customers, investors and recruits. " + "Include details of company culture, customers and careers/jobs if information is available. ") + super().__init__(config, system_behavior) + self.__website: Website = website + self.__extractor: ExtractorOfRelevantLinks = ExtractorOfRelevantLinks(config, website) + + def create_brochure(self) -> str: + """ + Create a short Markdown brochure based on the website's content. + + Returns: + A Markdown string with the brochure, or a fallback message if no relevant pages were found. + """ + relevant_pages: list[dict[str, str | Website]] = self._get_relevant_pages() + if not relevant_pages: + return "No relevant pages found to create a brochure." + + brochure_prompt_part: str = self._form_brochure_prompt(relevant_pages) + inferred_company_name, inferred_status = self._infer_entity(brochure_prompt_part) + + full_brochure_prompt: str = self._form_full_prompt(inferred_company_name, inferred_status) + response: str = self.ask(full_brochure_prompt) + return response + + def _get_relevant_pages(self) -> list[dict[str, str | Website]]: + """ + Resolve relevant links into Website objects using a shared session and concurrency. + """ + relevant_pages: list[dict[str, str | Website]] = [] + relevant_links: list[dict[str, str]] = self._extractor.extract_relevant_links()["links"] + # Limit the number of pages to fetch to keep latency and token usage reasonable. + MAX_PAGES: int = 6 + links_subset = relevant_links[:MAX_PAGES] + + def build_page(item: dict[str, str], session: Session) -> dict[str, str | Website] | None: + try: + url = str(item["url"]) + page_type = str(item["type"]) + return {"type": page_type, "page": Website(url, session=session)} + except Exception: + return None + + with Session() as session, ThreadPoolExecutor(max_workers=4) as executor: + futures = [executor.submit(build_page, link, session) for link in links_subset] + for fut in as_completed(futures): + res = fut.result() + if res: + relevant_pages.append(res) + + return relevant_pages + + def _truncate_text(self, text: str, limit: int) -> str: + """ + Truncate text to 'limit' characters to reduce tokens and latency. + """ + if len(text) <= limit: + return text + return text[: max(0, limit - 20)] + "... [truncated]" + + def _form_brochure_prompt(self, relevant_pages: list[dict[str, str | Website]]) -> str: + """ + Assemble a prompt that includes the main page and relevant pages' titles and text. + + Parameters: + relevant_pages: List of page descriptors returned by _get_relevant_pages. + + Returns: + A prompt string containing quoted sections per page. + """ + QUOTE_DELIMITER: str = "\n\"\"\"\n" + MAX_MAIN_CHARS = 6000 + MAX_PAGE_CHARS = 3000 + prompt: str = ( + f"Main page:{QUOTE_DELIMITER}" + f"Title: {self._website.title}\n" + f"Text:\n{self._truncate_text(self._website.text, MAX_MAIN_CHARS)}{QUOTE_DELIMITER}\n" + ) + + for page in relevant_pages: + if isinstance(page['page'], Website) and not page['page'].fetch_failed: + prompt += ( + f"{page['type']}:{QUOTE_DELIMITER}" + f"Title: {page['page'].title}\n" + f"Text:\n{self._truncate_text(page['page'].text, MAX_PAGE_CHARS)}{QUOTE_DELIMITER}\n" + ) + + return prompt + + def _infer_entity(self, brochure_prompt_part: str) -> tuple[str, str]: + """ + Infer both the entity name and status in a single model call to reduce latency. + Returns: + (name, status) where status is 'company' or 'individual'. + """ + prompt = ( + "From the following website excerpts, infer the entity name and whether it is a company or an individual. " + "Respond strictly as JSON with keys 'name' and 'status' (status must be 'company' or 'individual').\n" + f"{brochure_prompt_part}" + ) + raw = self.ask(prompt) + try: + data: dict[str, str] = loads(raw) + name: str = str(data.get("name", "")).strip() or "Unknown" + status: str = str(data.get("status", "")).strip().lower() + if status not in ("company", "individual"): + status = "company" + return name, status + except Exception: + # Fallback: use entire output as name, assume company + return raw.strip() or "Unknown", "company" + + def _form_full_prompt(self, inferred_company_name: str, inferred_status: str) -> str: + """ + Build the final brochure-generation prompt using the inferred entity and prior history. + + Parameters: + inferred_company_name: The inferred entity name. + inferred_status: Either 'company' or 'individual'. + + Returns: + A final prompt instructing the model to produce a Markdown brochure. + """ + full_prompt: str = (f"You are looking at a {inferred_status} called {inferred_company_name}, to whom website {self._website.website_url} belongs.\n" + f"Build a short brochure about the {inferred_status}. Use the information from the website that is already stored in the history.\n" + "Your response must be in a Markdown format.") + return full_prompt + + def ask(self, question: str) -> str: + """ + Send a question to the model, update chat history, and return the text output. + + Parameters: + question: The user prompt. + + Returns: + The model output text. + """ + self.history_manager.add_user_message(question) + response: Response = self._ai_api.responses.create( + model=self.config.model_name, + instructions=self.history_manager.system_behavior, + input=self.history_manager.chat_history, + reasoning={ "effort": "low" } + ) + self.history_manager.add_assistant_message(response) + return response.output_text + +console: Console = Console() + +def display_markdown(content: str) -> None: + """ + Render Markdown content to the console using rich. + """ + console.print(Markdown(content)) + +def show_summary(summary: str) -> None: + """ + Print a Markdown summary if provided; otherwise print a fallback message. + """ + if summary: + display_markdown(summary) + else: + console.print("No summary found.") + +if __name__ == "__main__": + website: Website = Website("") + brochure_creator: BrochureCreator = BrochureCreator(AIBrochureConfig(), website) + brochure: str = brochure_creator.create_brochure() + display_markdown(brochure) \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_brochure_config.py b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_brochure_config.py new file mode 100644 index 0000000..9a0e2bd --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_brochure_config.py @@ -0,0 +1,59 @@ +import os +from dotenv import load_dotenv + +class AIBrochureConfig: + """ + Configuration class to load environment variables. + """ + + def __get_config_value(self, key: str): + """ + Get the value of an environment variable. + """ + if not key: + raise ValueError("Key must be provided") + + value: str | None = os.getenv(key) + if not value: + raise ValueError(f"Environment variable '{key}' not found") + + return value + + def _get_str(self, key: str) -> str: + """ + Get a string value from the environment variables. + """ + return self.__get_config_value(key) + + def _get_int(self, key: str) -> int: + """ + Get an integer value from the environment variables. + """ + value = self.__get_config_value(key) + try: + return int(value) + except ValueError: + raise ValueError(f"Environment variable '{key}' must be an integer") + + @property + def openai_api_key(self) -> str: + """ + Get the OpenAI API key from the environment variables. + """ + if self.__openai_api_key == "": + self.__openai_api_key = self._get_str("OPENAI_API_KEY") + return self.__openai_api_key + + @property + def model_name(self) -> str: + """ + Get the model name from the environment variables. + """ + if self.__model_name == "": + self.__model_name = self._get_str("MODEL_NAME") + return self.__model_name + + def __init__(self) -> None: + load_dotenv(dotenv_path=".env") + self.__openai_api_key: str = "" + self.__model_name: str = "" diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_core.py b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_core.py new file mode 100644 index 0000000..e517f9d --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/ai_core.py @@ -0,0 +1,181 @@ +import openai +from abc import ABC, abstractmethod +from ai_brochure_config import AIBrochureConfig +from typing import Any, cast, Generic, TypeVar +from openai.types.responses import ResponseInputItemParam, Response, ResponseOutputMessage + +TAiResponse = TypeVar('TAiResponse', default=Any) + +class HistoryManager: + """ + Manage chat history and system behavior for a conversation with the model. + """ + @property + def chat_history(self) -> list[ResponseInputItemParam]: + """ + Return the accumulated conversation as a list of response input items. + """ + return self.__chat_history + + @property + def system_behavior(self) -> str: + """ + Return the system behavior (instructions) used for this conversation. + """ + return self.__system_behavior + + def __init__(self, system_behavior: str) -> None: + """ + Initialize the history manager. + + Parameters: + system_behavior: The system instruction string for the conversation. + """ + self.__chat_history: list[ResponseInputItemParam] = [] + self.__system_behavior: str = system_behavior + + def add_user_message(self, message: str) -> None: + """ + Append a user message to the chat history. + + Parameters: + message: The user text to add. + """ + self.__chat_history.append({ + "role": "user", + "content": [{"type": "input_text", "text": message}], + }) + + def add_assistant_message(self, output_message: Response) -> None: + """ + Append the assistant's output to the chat history. + + Parameters: + output_message: The model response to convert and store. + """ + for out in output_message.output: + # Convert the Pydantic output model to an input item shape + self.__chat_history.append( + cast(ResponseInputItemParam, out.model_dump(exclude_unset=True)) + ) + + +class AICore(ABC, Generic[TAiResponse]): + """ + Abstract base class for AI core functionalities. + """ + @property + def config(self) -> AIBrochureConfig: + """ + Return the stored AIBrochureConfig for this instance. + + Returns: + AIBrochureConfig: The current configuration used by this object. + + Notes: + - This accessor returns the internal configuration reference. Mutating the returned + object may affect the internal state of this instance. + - To change the configuration, use the appropriate setter or factory method rather + than modifying the returned value in-place. + """ + return self.__config + + @config.setter + def config(self, config: AIBrochureConfig | None) -> None: + """ + Set the instance configuration for the AI brochure generator. + + Parameters + ---------- + config : AIBrochureConfig | None + The configuration to assign to the instance. If None, the instance's + configuration will be reset to a newly created default AIBrochureConfig. + + Returns + ------- + None + + Notes + ----- + This method stores the provided configuration on a private attribute + """ + if config is None: + self.__config = AIBrochureConfig() + else: + self.__config = config + + @property + def _ai_api(self) -> openai.OpenAI: + """ + Return the cached OpenAI API client, initializing it on first access. + + This private helper lazily constructs and caches an openai.OpenAI client using + the API key found on self.config.openai_api_key. On the first call, if the + client has not yet been created, the method verifies that self.config is set, + creates the client with openai.OpenAI(api_key=...), stores it on + self.__ai_api, and returns it. Subsequent calls return the same cached + instance. + + Returns: + openai.OpenAI: A configured OpenAI API client. + + Raises: + ValueError: If self.config is None when attempting to initialize the client. + + Notes: + - The method mutates self.__ai_api as a side effect (caching). + - The caller should treat this as a private implementation detail. + - Thread safety is not guaranteed; concurrent initialization may result in + multiple client instances if invoked from multiple threads simultaneously. + """ + if self.__ai_api is None: + if self.config is None: + raise ValueError("Configuration must be set before accessing AI API") + self.__ai_api = openai.OpenAI(api_key=self.config.openai_api_key) + return self.__ai_api + + @property + def history_manager(self) -> HistoryManager: + """ + Return the history manager for this AI core instance. + + This property provides access to the HistoryManager that tracks the chat + history and system behavior. + + Returns: + HistoryManager: The current history manager. This property always returns + a HistoryManager instance and never None. + """ + return self.__history_manager + + def __init__(self, config: AIBrochureConfig, system_behavior: str) -> None: + """ + Initializes the AI core with the provided configuration. + + Parameters: + config (AIBrochureConfig): The configuration object for the AI core. + system_behavior (str): The behavior of the system. + """ + # Initialize all instance-level attributes here + self.__config: AIBrochureConfig = config + self.__history_manager: HistoryManager = HistoryManager(system_behavior) + self.__ai_api: openai.OpenAI | None = None + + if __debug__: + # Sanity check: confirm attributes are initialized + assert hasattr(self, "_AICore__config") + assert hasattr(self, "_AICore__history_manager") + assert hasattr(self, "_AICore__ai_api") + + @abstractmethod + def ask(self, question: str) -> TAiResponse: + """ + Ask a question to the AI model. + + Parameters: + question: The question to ask. + + Returns: + TAiResponse: The model's response type defined by the subclass. + """ + pass \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/extractor_of_relevant_links.py b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/extractor_of_relevant_links.py new file mode 100644 index 0000000..e94fa38 --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/extractor_of_relevant_links.py @@ -0,0 +1,91 @@ +from ai_brochure_config import AIBrochureConfig +from website import Website +from ai_core import AICore +from openai.types.responses import Response +from json import loads + +RelevantLinksDict = dict[str, list[dict[str, str]]] + +class ExtractorOfRelevantLinks(AICore[RelevantLinksDict]): + """ + Extractor for relevant links from a website. + """ + + @property + def website(self) -> Website: + """Return the root Website whose links are being analyzed.""" + return self.__website + + def __init__(self, config: AIBrochureConfig, website: Website) -> None: + """ + Initialize the extractor with configuration and target website. + + Parameters: + config: AI and runtime configuration. + website: The Website from which links were collected. + """ + system_behavior: str = ("You are an expert in creation of online advertisement materials." + "You are going to be provided with a list of links found on a website." + "You are able to decide which of the links would be most relevant to include in a brochure about the company," + "such as links to an About page or a Company page or Careers/Jobs pages.\n" + "You should respond in JSON as in this example:") + system_behavior += """ + { + "links": [ + {"type": "about page", "url": "https://www.example.com/about"}, + {"type": "company page", "url": "https://www.another_example.net/company"}, + {"type": "careers page", "url": "https://ex.one_more_example.org/careers"} + ] + } + """ + super().__init__(config, system_behavior) + self.__website: Website = website + + def get_links_user_prompt(self) -> str: + """ + Build a user prompt listing discovered links and instructions for relevance filtering. + + Returns: + A string to send to the model listing links and guidance. + """ + starter_part: str = (f"Here is a list of links found on the website of {self.website.website_url} - " + "please decide which of these links are relevant web links for a brochure about company." + "Respond with full HTTPS URLs. Avoid including Terms of Service, Privacy, email links.\n" + "Links (some might be relative links):\n") + + links_part: str = "\n".join(f"- {link}" for link in self.website.links_on_page) if self.website.links_on_page else "No links found." + + return starter_part + links_part + + def extract_relevant_links(self) -> RelevantLinksDict: + """ + Request the model to select relevant links for brochure creation. + + Returns: + A dictionary with a 'links' array containing objects with 'type' and 'url'. + """ + user_prompt = self.get_links_user_prompt() + response = self.ask(user_prompt) + return response + + def ask(self, question: str) -> RelevantLinksDict: + """ + Send a question to the model and parse the JSON response. + + Parameters: + question: The prompt to submit. + + Returns: + RelevantLinksDict: Parsed JSON containing selected links. + """ + self.history_manager.add_user_message(question) + + response: Response = self._ai_api.responses.create( + model=self.config.model_name, + instructions=self.history_manager.system_behavior, + reasoning={ "effort": "low" }, + input=self.history_manager.chat_history + ) + + self.history_manager.add_assistant_message(response) + return loads(response.output_text) \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/requirements.txt b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/requirements.txt new file mode 100644 index 0000000..9747210 --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/requirements.txt @@ -0,0 +1,5 @@ +python-dotenv +openai +bs4 +requests +rich \ No newline at end of file diff --git a/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/website.py b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/website.py new file mode 100644 index 0000000..ac9bb9d --- /dev/null +++ b/week1/community-contributions/ai-powered-marketing-brochures-gpt-5/website.py @@ -0,0 +1,286 @@ +from ipaddress import ip_address, IPv4Address, IPv6Address +from urllib.parse import ParseResult, urlparse +from bs4 import BeautifulSoup, Tag +from requests import get, RequestException, Session + +class Extractor: + """ + Extracts and processes content from HTML response text using BeautifulSoup. + """ + __soup: BeautifulSoup + + __extracted_title: str = "" + @property + def extracted_title(self) -> str: + """ + Returns the extracted title from the HTML content. + """ + if not self.__extracted_title: + self.__extracted_title = self.get_title() + return self.__extracted_title + + __extracted_text: str = "" + @property + def extracted_text(self) -> str: + """ + Returns the extracted main text content from the HTML, excluding irrelevant tags. + """ + if not self.__extracted_text: + self.__extracted_text = self.get_text() + return self.__extracted_text + + __extracted_links_on_page: list[str] | None = None + @property + def extracted_links_on_page(self) -> list[str]: + """ + Return all href values found on the page. + + Notes: + - Only anchor tags with an href are included. + - Values are returned as-is (may be relative or absolute). + """ + if self.__extracted_links_on_page is None: + self.__extracted_links_on_page = [str(a.get("href")) for a in self._soup.find_all('a', href=True) if isinstance(a, Tag)] + return self.__extracted_links_on_page + + @property + def _soup(self) -> BeautifulSoup: + """ + Returns the BeautifulSoup object for the HTML content. + """ + return self.__soup + + def __init__(self, response_text_content: str) -> None: + """ + Initializes the Extractor with HTML response text. + + Parameters: + response_text_content (str): The HTML response text to be processed. + """ + self.__soup = BeautifulSoup(response_text_content, "html.parser") + self.__extracted_links_on_page = None + + def get_title(self) -> str: + """ + Extracts the title from the HTML content. + """ + return self._soup.title.get_text() if self._soup.title is not None else "No title" + + def get_text(self) -> str: + """ + Extracts and cleans the main text content from the HTML, removing irrelevant tags. + """ + for irrelevant in self._soup.find_all(["script", "style", "img", "figure", "video", "audio", "button", "svg", "canvas", "input", "form", "meta"]): + irrelevant.decompose() + raw_text: str = self._soup.get_text(separator="\n") + cleaned_text: str = " ".join(raw_text.split()) + return cleaned_text if cleaned_text else "No content" + +class Website: + """ + A class to represent a website. + """ + + __DEFAULT_ALLOWED_DOMAINS: list[str] = [".com", ".org", ".net"] + + __title: str = "" + __website_url: str = "" + __text: str = "" + __allowed_domains: list[str] = [] + __links_on_page: list[str] | None = None + + @property + def title(self) -> str: + """ + Returns the title of the website. + """ + return self.__title + + @property + def text(self) -> str: + """ + Returns the main text content of the website. + """ + return self.__text + + @property + def website_url(self) -> str: + """ + Returns the URL of the website. + """ + return self.__website_url + + @property + def links_on_page(self) -> list[str] | None: + """ + Returns the list of links extracted from the website. + """ + return self.__links_on_page + + @property + def _allowed_domains(self) -> list[str]: + """ + Returns the list of allowed domain suffixes. + """ + return self.__allowed_domains + + @_allowed_domains.setter + def _allowed_domains(self, value: list[str] | str) -> None: + """ + Sets the list of allowed domain suffixes. + Filters out empty strings and ensures each suffix starts with a dot. + """ + if isinstance(value, str): + value = [ + item.strip() if item.strip().startswith(".") else f".{item.strip()}" + for item in value.split(",") + if item.strip() + ] + else: + value = [ + item if item.startswith(".") else f".{item}" + for item in value + if item + ] + self.__allowed_domains = value + + def _set_website_url(self, value: str) -> None: + """ + Protected: set the website URL after validating and fetch website data. + Use this from inside the class to initialize or change the URL. + """ + if not value: + raise ValueError("Website URL must be provided") + + parsed_url: ParseResult = urlparse(value) + + self._validate(parsed_url) + + self.__website_url = value + self.__fetch_website_data() + + @property + def fetch_failed(self) -> bool: + """ + Returns whether the website data fetch failed. + """ + return self.__fetch_failed + + def _validate(self, parsed_url: ParseResult) -> None: + """ + Validate the parsed URL. + + Parameters: + parsed_url: The parsed URL to validate. + + Raises: + ValueError: If the URL is missing parts, uses an invalid scheme, + points to a local/private address, or is not in allowed domains. + """ + if not parsed_url.netloc or parsed_url.scheme not in ("http", "https"): + raise ValueError("Website URL must be a valid URL") + + if not parsed_url.hostname: + raise ValueError("Website URL must contain a valid hostname") + + if self.__is_local_address(parsed_url.hostname): + raise ValueError("Website URL must not be a local address") + + if not self.__is_allowed_domain(parsed_url.hostname): + raise ValueError("Website URL must be an allowed domain") + + def __is_local_address(self, hostname: str) -> bool: + """ + Check if the given hostname is a local address. + + Parameters: + hostname (str): The hostname to check. + + Returns: + bool: True if the hostname is a local address, False otherwise. + """ + if hostname in ("localhost", "127.0.0.1", "::1"): + return True + + try: + ip: IPv4Address | IPv6Address = ip_address(hostname) + if ip.is_loopback or ip.is_private or ip.is_link_local or ip.is_reserved: + return True + except ValueError: + return False + + return False + + def __is_allowed_domain(self, hostname: str) -> bool: + """ + Check if the given hostname is an allowed domain. + + Parameters: + hostname (str): The hostname to check. + + Returns: + bool: True if the hostname is an allowed domain, False otherwise. + """ + allowed_domains = [".com", ".org", ".net", ".io"] + return any(hostname.endswith(domain) for domain in allowed_domains) + + def __fetch_website_data(self) -> None: + """ + Fetch website content and populate title, text, and links. + + Side effects: + - Sets internal state: __title, __text, __links_on_page, __fetch_failed. + - Performs an HTTP GET with a browser-like User-Agent. + """ + try: + get_fn = self.__session.get if self.__session else get + response = get_fn( + self.website_url, + timeout=10, + verify=True, + headers={"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/115.0.0.0 Safari/537.36"} + ) + except RequestException as e: + self.__title = "Error" + self.__text = str(e) + self.__fetch_failed = True + return + + if response.ok: + extractor: Extractor = Extractor(response.text) + self.__title = extractor.extracted_title + self.__text = extractor.extracted_text + self.__links_on_page = extractor.extracted_links_on_page + else: + if response.status_code == 404: + self.__title = "Not Found" + self.__text = "The requested page was not found (404)." + else: + self.__title = "Error" + self.__text = f"Error: {response.status_code} - {response.reason}" + self.__fetch_failed = True + + def __init__(self, website_url: str, allowed_domains: list[str] | str | None = None, session: Session | None = None) -> None: + """ + Initializes the Website object and fetches its data. + + Parameters: + website_url (str): The URL of the website to fetch. + allowed_domains (list[str] | str, optional): A list of allowed domain suffixes. + If a string is provided, it should be a comma-separated list of domain suffixes (e.g., ".com,.org,.net"). + session (requests.Session | None, optional): Reused HTTP session for connection pooling. + """ + self.__fetch_failed: bool = False + self.__session: Session | None = session + if allowed_domains is None: + self._allowed_domains = self.__DEFAULT_ALLOWED_DOMAINS.copy() + else: + self._allowed_domains = allowed_domains + # Use protected setter internally so the public API exposes only the getter. + self._set_website_url(website_url) + + def __str__(self) -> str: + """ + Returns a string representation of the Website object. + """ + return f"Website(title={self.title}, url={self.website_url})" \ No newline at end of file diff --git a/week1/community-contributions/brochure-builder-with-multishot-prompting.ipynb b/week1/community-contributions/brochure-builder-with-multishot-prompting.ipynb new file mode 100644 index 0000000..3427a82 --- /dev/null +++ b/week1/community-contributions/brochure-builder-with-multishot-prompting.ipynb @@ -0,0 +1,402 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "9905f163-759f-474b-8f7a-7d14da0df44d", + "metadata": {}, + "source": [ + "### BUSINESS CHALLENGE: Using Multi-shot Prompting\n", + "#### Day 5\n", + "\n", + "Create a product that builds a Brochure for a company to be used for prospective clients, investors and potential recruits.\n", + "\n", + "We will be provided a company name and their primary website." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a0895f24-65ff-4624-8ae0-15d2d400d8f0", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "# If these fail, please check you're running from an 'activated' environment with (llms) in the command prompt\n", + "\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7794aa70-5962-4669-b86f-b53639f4f9ea", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize and constants\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + " \n", + "MODEL = 'gpt-4o-mini'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "63bf8631-2746-4255-bec1-522855d3e812", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "markdown", + "id": "1e7bb527-e769-4245-bb91-ae65e64593ff", + "metadata": {}, + "source": [ + "## First step: Have GPT-4o-mini figure out which links are relevant\n", + "\n", + "### Use a call to gpt-4o-mini to read the links on a webpage, and respond in structured JSON. " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1ce303ae-b967-4261-aadc-02dafa54db4a", + "metadata": {}, + "outputs": [], + "source": [ + "link_system_prompt = \"You are provided with a list of links found on a webpage. \\\n", + "You are able to decide which of the links would be most relevant to include in a brochure about the company, \\\n", + "such as links to an About page, or a Company page, or Careers/Jobs pages.\\n\"\n", + "link_system_prompt += \"You should respond in JSON as in this example:\"\n", + "link_system_prompt += \"\"\"\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"about page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"careers page\", \"url\": \"https://another.full.url/careers\"}\n", + " ]\n", + "}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d24a4c0c-a1d1-4897-b2a7-4128d25c2e08", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links_user_prompt(website):\n", + " user_prompt = f\"Here is the list of links on the website of {website.url} - \"\n", + " user_prompt += \"please decide which of these are relevant web links for a brochure about the company, respond with the full https URL in JSON format. \\\n", + "Do not include Terms of Service, Privacy, email links.\\n\"\n", + " user_prompt += \"Links (some might be relative links):\\n\"\n", + " user_prompt += \"\\n\".join(website.links)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8103fc11-5bc0-41c4-8c97-502c9e96429c", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links(url): # 1st inference\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": link_system_prompt},\n", + " {\"role\": \"user\", \"content\": get_links_user_prompt(website)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dc84a695-515d-4292-9a95-818f4fe3d20e", + "metadata": {}, + "outputs": [], + "source": [ + "huggingface = Website(\"https://huggingface.co\")" + ] + }, + { + "cell_type": "markdown", + "id": "91896908-1632-41fc-9b8b-39a7638d8dd1", + "metadata": {}, + "source": [ + "## Second step: make the brochure!\n", + "\n", + "Assemble all the details into another prompt to GPT4-o" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ab7c54e3-e654-4b1f-8671-09194b628aa0", + "metadata": {}, + "outputs": [], + "source": [ + "def get_all_details(url): # 1st inference wrapper\n", + " result = \"Landing page:\\n\"\n", + " result += Website(url).get_contents()\n", + " links = get_links(url) # inference\n", + " # print(\"Found links:\", links)\n", + " for link in links[\"links\"]:\n", + " result += f\"\\n\\n{link['type']}\\n\"\n", + " result += Website(link[\"url\"]).get_contents()\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ea9f54d1-a248-4c56-a1de-6633193de5bf", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "and creates a short humorous, entertaining, jokey brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "Include details of company culture, customers and careers/jobs if you have the information.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "13412c85-badd-4d79-a5ac-8283e4bb832f", + "metadata": {}, + "outputs": [], + "source": [ + "def get_brochure_user_prompt(company_name, url):\n", + " user_prompt = f\"You are looking at a company called: {company_name}\\n\"\n", + " user_prompt += f\"Here are the contents of its landing page and other relevant pages; use this information to build a short brochure of the company in markdown.\\n\"\n", + " user_prompt += get_all_details(url) # inference wrapper\n", + " user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "addc0047-ea73-4748-abc3-747ff343c134", + "metadata": {}, + "outputs": [], + "source": [ + "def create_brochure(company_name, url): # 2nd inference\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "82a3b61a-da26-4265-840a-0a93f81cd048", + "metadata": {}, + "outputs": [], + "source": [ + "brochure_english = create_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5d165e3f-8fe2-4712-b098-d34d9fabe583", + "metadata": {}, + "outputs": [], + "source": [ + "display(Markdown(brochure_english))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "107a2100-3f7d-4f16-8ba7-b5da602393c6", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_brochure(company_name, url):\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26cbe9b5-3603-49a1-a676-75c7ddaacdb8", + "metadata": {}, + "outputs": [], + "source": [ + "stream_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "markdown", + "id": "c10d8189-7f79-4991-abc4-0764369b7d64", + "metadata": {}, + "source": [ + "### Third step: Translate the entire brochure to Spanish" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "666817eb-1e8b-4fee-bbab-c0dbfe2ea7c0", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a brochure \\\n", + "and translates to Spanish. Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c48adb12-bc3c-48f9-ab38-b7ca895195f6", + "metadata": {}, + "outputs": [], + "source": [ + "def translate_user_prompt(company_name, url):\n", + " user_prompt = f\"Please translate the following brochure content to Spanish\\n\"\n", + " user_prompt += create_brochure(company_name, url) # inference wrapper\n", + " # user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b92b61ac-3be3-4e84-9000-ec8233697b81", + "metadata": {}, + "outputs": [], + "source": [ + "translate_user_prompt(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6bfd04f4-4381-4730-ac5d-c9fa02f906df", + "metadata": {}, + "outputs": [], + "source": [ + "def translate_brochure(): # 3rd inference\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": translate_user_prompt(\"HuggingFace\", \"https://huggingface.co\")}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bb78ed28-9ecd-4c08-ae96-d7473cbc97dd", + "metadata": {}, + "outputs": [], + "source": [ + "translate_brochure()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day-1-bank-account-summarization.ipynb b/week1/community-contributions/day-1-bank-account-summarization.ipynb new file mode 100644 index 0000000..bae0cfe --- /dev/null +++ b/week1/community-contributions/day-1-bank-account-summarization.ipynb @@ -0,0 +1,270 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "f60dab2a-a377-4761-8be3-69a3b8124ca6", + "metadata": { + "editable": true, + "slideshow": { + "slide_type": "" + }, + "tags": [] + }, + "outputs": [], + "source": [ + "import pdfplumber\n", + "import re\n", + "import json\n", + "\n", + "def parse_transaction_line(line):\n", + " # More specific pattern that captures each component'\n", + " pattern = r'^(\\d{2}/\\d{2})\\s+(.+?)\\s+(-?[\\d,]+\\.\\d{2})\\s+(-?[\\d,]+\\.\\d{2})$'\n", + " match = re.match(pattern, line.strip())\n", + " \n", + " if match:\n", + " date, description, amount, balance = match.groups()\n", + " return {\n", + " 'date': date,\n", + " 'description': description.strip(),\n", + " 'amount': amount,\n", + " 'balance': balance\n", + " }\n", + " return None\n", + "\n", + "def parse_Credit_Card_transaction_line(line):\n", + " # More specific pattern that captures each component'\n", + " pattern = r'^(\\d{2}/\\d{2})\\s+(.+?)\\s+(-?[\\d,]+\\.\\d{2})$'\n", + " match = re.match(pattern, line.strip())\n", + " \n", + " if match:\n", + " date, description, amount = match.groups()\n", + " return {\n", + " 'date': date,\n", + " 'description': description.strip(),\n", + " 'amount': amount\n", + " }\n", + " return None\n", + "\n", + "# \n", + "def extract_transactions_CA_from_pdf(pdf_path):\n", + " transactions = []\n", + " \n", + " with pdfplumber.open(pdf_path) as pdf:\n", + " for page in pdf.pages:\n", + " text = page.extract_text()\n", + " for line in text.split(\"\\n\"):\n", + " parsed = parse_transaction_line(line)\n", + " if parsed:\n", + " transactions.append(parsed)\n", + " return transactions\n", + "\n", + "def extract_transactions_CreditCard_from_pdf(pdf_path):\n", + " transactions = []\n", + " \n", + " with pdfplumber.open(pdf_path) as pdf:\n", + " for page in pdf.pages:\n", + " text = page.extract_text()\n", + " for line in text.split(\"\\n\"):\n", + " parsed = parse_Credit_Card_transaction_line(line)\n", + " if parsed:\n", + " transactions.append(parsed)\n", + " return transactions\n", + "# print(transactions, len(transactions)) # check first 10 extracted lines\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "82c34eac-fc30-41d6-8325-77efc48d0dd8", + "metadata": { + "editable": true, + "slideshow": { + "slide_type": "" + }, + "tags": [] + }, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import os\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")" + ] + }, + { + "cell_type": "code", + "execution_count": 49, + "id": "769ee512-75f5-480a-9407-f9c4cd46b679", + "metadata": { + "editable": true, + "slideshow": { + "slide_type": "" + }, + "tags": [] + }, + "outputs": [], + "source": [ + "import pandas as pd\n", + "\n", + "# ---------- STEP 3: Build prompts ----------\n", + "\n", + "def build_prompts(transactions):\n", + " system_prompt = \"\"\"\n", + "You are a personal financial assistant.\n", + "Your job is to analyze bank transactions, categorize each expense into categories such as:\n", + "Food, Clothing, Rent, Utilities, Entertainment, Travel, Health, Miscellaneous, and Others.\n", + "\n", + "Your responsibilities:\n", + "\n", + "Categorize all transactions and compute total spending per category.\n", + "\n", + "Identify the top 5 categories by total spending.\n", + "\n", + "Detect high-frequency purchases, even if individual amounts are small (e.g., $4 coffee bought 40 times).\n", + "\n", + "For these, group transactions by merchant/description and count frequency.\n", + "\n", + "Highlight the top 5 frequent purchases, with both frequency and total spend.\n", + "\n", + "Provide a practical summary of spending habits, covering both biggest expenses and frequent small purchases.\n", + "\n", + "Suggest 2–3 actionable recommendations to reduce spending, targeting both:\n", + "\n", + "Big categories (e.g., Rent, Travel, Entertainment).\n", + "\n", + "Small but frequent “habit expenses” (e.g., coffee, fast food, subscriptions).\n", + "\n", + "The output should be a valid JSON object with this structure:\n", + "{\n", + " \"summary\": {\n", + " \"Food\": ,\n", + " \"Clothing\": ,\n", + " \"Rent\": ,\n", + " \"Utilities\": ,\n", + " \"Entertainment\": ,\n", + " \"Travel\": ,\n", + " \"Health\": ,\n", + " \"Miscellaneous\": ,\n", + " \"Others\": \n", + " },\n", + " \"total_expenses\": ,\n", + " \"top_5_categories\": [ {\"category\": , \"amount\": } ],\n", + " \"top_5_frequent_purchases\": [ {\"item\": , \"count\": , \"total\": } ],\n", + " \"insights\": \"\",\n", + " \"recommendations\": [ \"\", \"\", \"\" ]\n", + "}\n", + "\n", + "\"\"\"\n", + "\n", + " user_prompt = \"Here are my bank account transactions for the past few months:\\n\\n\"\n", + " for txn in transactions:\n", + " user_prompt += f\"- Date: {txn['date']}, Description: {txn['description']}, Amount: {txn['amount']}\\n\"\n", + "\n", + " user_prompt += \"\"\"\n", + "Please analyze these transactions according to the instructions in the system prompt.\n", + "\"\"\"\n", + "\n", + " return system_prompt, user_prompt\n", + "\n", + "\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "307ca02b-2df6-4996-85e7-d073f74592f5", + "metadata": { + "editable": true, + "slideshow": { + "slide_type": "" + }, + "tags": [] + }, + "outputs": [], + "source": [ + "# ---------- STEP 4: Call OpenAI ----------\n", + "def analyze_transactions(pdf_path):\n", + " transactions = extract_transactions_CreditCard_from_pdf(pdf_path)\n", + " system_prompt, user_prompt = build_prompts(transactions)\n", + "\n", + " client = OpenAI() # assumes OPENAI_API_KEY is set in env\n", + "\n", + " response = client.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ],\n", + " response_format={\"type\": \"json_object\"} # ensures valid JSON\n", + " )\n", + "\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)\n", + "\n", + "# ---------- MAIN ----------\n", + "if __name__ == \"__main__\":\n", + " cc_pdf_file = \"cc_statement.pdf\"\n", + " # To Debug in case of failures\n", + " # transactions = extract_transactions_from_pdf(pdf_file)\n", + " # print(cc_transactions,len(cc_transactions))\n", + " # system_prompt, user_prompt = build_prompts(cc_transactions)\n", + " # print(system_prompt, user_prompt)\n", + "\n", + " # Analyse the function to create a smart alert\n", + " cc_transactions = extract_transactions_CreditCard_from_pdf(cc_pdf_file)\n", + " analysis = analyze_transactions(cc_pdf_file)\n", + " print(\"=========================================\")\n", + " print(\"=== Top 5 Spending Habits & Insights ====\")\n", + " print(\"=========================================\")\n", + " print(json.dumps(analysis, indent=2))\n", + " print(\"=========================================\")\n", + " print(\"=========================================\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "831922f4-5efd-4cba-9975-54767b65f6d6", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.9.6" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day-1-thesis_pdf_summarizer.ipynb b/week1/community-contributions/day-1-thesis_pdf_summarizer.ipynb new file mode 100644 index 0000000..e18c68f --- /dev/null +++ b/week1/community-contributions/day-1-thesis_pdf_summarizer.ipynb @@ -0,0 +1,305 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "581151c0-941e-47b3-a3e0-2da65ba70087", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "47353a41-4b47-499e-9460-fd645345f591", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found and looks good so far\n" + ] + } + ], + "source": [ + "load_dotenv()\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if not api_key:\n", + " print('No API key was found')\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"API key is found but is not in the proper format\")\n", + "else:\n", + " print(\"API key found and looks good so far\")" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "dbfbb29a-3452-45a0-b9b3-4e329ac776fb", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "88ffe256-e46a-45e8-a616-0ac574aa7085", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"\"\"You are a research summarizer specialized in wireless communication systems and propagation modeling. Your task is to summarize a research thesis in no more than 1000 words. The summary must be clear, structured, and written in markdown format.\n", + "\n", + "The summary should include the following sections:\n", + "\n", + "1. **Title and Authors** – Provide the full title of the thesis and author name(s).\n", + "2. **Objective / Research Problem** – Clearly state the core research goal or question addressed in the thesis.\n", + "3. **Scientific and Regional Background** – Explain the technical context of radio wave propagation, and why studying it in the Horn of Africa region is important.\n", + "4. **Methodology** – Summarize the modeling techniques, data sources, simulation tools, frequency bands (e.g., microwave, millimeter), and measurement or evaluation methods used.\n", + "5. **Key Findings** – Highlight the quantitative and qualitative results, including differences between precipitation and clear-air conditions, and observed trends across geographic locations.\n", + "6. **Conclusion** – Describe the primary outcomes and how they advance understanding in wireless communications.\n", + "7. **Limitations** – Point out any constraints (e.g., lack of in-situ measurement, simulation assumptions).\n", + "8. **Future Work** – Suggest next steps for improving or extending this research.\n", + "9. **Real-World Applications** – Discuss how the models or findings could improve wireless network planning, 5G deployment, or link budgeting in East Africa and similar regions.\n", + "\n", + "Use academic language but keep it concise, clear, and structured for a technical reader. Output in markdown format only.\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "5f3f7b1a-865f-44cc-854d-9e9e7771eb82", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Requirement already satisfied: ipywidgets in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (8.1.7)\n", + "Collecting pdfplumber\n", + " Downloading pdfplumber-0.11.7-py3-none-any.whl.metadata (42 kB)\n", + "Requirement already satisfied: comm>=0.1.3 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipywidgets) (0.2.3)\n", + "Requirement already satisfied: ipython>=6.1.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipywidgets) (9.4.0)\n", + "Requirement already satisfied: traitlets>=4.3.1 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipywidgets) (5.14.3)\n", + "Requirement already satisfied: widgetsnbextension~=4.0.14 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipywidgets) (4.0.14)\n", + "Requirement already satisfied: jupyterlab_widgets~=3.0.15 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipywidgets) (3.0.15)\n", + "Collecting pdfminer.six==20250506 (from pdfplumber)\n", + " Downloading pdfminer_six-20250506-py3-none-any.whl.metadata (4.2 kB)\n", + "Requirement already satisfied: Pillow>=9.1 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from pdfplumber) (11.3.0)\n", + "Collecting pypdfium2>=4.18.0 (from pdfplumber)\n", + " Downloading pypdfium2-4.30.0-py3-none-win_amd64.whl.metadata (48 kB)\n", + "Requirement already satisfied: charset-normalizer>=2.0.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from pdfminer.six==20250506->pdfplumber) (3.4.3)\n", + "Requirement already satisfied: cryptography>=36.0.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from pdfminer.six==20250506->pdfplumber) (45.0.6)\n", + "Requirement already satisfied: cffi>=1.14 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from cryptography>=36.0.0->pdfminer.six==20250506->pdfplumber) (1.17.1)\n", + "Requirement already satisfied: pycparser in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from cffi>=1.14->cryptography>=36.0.0->pdfminer.six==20250506->pdfplumber) (2.22)\n", + "Requirement already satisfied: colorama in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (0.4.6)\n", + "Requirement already satisfied: decorator in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (5.2.1)\n", + "Requirement already satisfied: ipython-pygments-lexers in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (1.1.1)\n", + "Requirement already satisfied: jedi>=0.16 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (0.19.2)\n", + "Requirement already satisfied: matplotlib-inline in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (0.1.7)\n", + "Requirement already satisfied: prompt_toolkit<3.1.0,>=3.0.41 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (3.0.51)\n", + "Requirement already satisfied: pygments>=2.4.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (2.19.2)\n", + "Requirement already satisfied: stack_data in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (0.6.3)\n", + "Requirement already satisfied: typing_extensions>=4.6 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from ipython>=6.1.0->ipywidgets) (4.14.1)\n", + "Requirement already satisfied: wcwidth in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from prompt_toolkit<3.1.0,>=3.0.41->ipython>=6.1.0->ipywidgets) (0.2.13)\n", + "Requirement already satisfied: parso<0.9.0,>=0.8.4 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from jedi>=0.16->ipython>=6.1.0->ipywidgets) (0.8.4)\n", + "Requirement already satisfied: executing>=1.2.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from stack_data->ipython>=6.1.0->ipywidgets) (2.2.0)\n", + "Requirement already satisfied: asttokens>=2.1.0 in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from stack_data->ipython>=6.1.0->ipywidgets) (3.0.0)\n", + "Requirement already satisfied: pure_eval in c:\\users\\esku4\\anaconda3\\envs\\llms\\lib\\site-packages (from stack_data->ipython>=6.1.0->ipywidgets) (0.2.3)\n", + "Downloading pdfplumber-0.11.7-py3-none-any.whl (60 kB)\n", + "Downloading pdfminer_six-20250506-py3-none-any.whl (5.6 MB)\n", + " ---------------------------------------- 0.0/5.6 MB ? eta -:--:--\n", + " --------------------------------------- 5.5/5.6 MB 30.7 MB/s eta 0:00:01\n", + " ---------------------------------------- 5.6/5.6 MB 22.9 MB/s 0:00:00\n", + "Downloading pypdfium2-4.30.0-py3-none-win_amd64.whl (2.9 MB)\n", + " ---------------------------------------- 0.0/2.9 MB ? eta -:--:--\n", + " ---------------------------------------- 2.9/2.9 MB 28.0 MB/s 0:00:00\n", + "Installing collected packages: pypdfium2, pdfminer.six, pdfplumber\n", + "\n", + " ---------------------------------------- 0/3 [pypdfium2]\n", + " ---------------------------------------- 0/3 [pypdfium2]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " ------------- -------------------------- 1/3 [pdfminer.six]\n", + " -------------------------- ------------- 2/3 [pdfplumber]\n", + " ---------------------------------------- 3/3 [pdfplumber]\n", + "\n", + "Successfully installed pdfminer.six-20250506 pdfplumber-0.11.7 pypdfium2-4.30.0\n", + "Note: you may need to restart the kernel to use updated packages.\n" + ] + } + ], + "source": [ + "pip install ipywidgets pdfplumber" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "06dcfc1d-b106-4b9a-9346-6dd6af4a4015", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "UNIVERSITY OF KWAZULU-NATAL\n", + "Radio Wave Propagation Modeling under\n", + "Precipitation and Clear-air at Microwave\n", + "and Millimetric Bands over Wireless Links\n", + "in the Horn of Africa\n", + "Feyisa Debo Diba\n", + "February, 2017\n", + "Supervisor: Professor Thomas J. Afullo\n", + "Co-supervisor: Dr. Akintunde Ayodeji Alonge\n", + "Radio Wave Propagation Modeling under\n", + "Precipitation and Clear-air at Microwave\n", + "and Millimetric Bands over Wireless Links\n", + "in the Horn of Africa\n", + "Feyisa Debo Diba\n", + "In fulfillment of the Degree of Doctor of Philosophy in\n", + "Electronic Engineering, College of Agriculture, Engineering\n", + "and Science, University of KwaZulu-Natal, Durban\n", + "February, 2017\n", + "Supervisor:\n", + "As the candidate’s Supervisor, I agree/do not agree to the submission of this thesis\n", + "Professor T.J. Afullo ———————————-\n", + "Date—————————————————\n", + "Co-Supervisor:\n", + "Dr. Akintunde Ayodeji Alonge\n", + "As the candidate’s Co.Supervisor, I agree to the submission of this thesis\n", + "Dr. A. A. Alonge ———————————-\n", + "Date—————————————————\n", + "ii\n", + "DECLARATION 1 - PLAGIARISM\n", + "I, Feyisa Debo Diba\n" + ] + } + ], + "source": [ + "# Cell 3: Download and extract from PDF URL\n", + "pdf_url = (\n", + " \"https://researchspace.ukzn.ac.za/server/api/core/bitstreams/\"\n", + " \"29218203-bfc8-4fcb-bc63-9afba3341910/content\"\n", + ")\n", + "\n", + "response = requests.get(pdf_url)\n", + "if response.status_code != 200:\n", + " raise Exception(f\"Failed to download PDF (Status code: {response.status_code})\")\n", + "\n", + "with pdfplumber.open(BytesIO(response.content)) as pdf:\n", + " thesis_text = \"\\n\".join(page.extract_text() for page in pdf.pages if page.extract_text())\n", + "\n", + "# Optional Preview\n", + "print(thesis_text[:1000])\n" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "84c544db-64a0-4181-beb0-1cc72bc88466", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "# Summary of the Research Thesis\n", + "\n", + "## 1. Title and Authors\n", + "**Title:** Radio Wave Propagation Modeling under Precipitation and Clear-air at Microwave and Millimetric Bands over Wireless Links in the Horn of Africa \n", + "**Author:** Feyisa Debo Diba \n", + "**Supervisors:** Professor Thomas J. Afullo, Dr. Akintunde Ayodeji Alonge \n", + "\n", + "## 2. Objective / Research Problem\n", + "The thesis investigates radio wave propagation modeling in clear air and precipitation conditions over wireless communication systems in the Horn of Africa, specifically Ethiopia. The research aims to address the attenuation problem caused by precipitation for systems operating at higher frequency bands.\n", + "\n", + "## 3. Scientific and Regional Background\n", + "The congestion of lower operating frequency bands has led to the rapid growth of utilizing higher frequency spectrum for wireless communication systems. However, the Horn of Africa, particularly Ethiopia, lacks comprehensive studies on propagation modeling under different atmospheric conditions. This research provides valuable insights for the region, contributing to the efficient operation of wireless networks.\n", + "\n", + "## 4. Methodology\n", + "The research uses three years of atmospheric data (temperature, pressure, relative humidity) from the National Meteorological Agency of Ethiopia and clear air signal measurements over terrestrial Line-of-Sight (LOS) links from EthioTelecom. Rainfall data from a Davis Vantage weather station installed at Jimma University, Ethiopia, are also used. The study applies the ITU-R model for refractivity gradient prediction and the Rice-Holmberg (R-H) model for one-minute rain rate distribution. A semi-Markovian model is used for rainfall event characterization and generation.\n", + "\n", + "## 5. Key Findings\n", + "The research derived radio climatological parameters for different rain and clear air fade models. It also proposed rainfall rate conversion factors for Ethiopian sites and developed rainfall rate and fade margin contour maps for Ethiopia. The study found that the sojourn time of spikes in every rain regime is appropriately described by Erlang-k distribution. The number of spikes of generated rainfall events and the corresponding sojourn times follow the power-law relationship.\n", + "\n", + "## 6. Conclusion\n", + "The research provides a comprehensive analysis of radio wave propagation under different atmospheric conditions in Ethiopia. The findings contribute to the understanding of the impact of atmospheric conditions on wireless communication systems operating at higher frequency bands.\n", + "\n", + "## 7. Limitations\n", + "The research is limited by the availability and quality of atmospheric and signal level data. The simulation models also have inherent assumptions that may affect the accuracy of the results.\n", + "\n", + "## 8. Future Work\n", + "Future research could focus on refining the models used in this study by incorporating more data and improving the simulation techniques. Studies could also be extended to other regions in the Horn of Africa.\n", + "\n", + "## 9. Real-World Applications\n", + "The findings of this research can improve wireless network planning and 5G deployment in East Africa. The models developed can also be used in link budgeting, which is crucial for the design and operation of wireless communication systems." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "# Cell 4: Summarize via OpenAI\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": f\"Here is the thesis text (truncated):\\n\\n{thesis_text[:10000]}\"}\n", + "]\n", + "\n", + "response = openai.chat.completions.create(\n", + " model=\"gpt-4\",\n", + " messages=messages,\n", + " temperature=0.3\n", + ")\n", + "\n", + "summary = response.choices[0].message.content.strip()\n", + "display(Markdown(summary))\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e1cdf9ec-5efb-4d4b-8de2-83648865f092", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day01_email_subjectLine_creator.ipynb b/week1/community-contributions/day01_email_subjectLine_creator.ipynb new file mode 100644 index 0000000..5ff45d4 --- /dev/null +++ b/week1/community-contributions/day01_email_subjectLine_creator.ipynb @@ -0,0 +1,124 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "36ef4c36-2905-4485-a46e-dead68cc2dcb", + "metadata": {}, + "outputs": [], + "source": [ + "from openai import OpenAI\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown\n", + "\n", + "# Step 1: Create your prompts\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of \\\n", + " email texts and suggests short subject lines for the email based \\\n", + " on the requested tone and language. Respond in markdown.\"\n", + "user_prompt = \"\"\"\n", + " What’s the hardest-working organ in your body? A lot of people will say the heart. After all, it beats 100,000 times a day. It doesn’t rest. It doesn’t sleep. And it keeps you alive.\n", + "\n", + "\n", + "But we wanted a second opinion. For this edition of From the Armchair, our monthly mental health newsletter, we asked our psychiatrists. Some of them had a different answer: The brain.\n", + "\n", + "\n", + "If that surprises you, think about how your body feels after a long workout—then compare it to how your brain feels after a day of meetings, emails, problem-solving, or just trying to hold it all together. That drained, foggy, overstimulated sensation? That’s cognitive fatigue. And even if your body hasn’t moved much, the exhaustion is very real.\n", + "\n", + "The brain’s quiet hustle\n", + "\n", + "Cognitive fatigue happens when the mental load we’re carrying uses up more fuel than we realize. And the brain is a gas-guzzler. It makes up only 2% of our body weight but consumes about 20% of our energy—just to keep us functioning.\n", + "\n", + "That’s not just because we’re thinking deeply or making big decisions. It’s because the brain is always on: Absorbing information, interpreting social cues, navigating ambiguity, switching between tasks, and trying to make sense of a noisy world. All of that takes effort.\n", + "\n", + "Which brings us to a fallacy: We sometimes think we haven’t “done anything” if we haven’t physically moved or checked something off a list. But the brain doesn’t lift weights to get tired. Its heavy lifting is invisible.\n", + "\n", + "The myth: Motion = accomplishment\n", + "\n", + "There’s a cultural bias that equates movement with productivity. A tired body is seen as earned. A tired mind is often seen as weakness—or worse. Neuroscience disagrees.\n", + "\n", + "The truth is, mental labor—especially the constant decision-making, emotional regulation, and alertness that life demands—can be deeply taxing. Unlike a workout, there's often no clear beginning, middle, or end. Just a low-grade hum that builds over time.\n", + "\n", + "So if you’ve ever said, “Why am I so tired? I didn’t even do anything today,” this could be your answer: Your brain has been sprinting in place.\n", + "\n", + "Mental health and the weight of thinking\n", + "\n", + "Talkiatry psychiatrists note that if you’re living with a mental health condition, this load can feel even heavier. Decisions feel high stakes. Basic tasks can turn into uphill climbs. We can get overloaded with competing signals. Every day stress taxes the circuits we rely on to cope.\n", + "\n", + "While the brain is incredibly resilient, nearly every mental health condition adds friction to thought. That friction requires effort to overcome. That effort burns fuel. And that fuel runs out faster than we expect.\n", + "\n", + "Rest isn’t laziness—it’s repair\n", + "\n", + "This isn’t meant to sound hopeless. In fact, it’s the opposite.\n", + "\n", + " \n", + "\n", + "Recognizing that your brain works hard—even when you don’t realize it—is the first step towards giving it the care it deserves. That might mean rest. It might mean therapy or medication to help you find balance. It might just mean giving yourself credit for the things no one else can see.\n", + "\n", + "\n", + "So if your mind feels tired, believe it. You’re not lazy. You’re human. And you’ve probably done a lot more than you think.\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "AIInputMessages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "] # fill this in\n", + "\n", + "# Step 3: Call OpenAI\n", + "openAI = OpenAI()\n", + "\n", + "response = openAI.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages = AIInputMessages\n", + ")\n", + "\n", + "responseString = response.choices[0].message.content\n", + "# Step 4: print the result\n", + "\n", + "print(\"Printing....\")\n", + "print(responseString)\n", + "\n", + "print(\"\\n\")\n", + "print(\"Displaying....\")\n", + "display(responseString)\n", + "\n", + "print(\"\\n\")\n", + "print(\"Displaying Markdown....\")\n", + "display(Markdown(responseString))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0f8cc568-428d-4ff3-988d-6a31c35db5ba", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-BitcoinMarketPrediction.ipynb b/week1/community-contributions/day1-BitcoinMarketPrediction.ipynb new file mode 100644 index 0000000..dd5265d --- /dev/null +++ b/week1/community-contributions/day1-BitcoinMarketPrediction.ipynb @@ -0,0 +1,230 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "7db973a2-c95e-4939-a0d7-b54edec4d2cf", + "metadata": {}, + "source": [ + "# Bitcoin Market Prediction uisng CoinmarketCap\n", + "An AI-powered project using historical CoinMarketCap data to predict Bitcoin price trends and offer actionable insights for traders." + ] + }, + { + "cell_type": "markdown", + "id": "b792b517-bbc8-4e2c-bff2-45fad1a784dc", + "metadata": {}, + "source": [ + "## Imports" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "51523d62-825a-4a15-aec2-7c910beb5fda", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "markdown", + "id": "2e3816b0-4557-4225-bfb9-9933d813548a", + "metadata": {}, + "source": [ + "## .env configuration" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "02be59e7-01cc-41b5-88c3-a47860570078", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")" + ] + }, + { + "cell_type": "markdown", + "id": "3fc32555-ea4e-45fe-ad44-9dbf4441afd1", + "metadata": {}, + "source": [ + "### This line creates an authenticated OpenAI client instance, used to make API requests in your code." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0845c687-6610-4f83-89e8-fb94bc47ddd2", + "metadata": {}, + "outputs": [], + "source": [ + "from openai import OpenAI\n", + "openai = OpenAI(api_key=api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d140db1a-dd72-4986-8f38-09f8d8f97b00", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fdc96768-94a8-4a08-acf1-32a62b699b94", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"\"\"\n", + "You are an intelligent assistant specialized in Bitcoin market prediction. Your tasks are:\n", + "\n", + "- Collect, preprocess, and analyze historical Bitcoin price and volume data sourced from CoinMarketCap historical data tables or API.\n", + "- Extract relevant time series and technical features from OHLC (open, high, low, close) and volume data.\n", + "- Use machine learning or statistical models to forecast future Bitcoin price trends.\n", + "- Output clear, concise, and actionable insights, focusing on predicted price direction and potential trading signals.\n", + "- Ensure all data collection respects CoinMarketCap’s terms of service.\n", + "- Present findings in user-friendly language, explaining prediction confidence and market risks.\n", + "- Continuously improve prediction accuracy through back-testing on updated datasets.\n", + "\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7d39e983-5b65-4de1-bdf0-e4239c3eb03f", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_for(website):\n", + " user_prompt = f\"You are analyzing historical Bitcoin market data from the webpage titled '{website.title}'.\\n\"\n", + " user_prompt += (\n", + " \"The data includes daily open, high, low, close prices, trading volume, \"\n", + " \"and market capitalization presented in a table format.\\n\"\n", + " \"Please provide a clear and concise analysis in Markdown format, focusing on recent trends, \"\n", + " \"price movements, volatility, and any insights that could help forecast Bitcoin price directions.\\n\"\n", + " \"If possible, include technical indicators, significant patterns, or notable market events mentioned in the data.\\n\\n\"\n", + " )\n", + " user_prompt += website.text\n", + " return user_prompt\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3d41ed3-4753-49f2-b51f-37e8be43102c", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0eb99fcf-75a2-41b8-bf53-568f94264438", + "metadata": {}, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(summary): \n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a0e57921-5132-40c6-834b-03a11a96425c", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://coinmarketcap.com/currencies/bitcoin/historical-data/3\"\n", + "summary = summarize(url)\n", + "display_summary(summary)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "19d9b69a-6493-402d-a0b4-a486c322c816", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-email-subject-creation.ipynb b/week1/community-contributions/day1-email-subject-creation.ipynb new file mode 100644 index 0000000..35e18af --- /dev/null +++ b/week1/community-contributions/day1-email-subject-creation.ipynb @@ -0,0 +1,632 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# YOUR FIRST LAB\n", + "### Please read this section. This is valuable to get you prepared, even if it's a long read -- it's important stuff.\n", + "\n", + "## Your first Frontier LLM Project\n", + "\n", + "Let's build a useful LLM solution - in a matter of minutes.\n", + "\n", + "By the end of this course, you will have built an autonomous Agentic AI solution with 7 agents that collaborate to solve a business problem. All in good time! We will start with something smaller...\n", + "\n", + "Our goal is to code a new kind of Web Browser. Give it a URL, and it will respond with a summary. The Reader's Digest of the internet!!\n", + "\n", + "Before starting, you should have completed the setup for [PC](../SETUP-PC.md) or [Mac](../SETUP-mac.md) and you hopefully launched this jupyter lab from within the project root directory, with your environment activated.\n", + "\n", + "## If you're new to Jupyter Lab\n", + "\n", + "Welcome to the wonderful world of Data Science experimentation! Once you've used Jupyter Lab, you'll wonder how you ever lived without it. Simply click in each \"cell\" with code in it, such as the cell immediately below this text, and hit Shift+Return to execute that cell. As you wish, you can add a cell with the + button in the toolbar, and print values of variables, or try out variations. \n", + "\n", + "I've written a notebook called [Guide to Jupyter](Guide%20to%20Jupyter.ipynb) to help you get more familiar with Jupyter Labs, including adding Markdown comments, using `!` to run shell commands, and `tqdm` to show progress.\n", + "\n", + "## If you're new to the Command Line\n", + "\n", + "Please see these excellent guides: [Command line on PC](https://chatgpt.com/share/67b0acea-ba38-8012-9c34-7a2541052665) and [Command line on Mac](https://chatgpt.com/canvas/shared/67b0b10c93a081918210723867525d2b). \n", + "\n", + "## If you'd prefer to work in IDEs\n", + "\n", + "If you're more comfortable in IDEs like VSCode, Cursor or PyCharm, they both work great with these lab notebooks too. \n", + "If you'd prefer to work in VSCode, [here](https://chatgpt.com/share/676f2e19-c228-8012-9911-6ca42f8ed766) are instructions from an AI friend on how to configure it for the course.\n", + "\n", + "## If you'd like to brush up your Python\n", + "\n", + "I've added a notebook called [Intermediate Python](Intermediate%20Python.ipynb) to get you up to speed. But you should give it a miss if you already have a good idea what this code does: \n", + "`yield from {book.get(\"author\") for book in books if book.get(\"author\")}`\n", + "\n", + "## I am here to help\n", + "\n", + "If you have any problems at all, please do reach out. \n", + "I'm available through the platform, or at ed@edwarddonner.com, or at https://www.linkedin.com/in/eddonner/ if you'd like to connect (and I love connecting!) \n", + "And this is new to me, but I'm also trying out X/Twitter at [@edwarddonner](https://x.com/edwarddonner) - if you're on X, please show me how it's done 😂 \n", + "\n", + "## More troubleshooting\n", + "\n", + "Please see the [troubleshooting](troubleshooting.ipynb) notebook in this folder to diagnose and fix common problems. At the very end of it is a diagnostics script with some useful debug info.\n", + "\n", + "## For foundational technical knowledge (eg Git, APIs, debugging) \n", + "\n", + "If you're relatively new to programming -- I've got your back! While it's ideal to have some programming experience for this course, there's only one mandatory prerequisite: plenty of patience. 😁 I've put together a set of self-study guides that cover Git and GitHub, APIs and endpoints, beginner python and more.\n", + "\n", + "This covers Git and GitHub; what they are, the difference, and how to use them: \n", + "https://github.com/ed-donner/agents/blob/main/guides/03_git_and_github.ipynb\n", + "\n", + "This covers technical foundations: \n", + "ChatGPT vs API; taking screenshots; Environment Variables; Networking basics; APIs and endpoints: \n", + "https://github.com/ed-donner/agents/blob/main/guides/04_technical_foundations.ipynb\n", + "\n", + "This covers Python for beginners, and making sure that a `NameError` never trips you up: \n", + "https://github.com/ed-donner/agents/blob/main/guides/06_python_foundations.ipynb\n", + "\n", + "This covers the essential techniques for figuring out errors: \n", + "https://github.com/ed-donner/agents/blob/main/guides/08_debugging.ipynb\n", + "\n", + "And you'll find other useful guides in the same folder in GitHub. Some information applies to my other Udemy course (eg Async Python) but most of it is very relevant for LLM engineering.\n", + "\n", + "## If this is old hat!\n", + "\n", + "If you're already comfortable with today's material, please hang in there; you can move swiftly through the first few labs - we will get much more in depth as the weeks progress. Ultimately we will fine-tune our own LLM to compete with OpenAI!\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Please read - important note

\n", + " The way I collaborate with you may be different to other courses you've taken. I prefer not to type code while you watch. Rather, I execute Jupyter Labs, like this, and give you an intuition for what's going on. My suggestion is that you carefully execute this yourself, after watching the lecture. Add print statements to understand what's going on, and then come up with your own variations. If you have a Github account, use this to showcase your variations. Not only is this essential practice, but it demonstrates your skills to others, including perhaps future clients or employers...\n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

This code is a live resource - keep an eye out for my emails

\n", + " I push updates to the code regularly. As people ask questions, I add more examples or improved commentary. As a result, you'll notice that the code below isn't identical to the videos. Everything from the videos is here; but I've also added better explanations and new models like DeepSeek. Consider this like an interactive book.

\n", + " I try to send emails regularly with important updates related to the course. You can find this in the 'Announcements' section of Udemy in the left sidebar. You can also choose to receive my emails via your Notification Settings in Udemy. I'm respectful of your inbox and always try to add value with my emails!\n", + " \n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business value of these exercises

\n", + " A final thought. While I've designed these notebooks to be educational, I've also tried to make them enjoyable. We'll do fun things like have LLMs tell jokes and argue with each other. But fundamentally, my goal is to teach skills you can apply in business. I'll explain business implications as we go, and it's worth keeping this in mind: as you build experience with models and techniques, think of ways you could put this into action at work today. Please do contact me if you'd like to discuss more or if you have ideas to bounce off me.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "markdown", + "id": "6900b2a8-6384-4316-8aaa-5e519fca4254", + "metadata": {}, + "source": [ + "# Connecting to OpenAI (or Ollama)\n", + "\n", + "The next cell is where we load in the environment variables in your `.env` file and connect to OpenAI. \n", + "\n", + "If you'd like to use free Ollama instead, please see the README section \"Free Alternative to Paid APIs\", and if you're not sure how to do this, there's a full solution in the solutions folder (day1_with_ollama.ipynb).\n", + "\n", + "## Troubleshooting if you have problems:\n", + "\n", + "Head over to the [troubleshooting](troubleshooting.ipynb) notebook in this folder for step by step code to identify the root cause and fix it!\n", + "\n", + "If you make a change, try restarting the \"Kernel\" (the python process sitting behind this notebook) by Kernel menu >> Restart Kernel and Clear Outputs of All Cells. Then try this notebook again, starting at the top.\n", + "\n", + "Or, contact me! Message me or email ed@edwarddonner.com and we will get this to work.\n", + "\n", + "Any concerns about API costs? See my notes in the README - costs should be minimal, and you can control it at every point. You can also use Ollama as a free alternative, which we discuss during Day 2." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "markdown", + "id": "442fc84b-0815-4f40-99ab-d9a5da6bda91", + "metadata": {}, + "source": [ + "# Let's make a quick call to a Frontier model to get started, as a preview!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a58394bf-1e45-46af-9bfd-01e24da6f49a", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with these messages is this easy. Any problems, head over to the Troubleshooting notebook.\n", + "\n", + "message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\":\"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "2aa190e5-cb31-456a-96cc-db109919cd78", + "metadata": {}, + "source": [ + "## OK onwards with our first project" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's try one out. Change the website and add print statements to follow along.\n", + "\n", + "ed = Website(\"https://edwarddonner.com\")\n", + "print(ed.title)\n", + "print(ed.text)" + ] + }, + { + "cell_type": "markdown", + "id": "6a478a0c-2c53-48ff-869c-4d08199931e1", + "metadata": {}, + "source": [ + "## Types of prompts\n", + "\n", + "You may know this already - but if not, you will get very familiar with it!\n", + "\n", + "Models like GPT4o have been trained to receive instructions in a particular way.\n", + "\n", + "They expect to receive:\n", + "\n", + "**A system prompt** that tells them what task they are performing and what tone they should use\n", + "\n", + "**A user prompt** -- the conversation starter that they should reply to" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0275b1b-7cfe-4f9d-abfa-7650d378da0c", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bcd09f4b-a2c4-4274-acec-b3b7c0ac883a", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(ed))" + ] + }, + { + "cell_type": "markdown", + "id": "ea211b5f-28e1-4a86-8e52-c0b7677cadcc", + "metadata": {}, + "source": [ + "## Messages\n", + "\n", + "The API from OpenAI expects to receive messages in a particular structure.\n", + "Many of the other APIs share this structure:\n", + "\n", + "```python\n", + "[\n", + " {\"role\": \"system\", \"content\": \"system message goes here\"},\n", + " {\"role\": \"user\", \"content\": \"user message goes here\"}\n", + "]\n", + "```\n", + "To give you a preview, the next 2 cells make a rather simple call - we won't stretch the mighty GPT (yet!)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f25dcd35-0cd0-4235-9f64-ac37ed9eaaa5", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"system\", \"content\": \"You are a snarky assistant\"},\n", + " {\"role\": \"user\", \"content\": \"What is 2 + 2?\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "21ed95c5-7001-47de-a36d-1d6673b403ce", + "metadata": {}, + "outputs": [], + "source": [ + "# To give you a preview -- calling OpenAI with system and user messages:\n", + "\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "d06e8d78-ce4c-4b05-aa8e-17050c82bb47", + "metadata": {}, + "source": [ + "## And now let's build useful messages for GPT-4o-mini, using a function" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "# See how this function creates exactly the format above\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "36478464-39ee-485c-9f3f-6a4e458dbc9c", + "metadata": {}, + "outputs": [], + "source": [ + "# Try this out, and then try for a few more websites\n", + "\n", + "messages_for(ed)" + ] + }, + { + "cell_type": "markdown", + "id": "16f49d46-bf55-4c3e-928f-68fc0bf715b0", + "metadata": {}, + "source": [ + "## Time to bring it together - the API for OpenAI is very simple!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [], + "source": [ + "summarize(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "b3bcf6f4-adce-45e9-97ad-d9a5d7a3a624", + "metadata": {}, + "source": [ + "# Let's try more websites\n", + "\n", + "Note that this will only work on websites that can be scraped using this simplistic approach.\n", + "\n", + "Websites that are rendered with Javascript, like React apps, won't show up. See the community-contributions folder for a Selenium implementation that gets around this. You'll need to read up on installing Selenium (ask ChatGPT!)\n", + "\n", + "Also Websites protected with CloudFront (and similar) may give 403 errors - many thanks Andy J for pointing this out.\n", + "\n", + "But many websites will work just fine!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "45d83403-a24c-44b5-84ac-961449b4008f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "75e9fd40-b354-4341-991e-863ef2e59db7", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://anthropic.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "c951be1a-7f1b-448f-af1f-845978e47e2c", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business applications

\n", + " In this exercise, you experienced calling the Cloud API of a Frontier Model (a leading model at the frontier of AI) for the first time. We will be using APIs like OpenAI at many stages in the course, in addition to building our own LLMs.\n", + "\n", + "More specifically, we've applied this to Summarization - a classic Gen AI use case to make a summary. This can be applied to any business vertical - summarizing the news, summarizing financial performance, summarizing a resume in a cover letter - the applications are limitless. Consider how you could apply Summarization in your business, and try prototyping a solution.\n", + "
\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Before you continue - now try yourself

\n", + " Use the cell below to make your own simple commercial example. Stick with the summarization use case for now. Here's an idea: write something that will take the contents of an email, and will suggest an appropriate short subject line for the email. That's the kind of feature that might be built into a commercial email tool.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the data of an email and suggest appropriate subject line for the email\"\n", + "user_prompt = \"\"\"\n", + " Read the email below and suggest a short subject line for the email\n", + " Dear John\n", + "I hope this email finds you well.\n", + "I am writing to request a meeting to discuss the timeline for Project X. I believe a discussion would be beneficial to ensure alignment on deliverables and milestones.\n", + "Would you be available for a 30-minute meeting on Thursday, August 14th at 2:00 PM PDT, or Friday, August 15th at 10:00 AM PDT? Please let me know if either of these times work for you, or feel free to suggest an alternative time that better suits your schedule.\n", + "Thank you for your time and consideration. I look forward to hearing from you soon.\n", + "Best regards,\n", + "Scott\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [ {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}] # fill this in\n", + "\n", + "# Step 3: Call OpenAI\n", + "\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "\n", + "# Step 4: print the result\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "36ed9f14-b349-40e9-a42c-b367e77f8bda", + "metadata": {}, + "source": [ + "## An extra exercise for those who enjoy web scraping\n", + "\n", + "You may notice that if you try `display_summary(\"https://openai.com\")` - it doesn't work! That's because OpenAI has a fancy website that uses Javascript. There are many ways around this that some of you might be familiar with. For example, Selenium is a hugely popular framework that runs a browser behind the scenes, renders the page, and allows you to query it. If you have experience with Selenium, Playwright or similar, then feel free to improve the Website class to use them. In the community-contributions folder, you'll find an example Selenium solution from a student (thank you!)" + ] + }, + { + "cell_type": "markdown", + "id": "eeab24dc-5f90-4570-b542-b0585aca3eb6", + "metadata": {}, + "source": [ + "# Sharing your code\n", + "\n", + "I'd love it if you share your code afterwards so I can share it with others! You'll notice that some students have already made changes (including a Selenium implementation) which you will find in the community-contributions folder. If you'd like add your changes to that folder, submit a Pull Request with your new versions in that folder and I'll merge your changes.\n", + "\n", + "If you're not an expert with git (and I am not!) then GPT has given some nice instructions on how to submit a Pull Request. It's a bit of an involved process, but once you've done it once it's pretty clear. As a pro-tip: it's best if you clear the outputs of your Jupyter notebooks (Edit >> Clean outputs of all cells, and then Save) for clean notebooks.\n", + "\n", + "Here are good instructions courtesy of an AI friend: \n", + "https://chatgpt.com/share/677a9cb5-c64c-8012-99e0-e06e88afd293" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-email-subject-implementation.ipynb b/week1/community-contributions/day1-email-subject-implementation.ipynb new file mode 100644 index 0000000..e968e7c --- /dev/null +++ b/week1/community-contributions/day1-email-subject-implementation.ipynb @@ -0,0 +1,115 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"Eres un analista acostumbrado a trabajar con correos electrónicos que contiene un gran conocimiento sobre la mejor manera de resumir contenido releveante \\\n", + "dejando de lado cualquier información que no despierte interés o no sea el tema principal del correo. Tu función será leer contenido de correos y definir un listado de las 3 mejores opciones con el formato: Opción *numero de la opción*: *sujeto* Motivo: *que palabras clave dentro del texto has utilizado para llegar a esa conclusion y la relación semántica con tu idea\"\n", + "user_prompt = \"\"\"\n", + "Tengo un correo que le quiero enviar a mi profesor pero no se muy bien como llamarlo, ayudame. El correo es el siguiente:\n", + "Hola profe,\n", + "Ultimamente estoy disfrutando mucho sus clases y la información que presenta me parece muy importante. Este fin de semana me voy de vacaciones y no podré\n", + "ir a sus clases la semana que viene. Me gustaría si pudiera pasarme los pdfs de la siguiente semana para echarle un vistazo por mi cuenta durante mi ausencia en Francia.\n", + "\n", + "Un saludo,\n", + "Daniel.\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [{\"role\" : \"system\" , \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}]\n", + "\n", + "# Step 3: Call OpenAI\n", + "\n", + "response = openai.chat.completions.create( \n", + " model = \"gpt-4o-mini\",\n", + " messages = messages)\n", + "\n", + "# Step 4: print the result\n", + "\n", + "print(response.choices[0].message.content)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-job-search-assitant.ipynb b/week1/community-contributions/day1-job-search-assitant.ipynb new file mode 100644 index 0000000..20a5fd5 --- /dev/null +++ b/week1/community-contributions/day1-job-search-assitant.ipynb @@ -0,0 +1,278 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 23, + "id": "5cbb8ddf-bc86-4da0-96eb-b4971b9bf3a3", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from bs4 import BeautifulSoup\n", + "from openai import OpenAI\n", + "from dotenv import load_dotenv\n", + "load_dotenv()\n", + "import requests\n", + "from IPython.display import Markdown" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "id": "f76bea4c-95cf-47ae-9236-75e866320470", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " def __init__(self,url):\n", + " self.url = url\n", + " response = requests.get(self.url, headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " self.text = soup.get_text(separator=\" \").lower() " + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "id": "d0ed56fb-af44-42ad-9235-c588ca40edc8", + "metadata": {}, + "outputs": [], + "source": [ + "job_search = Website(\"https://www.google.com/about/careers/applications/jobs\")" + ] + }, + { + "cell_type": "code", + "execution_count": 28, + "id": "38468bd0-5d95-4944-b371-107300495ebf", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are a job search assistant who finds real-time DevOps-related job listings from career pages, job boards, and developer platforms. Return results with job title, company name, and a link to the listing. Focus on DevOps, SRE, Platform Engineering, and CI/CD tooling roles.\"\n", + "user_prompt = f\"\"\"\n", + "Here is a list of job postings:\n", + "\n", + "{job_search.text}\n", + "\n", + "Please extract only the jobs that are clearly related to:\n", + "- DevOps\n", + "- Site Reliability Engineering (SRE)\n", + "- Platform Engineering\n", + "- CI/CD or Infrastructure\n", + "\n", + "Exclude roles like sales, instructors, analysts, and anything not related to DevOps tools.\n", + "\n", + "For each DevOps-related job, return:\n", + "- Job Title\n", + "- Company\n", + "- Location\n", + "- Years of Experience\n", + "- Skill set required\n", + "- (if available) Whether it's remote\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "id": "927af617-0d87-48de-ac0a-751900b4a495", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " { \"role\": \"system\", \"content\": system_prompt },\n", + " { \"role\": \"user\", \"content\": user_prompt }\n", + "]\n", + "\n", + "openai = OpenAI(\n", + " api_key=os.getenv('GROQ_API_KEY'),\n", + " base_url=\"https://api.groq.com/openai/v1\" \n", + " )\n", + "response = openai.chat.completions.create(\n", + " model = \"gemma2-9b-it\",\n", + " messages = messages\n", + " )" + ] + }, + { + "cell_type": "code", + "execution_count": 30, + "id": "2ded75af-ba51-4e21-a581-d8da82439e2e", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Here are the DevOps-related jobs from your provided Google Careers listing:\n", + "\n", + "**1. Technical Lead Manager, Google Notifications** \n", + "\n", + "* **Company:** Google \n", + "* **Location:** Tel Aviv, Israel\n", + "* **Years of Experience:** 8+ years\n", + "* **Skillset:** 8+ years of software development experience, 5+ years of experience testing, launching software products, 3+ years of experience with software design and architecture, experience in a leadership role or people management.\n", + "\n", + " **2. Senior System Software Engineer, Embedded Systems, Firmware, Pixel**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** New Taipei, Banqiao District, New Taipei City, Taiwan\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** 5+ years of experience with software development in C, Android BSP and Linux drivers, 3+ years of experience testing, maintaining, or launching software products, 1+ years of experience with software design and architecture, 3+ years of experience working with embedded operating systems.\n", + "\n", + " **3. Senior Technical Program Manager, Infrastructure Deployment Software**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** Sunnyvale, CA, USA\n", + "* **Years of Experience:** 8+ years\n", + "* **Skillset:** Bachelor’s Degree in a relevant field, 8+ years of program management experience, 8+ years of infrastructure experience.\n", + "\n", + "**4. Network Engineer**\n", + "\n", + "* **Company:** Google\n", + "* **Location:** Bengaluru, Karnataka, India\n", + "* **Years of Experience:** 3+ years\n", + "* **Skillset:** Bachelor’s Degree in Computer Science, Engineering, a related field, or equivalent practical experience; 3+ years of experience with network routing protocols, design and troubleshooting, with network equipment providers; Experience with network operations at scale; broad understanding of cloud computing, Linux server environments, network design and deployment, managing large scale infrastructure; experience working in a customer-facing role at an ISP, carrier, content network, or CDN; experience programming in either Python or Go.\n", + "\n", + "**5. Field Activation Lead, Google Cloud**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** Warsaw, Poland\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** Bachelor's degree or equivalent practical experience; 5+ years of experience in program or project management in an enterprise, cloud, or technology environment; experience working with executive-level clients or stakeholders.\n", + "\n", + "**6. Software Engineer III, Infrastructure, Google Cloud Business Platforms**\n", + "* **Company:** Google \n", + "* **Location:** Cambridge, MA, USA\n", + "* **Years of Experience:** 2+ years \n", + "* **Skillset:** Bachelor’s degree or equivalent practical experience; 2 years of experience with software development in one or more programming languages (C, C++, Python, or Go), or 1 year of experience with an advanced degree; 2 years of experience with developing large-scale infrastructure, distributed systems or networks, or experience with compute technologies, storage or hardware architecture.\n", + "\n", + " **7. Networking Performance Modeling Architect**\n", + "\n", + "* **Company:** Google\n", + "* **Location:** Tel Aviv, Israel; Haifa, Israel\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** Bachelor’s degree in Electrical engineering, computer engineering, Computer Science, or equivalent practical experience; 5 years of experience with software development in C++, and data structures/algorithms; experience in performance modeling, performance analysis, and workload characterization.\n", + "\n", + "\n", + "\n", + "Let me know if you'd like me to search for more specific roles or on other platforms!\n", + "\n" + ] + } + ], + "source": [ + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "id": "6b8bd531-c537-4792-a450-8c06e035172d", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "Here are the DevOps-related jobs from your provided Google Careers listing:\n", + "\n", + "**1. Technical Lead Manager, Google Notifications** \n", + "\n", + "* **Company:** Google \n", + "* **Location:** Tel Aviv, Israel\n", + "* **Years of Experience:** 8+ years\n", + "* **Skillset:** 8+ years of software development experience, 5+ years of experience testing, launching software products, 3+ years of experience with software design and architecture, experience in a leadership role or people management.\n", + "\n", + " **2. Senior System Software Engineer, Embedded Systems, Firmware, Pixel**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** New Taipei, Banqiao District, New Taipei City, Taiwan\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** 5+ years of experience with software development in C, Android BSP and Linux drivers, 3+ years of experience testing, maintaining, or launching software products, 1+ years of experience with software design and architecture, 3+ years of experience working with embedded operating systems.\n", + "\n", + " **3. Senior Technical Program Manager, Infrastructure Deployment Software**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** Sunnyvale, CA, USA\n", + "* **Years of Experience:** 8+ years\n", + "* **Skillset:** Bachelor’s Degree in a relevant field, 8+ years of program management experience, 8+ years of infrastructure experience.\n", + "\n", + "**4. Network Engineer**\n", + "\n", + "* **Company:** Google\n", + "* **Location:** Bengaluru, Karnataka, India\n", + "* **Years of Experience:** 3+ years\n", + "* **Skillset:** Bachelor’s Degree in Computer Science, Engineering, a related field, or equivalent practical experience; 3+ years of experience with network routing protocols, design and troubleshooting, with network equipment providers; Experience with network operations at scale; broad understanding of cloud computing, Linux server environments, network design and deployment, managing large scale infrastructure; experience working in a customer-facing role at an ISP, carrier, content network, or CDN; experience programming in either Python or Go.\n", + "\n", + "**5. Field Activation Lead, Google Cloud**\n", + "\n", + "* **Company:** Google \n", + "* **Location:** Warsaw, Poland\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** Bachelor's degree or equivalent practical experience; 5+ years of experience in program or project management in an enterprise, cloud, or technology environment; experience working with executive-level clients or stakeholders.\n", + "\n", + "**6. Software Engineer III, Infrastructure, Google Cloud Business Platforms**\n", + "* **Company:** Google \n", + "* **Location:** Cambridge, MA, USA\n", + "* **Years of Experience:** 2+ years \n", + "* **Skillset:** Bachelor’s degree or equivalent practical experience; 2 years of experience with software development in one or more programming languages (C, C++, Python, or Go), or 1 year of experience with an advanced degree; 2 years of experience with developing large-scale infrastructure, distributed systems or networks, or experience with compute technologies, storage or hardware architecture.\n", + "\n", + " **7. Networking Performance Modeling Architect**\n", + "\n", + "* **Company:** Google\n", + "* **Location:** Tel Aviv, Israel; Haifa, Israel\n", + "* **Years of Experience:** 5+ years\n", + "* **Skillset:** Bachelor’s degree in Electrical engineering, computer engineering, Computer Science, or equivalent practical experience; 5 years of experience with software development in C++, and data structures/algorithms; experience in performance modeling, performance analysis, and workload characterization.\n", + "\n", + "\n", + "\n", + "Let me know if you'd like me to search for more specific roles or on other platforms!\n" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "display(Markdown(response.choices[0].message.content))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c6bd6d14-e0df-45be-99e2-55aa4d96f53b", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-research-paper-summarizer-with-highlighter.ipynb b/week1/community-contributions/day1-research-paper-summarizer-with-highlighter.ipynb new file mode 100644 index 0000000..74a00f9 --- /dev/null +++ b/week1/community-contributions/day1-research-paper-summarizer-with-highlighter.ipynb @@ -0,0 +1,202 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "5c527a13-459e-4a46-b00e-f2c5056de155", + "metadata": {}, + "source": [ + "# Research Paper Summarizer with Text Highlighting" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "861a0be5-6da7-4f66-8f82-bc083a913f9f", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "74bf6765-53b6-457b-ac2d-0d1afa7fbf8f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found and looks good so far!\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "227ed7af-d539-4c87-988b-80e6e049c863", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "dcaadf8b-456d-48ca-af9d-9f57d3414308", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "6315093f-be68-408e-a5e1-6a2e4ea675e8", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at an article website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "I'm also looking for complete statements containing the following keywords (if found): \\\n", + "'large circuit model', 'ChipGPT' \\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt\n", + "\n", + "\n", + "article = Website(\"https://arxiv.org/html/2401.12224v1\")\n", + "# print(user_prompt_for(article))" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "ff8a4112-f118-4866-b6cf-82675de0a38d", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a scientific \\\n", + "article for a PhD student (who has to read a lot of papers and journals). The \\\n", + "user will provide the article website and keyword(s) they are looking to learn and \\\n", + "cite from. Your job is to summarize the paper and point out all the statements \\\n", + "containing the specific keyword(s) the user typed. \\\n", + "Respond in markdown.\"\n", + "\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]\n", + "\n", + " \n", + "#messages_for(article)" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "b5e47bea-403d-48c3-ab9d-4d6adef83241", + "metadata": {}, + "outputs": [], + "source": [ + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9f6ac1bc-5bc8-4daa-8174-d201400e517a", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://arxiv.org/html/2401.12224v1\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-reviewsSummary.ipynb b/week1/community-contributions/day1-reviewsSummary.ipynb new file mode 100644 index 0000000..910894f --- /dev/null +++ b/week1/community-contributions/day1-reviewsSummary.ipynb @@ -0,0 +1,130 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = f\"\"\"\n", + " You are an assistant that will analyze the ratings & reviews from :\\n\\n{reviews_text}\\n\\n and comeup with a summary of how many 5,4,3,2,1 star rating the restuarnat has. \n", + " You will also come up with a summary of the reviews showing what the customers love about the restaurant and what they dont like. Also extract the name of the restaurant,\n", + " the location and the cuisine. Respond in markdown\"\"\"\n", + "user_prompt = \"This is the summary for the restaurant: \"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "] \n", + "\n", + "def generate_review_summary(reviews_text):\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "try:\n", + " with open('restaurant_reviews.txt', 'r') as file:\n", + " reviews_text = file.read()\n", + " \n", + " # Generate review summary\n", + " summary = generate_review_summary(reviews_text)\n", + " display(Markdown(summary))\n", + "\n", + "except FileNotFoundError:\n", + " print(\"The specified reviews file was not found. Please ensure 'restaurant_reviews.txt' is in the correct directory.\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3eccbf35-0a0b-4a1b-b493-aa5c342109cc", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1-selenium-web-summary-es-mx.ipynb b/week1/community-contributions/day1-selenium-web-summary-es-mx.ipynb new file mode 100644 index 0000000..2a3de8b --- /dev/null +++ b/week1/community-contributions/day1-selenium-web-summary-es-mx.ipynb @@ -0,0 +1,260 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2588fbba", + "metadata": {}, + "source": [ + "# Website Analysis and Summarization with Selenium and OpenAI\n", + "\n", + "> This notebook demonstrates how to extract and summarize the main content of any website using Selenium for dynamic extraction and OpenAI for generating concise summaries in Mexican Spanish.\n", + "\n", + "## Overview\n", + "This notebook provides a workflow to automatically analyze websites, extract relevant text, and generate a short summary using a language model. Navigation elements are ignored, focusing on news, announcements, and main content.\n", + "\n", + "## Features\n", + "- Extracts relevant text from web pages using Selenium and BeautifulSoup.\n", + "- Generates automatic summaries using OpenAI's language models.\n", + "- Presents results in markdown format.\n", + "\n", + "## Requirements\n", + "- Python 3.8+\n", + "- Google Chrome browser installed\n", + "- The following Python packages:\n", + " - selenium\n", + " - webdriver-manager\n", + " - beautifulsoup4\n", + " - openai\n", + " - python-dotenv\n", + " - requests\n", + "- An OpenAI API key (project key, starting with `sk-proj-`)\n", + "- Internet connection\n", + "\n", + "## How to Use\n", + "1. Install the required packages:\n", + " ```bash\n", + " pip install selenium webdriver-manager undetected-chromedriver beautifulsoup4 openai python-dotenv requests\n", + " ```\n", + "2. Add your OpenAI API key to a `.env` file as `OPENAI_API_KEY`.\n", + "3. Run the notebook cells in order. You can change the target website URL in the code to analyze different sites.\n", + "4. The summary will be displayed in markdown format below the code cell.\n", + "\n", + "**Note:** Some websites may block automated access. The notebook includes options to simulate a real user and avoid bot detection, but results may vary depending on the site's protections.\n", + "\n", + "---" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dc7c2ade", + "metadata": {}, + "outputs": [], + "source": [ + "# Imports\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "from selenium import webdriver\n", + "from selenium.webdriver.chrome.service import Service\n", + "from selenium.webdriver.common.by import By\n", + "from selenium.webdriver.chrome.options import Options\n", + "from selenium.webdriver.support.ui import WebDriverWait\n", + "from selenium.webdriver.support import expected_conditions as EC\n", + "from webdriver_manager.chrome import ChromeDriverManager\n", + "import undetected_chromedriver as uc" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a2d21987", + "metadata": {}, + "outputs": [], + "source": [ + "# Load the environment variables from .env\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bbb3a8ed", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5313aa64", + "metadata": {}, + "outputs": [], + "source": [ + "class Website:\n", + " def __init__(self, url, headless=True, wait_time=10):\n", + " self.url = url # Website URL to analyze\n", + " self.title = None # Title of the website\n", + " self.text = None # Extracted text from the website\n", + " \n", + " # Chrome options configuration for Selenium\n", + " options = Options()\n", + " if headless:\n", + " options.add_argument(\"--headless=new\") # Run Chrome in headless mode (no window)\n", + " options.add_argument(\"--disable-gpu\") # Disable GPU acceleration\n", + " options.add_argument(\"--no-sandbox\") # Disable Chrome sandbox (required for some environments)\n", + " options.add_argument(\"--window-size=1920,1080\") # Set window size to simulate a real user\n", + " # Simulate a real user-agent to avoid bot detection\n", + " options.add_argument(\"user-agent=Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36\")\n", + " \n", + " # Initialize Chrome WebDriver\n", + " self.driver = webdriver.Chrome(service=Service(ChromeDriverManager().install()), options=options)\n", + " self.driver.get(url) # Open the URL in the browser\n", + " \n", + " try:\n", + " # Wait until the element is present in the page\n", + " WebDriverWait(self.driver, wait_time).until(EC.presence_of_element_located((By.TAG_NAME, \"body\")))\n", + " html = self.driver.page_source # Get the full HTML of the page\n", + " soup = BeautifulSoup(html, 'html.parser') # Parse HTML with BeautifulSoup\n", + " self.title = soup.title.string if soup.title else 'No title found' # Extract the title\n", + " if soup.body:\n", + " # Remove irrelevant elements from the body\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " # Extract clean text from the body\n", + " self.text = soup.body.get_text(separator='\\n', strip=True)\n", + " else:\n", + " self.text = \"No body found\" # If no body is found, indicate it\n", + " except Exception as e:\n", + " print(f\"Error accessing the site: {e}\") # Print error to console\n", + " self.text = \"Error accessing the site\" # Store error in the attribute\n", + " finally:\n", + " self.driver.quit() # Always close the browser, whether or not an error occurred" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e902c6b2", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown in Mexican Spanish.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eaee8f36", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9ac4ed8b", + "metadata": {}, + "outputs": [], + "source": [ + "# Creates messages for the OpenAI API\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1536d537", + "metadata": {}, + "outputs": [], + "source": [ + "# Creates a summary for the given URL\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fe135339", + "metadata": {}, + "outputs": [], + "source": [ + "# Shows the summary for the given URL\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a301ab4e", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://openai.com/\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "venv", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.9" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1.ipynb b/week1/community-contributions/day1.ipynb new file mode 100644 index 0000000..b876e38 --- /dev/null +++ b/week1/community-contributions/day1.ipynb @@ -0,0 +1,817 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# YOUR FIRST LAB\n", + "### Please read this section. This is valuable to get you prepared, even if it's a long read -- it's important stuff.\n", + "\n", + "## Your first Frontier LLM Project\n", + "\n", + "Let's build a useful LLM solution - in a matter of minutes.\n", + "\n", + "By the end of this course, you will have built an autonomous Agentic AI solution with 7 agents that collaborate to solve a business problem. All in good time! We will start with something smaller...\n", + "\n", + "Our goal is to code a new kind of Web Browser. Give it a URL, and it will respond with a summary. The Reader's Digest of the internet!!\n", + "\n", + "Before starting, you should have completed the setup for [PC](../SETUP-PC.md) or [Mac](../SETUP-mac.md) and you hopefully launched this jupyter lab from within the project root directory, with your environment activated.\n", + "\n", + "## If you're new to Jupyter Lab\n", + "\n", + "Welcome to the wonderful world of Data Science experimentation! Once you've used Jupyter Lab, you'll wonder how you ever lived without it. Simply click in each \"cell\" with code in it, such as the cell immediately below this text, and hit Shift+Return to execute that cell. As you wish, you can add a cell with the + button in the toolbar, and print values of variables, or try out variations. \n", + "\n", + "I've written a notebook called [Guide to Jupyter](Guide%20to%20Jupyter.ipynb) to help you get more familiar with Jupyter Labs, including adding Markdown comments, using `!` to run shell commands, and `tqdm` to show progress.\n", + "\n", + "## If you're new to the Command Line\n", + "\n", + "Please see these excellent guides: [Command line on PC](https://chatgpt.com/share/67b0acea-ba38-8012-9c34-7a2541052665) and [Command line on Mac](https://chatgpt.com/canvas/shared/67b0b10c93a081918210723867525d2b). \n", + "\n", + "## If you'd prefer to work in IDEs\n", + "\n", + "If you're more comfortable in IDEs like VSCode, Cursor or PyCharm, they both work great with these lab notebooks too. \n", + "If you'd prefer to work in VSCode, [here](https://chatgpt.com/share/676f2e19-c228-8012-9911-6ca42f8ed766) are instructions from an AI friend on how to configure it for the course.\n", + "\n", + "## If you'd like to brush up your Python\n", + "\n", + "I've added a notebook called [Intermediate Python](Intermediate%20Python.ipynb) to get you up to speed. But you should give it a miss if you already have a good idea what this code does: \n", + "`yield from {book.get(\"author\") for book in books if book.get(\"author\")}`\n", + "\n", + "## I am here to help\n", + "\n", + "If you have any problems at all, please do reach out. \n", + "I'm available through the platform, or at ed@edwarddonner.com, or at https://www.linkedin.com/in/eddonner/ if you'd like to connect (and I love connecting!) \n", + "And this is new to me, but I'm also trying out X/Twitter at [@edwarddonner](https://x.com/edwarddonner) - if you're on X, please show me how it's done 😂 \n", + "\n", + "## More troubleshooting\n", + "\n", + "Please see the [troubleshooting](troubleshooting.ipynb) notebook in this folder to diagnose and fix common problems. At the very end of it is a diagnostics script with some useful debug info.\n", + "\n", + "## For foundational technical knowledge (eg Git, APIs, debugging) \n", + "\n", + "If you're relatively new to programming -- I've got your back! While it's ideal to have some programming experience for this course, there's only one mandatory prerequisite: plenty of patience. 😁 I've put together a set of self-study guides that cover Git and GitHub, APIs and endpoints, beginner python and more.\n", + "\n", + "This covers Git and GitHub; what they are, the difference, and how to use them: \n", + "https://github.com/ed-donner/agents/blob/main/guides/03_git_and_github.ipynb\n", + "\n", + "This covers technical foundations: \n", + "ChatGPT vs API; taking screenshots; Environment Variables; Networking basics; APIs and endpoints: \n", + "https://github.com/ed-donner/agents/blob/main/guides/04_technical_foundations.ipynb\n", + "\n", + "This covers Python for beginners, and making sure that a `NameError` never trips you up: \n", + "https://github.com/ed-donner/agents/blob/main/guides/06_python_foundations.ipynb\n", + "\n", + "This covers the essential techniques for figuring out errors: \n", + "https://github.com/ed-donner/agents/blob/main/guides/08_debugging.ipynb\n", + "\n", + "And you'll find other useful guides in the same folder in GitHub. Some information applies to my other Udemy course (eg Async Python) but most of it is very relevant for LLM engineering.\n", + "\n", + "## If this is old hat!\n", + "\n", + "If you're already comfortable with today's material, please hang in there; you can move swiftly through the first few labs - we will get much more in depth as the weeks progress. Ultimately we will fine-tune our own LLM to compete with OpenAI!\n", + "\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Please read - important note

\n", + " The way I collaborate with you may be different to other courses you've taken. I prefer not to type code while you watch. Rather, I execute Jupyter Labs, like this, and give you an intuition for what's going on. My suggestion is that you carefully execute this yourself, after watching the lecture. Add print statements to understand what's going on, and then come up with your own variations. If you have a Github account, use this to showcase your variations. Not only is this essential practice, but it demonstrates your skills to others, including perhaps future clients or employers...\n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

This code is a live resource - keep an eye out for my emails

\n", + " I push updates to the code regularly. As people ask questions, I add more examples or improved commentary. As a result, you'll notice that the code below isn't identical to the videos. Everything from the videos is here; but I've also added better explanations and new models like DeepSeek. Consider this like an interactive book.

\n", + " I try to send emails regularly with important updates related to the course. You can find this in the 'Announcements' section of Udemy in the left sidebar. You can also choose to receive my emails via your Notification Settings in Udemy. I'm respectful of your inbox and always try to add value with my emails!\n", + " \n", + "
\n", + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Business value of these exercises

\n", + " A final thought. While I've designed these notebooks to be educational, I've also tried to make them enjoyable. We'll do fun things like have LLMs tell jokes and argue with each other. But fundamentally, my goal is to teach skills you can apply in business. I'll explain business implications as we go, and it's worth keeping this in mind: as you build experience with models and techniques, think of ways you could put this into action at work today. Please do contact me if you'd like to discuss more or if you have ideas to bounce off me.\n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "markdown", + "id": "6900b2a8-6384-4316-8aaa-5e519fca4254", + "metadata": {}, + "source": [ + "# Connecting to OpenAI (or Ollama)\n", + "\n", + "The next cell is where we load in the environment variables in your `.env` file and connect to OpenAI. \n", + "\n", + "If you'd like to use free Ollama instead, please see the README section \"Free Alternative to Paid APIs\", and if you're not sure how to do this, there's a full solution in the solutions folder (day1_with_ollama.ipynb).\n", + "\n", + "## Troubleshooting if you have problems:\n", + "\n", + "Head over to the [troubleshooting](troubleshooting.ipynb) notebook in this folder for step by step code to identify the root cause and fix it!\n", + "\n", + "If you make a change, try restarting the \"Kernel\" (the python process sitting behind this notebook) by Kernel menu >> Restart Kernel and Clear Outputs of All Cells. Then try this notebook again, starting at the top.\n", + "\n", + "Or, contact me! Message me or email ed@edwarddonner.com and we will get this to work.\n", + "\n", + "Any concerns about API costs? See my notes in the README - costs should be minimal, and you can control it at every point. You can also use Ollama as a free alternative, which we discuss during Day 2." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found - please head over to the troubleshooting notebook in this folder to identify & fix!\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key - see troubleshooting notebook\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them - see troubleshooting notebook\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n", + "# If it STILL doesn't work (horrors!) then please see the Troubleshooting notebook in this folder for full instructions" + ] + }, + { + "cell_type": "markdown", + "id": "442fc84b-0815-4f40-99ab-d9a5da6bda91", + "metadata": {}, + "source": [ + "# Let's make a quick call to a Frontier model to get started, as a preview!" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "a58394bf-1e45-46af-9bfd-01e24da6f49a", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Hello! It’s great to hear from you! How can I help you today?\n" + ] + } + ], + "source": [ + "# To give you a preview -- calling OpenAI with these messages is this easy. Any problems, head over to the Troubleshooting notebook.\n", + "\n", + "message = \"Hello, GPT! This is my first ever message to you! Hi!\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\":\"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "2aa190e5-cb31-456a-96cc-db109919cd78", + "metadata": {}, + "source": [ + "## OK onwards with our first project" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Home - Edward Donner\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Well, hi there.\n", + "I’m Ed. I like writing code and experimenting with LLMs, and hopefully you’re here because you do too. I also enjoy DJing (but I’m badly out of practice), amateur electronic music production (\n", + "very\n", + "amateur) and losing myself in\n", + "Hacker News\n", + ", nodding my head sagely to things I only half understand.\n", + "I’m the co-founder and CTO of\n", + "Nebula.io\n", + ". We’re applying AI to a field where it can make a massive, positive impact: helping people discover their potential and pursue their reason for being. Recruiters use our product today to source, understand, engage and manage talent. I’m previously the founder and CEO of AI startup untapt,\n", + "acquired in 2021\n", + ".\n", + "We work with groundbreaking, proprietary LLMs verticalized for talent, we’ve\n", + "patented\n", + "our matching model, and our award-winning platform has happy customers and tons of press coverage.\n", + "Connect\n", + "with me for more!\n", + "May 28, 2025\n", + "Connecting my courses – become an LLM expert and leader\n", + "May 18, 2025\n", + "2025 AI Executive Briefing\n", + "April 21, 2025\n", + "The Complete Agentic AI Engineering Course\n", + "January 23, 2025\n", + "LLM Workshop – Hands-on with Agents – resources\n", + "Navigation\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Get in touch\n", + "ed [at] edwarddonner [dot] com\n", + "www.edwarddonner.com\n", + "Follow me\n", + "LinkedIn\n", + "Twitter\n", + "Facebook\n", + "Subscribe to newsletter\n", + "Type your email…\n", + "Subscribe\n" + ] + } + ], + "source": [ + "# Let's try one out. Change the website and add print statements to follow along.\n", + "\n", + "ed = Website(\"https://edwarddonner.com\")\n", + "print(ed.title)\n", + "print(ed.text)" + ] + }, + { + "cell_type": "markdown", + "id": "6a478a0c-2c53-48ff-869c-4d08199931e1", + "metadata": {}, + "source": [ + "## Types of prompts\n", + "\n", + "You may know this already - but if not, you will get very familiar with it!\n", + "\n", + "Models like GPT4o have been trained to receive instructions in a particular way.\n", + "\n", + "They expect to receive:\n", + "\n", + "**A system prompt** that tells them what task they are performing and what tone they should use\n", + "\n", + "**A user prompt** -- the conversation starter that they should reply to" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# Define our system prompt - you can experiment with this later, changing the last sentence to 'Respond in markdown in Spanish.\"\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "f0275b1b-7cfe-4f9d-abfa-7650d378da0c", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "You are looking at a website titled Home - Edward Donner\n", + "The contents of this website is as follows; please provide a short summary of this website in markdown. If it includes news or announcements, then summarize these too.\n", + "\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Well, hi there.\n", + "I’m Ed. I like writing code and experimenting with LLMs, and hopefully you’re here because you do too. I also enjoy DJing (but I’m badly out of practice), amateur electronic music production (\n", + "very\n", + "amateur) and losing myself in\n", + "Hacker News\n", + ", nodding my head sagely to things I only half understand.\n", + "I’m the co-founder and CTO of\n", + "Nebula.io\n", + ". We’re applying AI to a field where it can make a massive, positive impact: helping people discover their potential and pursue their reason for being. Recruiters use our product today to source, understand, engage and manage talent. I’m previously the founder and CEO of AI startup untapt,\n", + "acquired in 2021\n", + ".\n", + "We work with groundbreaking, proprietary LLMs verticalized for talent, we’ve\n", + "patented\n", + "our matching model, and our award-winning platform has happy customers and tons of press coverage.\n", + "Connect\n", + "with me for more!\n", + "May 28, 2025\n", + "Connecting my courses – become an LLM expert and leader\n", + "May 18, 2025\n", + "2025 AI Executive Briefing\n", + "April 21, 2025\n", + "The Complete Agentic AI Engineering Course\n", + "January 23, 2025\n", + "LLM Workshop – Hands-on with Agents – resources\n", + "Navigation\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Get in touch\n", + "ed [at] edwarddonner [dot] com\n", + "www.edwarddonner.com\n", + "Follow me\n", + "LinkedIn\n", + "Twitter\n", + "Facebook\n", + "Subscribe to newsletter\n", + "Type your email…\n", + "Subscribe\n" + ] + } + ], + "source": [ + "print(user_prompt_for(ed))" + ] + }, + { + "cell_type": "markdown", + "id": "ea211b5f-28e1-4a86-8e52-c0b7677cadcc", + "metadata": {}, + "source": [ + "## Messages\n", + "\n", + "The API from OpenAI expects to receive messages in a particular structure.\n", + "Many of the other APIs share this structure:\n", + "\n", + "```python\n", + "[\n", + " {\"role\": \"system\", \"content\": \"system message goes here\"},\n", + " {\"role\": \"user\", \"content\": \"user message goes here\"}\n", + "]\n", + "```\n", + "To give you a preview, the next 2 cells make a rather simple call - we won't stretch the mighty GPT (yet!)" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "f25dcd35-0cd0-4235-9f64-ac37ed9eaaa5", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"system\", \"content\": \"You are a snarky assistant\"},\n", + " {\"role\": \"user\", \"content\": \"What is 2 + 2?\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "21ed95c5-7001-47de-a36d-1d6673b403ce", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Oh, you’re going for the big math questions now, huh? Well, if you insist on dragging me into elementary school territory, the answer is 4. Shocking, I know.\n" + ] + } + ], + "source": [ + "# To give you a preview -- calling OpenAI with system and user messages:\n", + "\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "d06e8d78-ce4c-4b05-aa8e-17050c82bb47", + "metadata": {}, + "source": [ + "## And now let's build useful messages for GPT-4o-mini, using a function" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "# See how this function creates exactly the format above\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "36478464-39ee-485c-9f3f-6a4e458dbc9c", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "[{'role': 'system',\n", + " 'content': 'You are an assistant that analyzes the contents of a website and provides a short summary, ignoring text that might be navigation related. Respond in markdown.'},\n", + " {'role': 'user',\n", + " 'content': 'You are looking at a website titled Home - Edward Donner\\nThe contents of this website is as follows; please provide a short summary of this website in markdown. If it includes news or announcements, then summarize these too.\\n\\nHome\\nConnect Four\\nOutsmart\\nAn arena that pits LLMs against each other in a battle of diplomacy and deviousness\\nAbout\\nPosts\\nWell, hi there.\\nI’m Ed. I like writing code and experimenting with LLMs, and hopefully you’re here because you do too. I also enjoy DJing (but I’m badly out of practice), amateur electronic music production (\\nvery\\namateur) and losing myself in\\nHacker News\\n, nodding my head sagely to things I only half understand.\\nI’m the co-founder and CTO of\\nNebula.io\\n. We’re applying AI to a field where it can make a massive, positive impact: helping people discover their potential and pursue their reason for being. Recruiters use our product today to source, understand, engage and manage talent. I’m previously the founder and CEO of AI startup untapt,\\nacquired in 2021\\n.\\nWe work with groundbreaking, proprietary LLMs verticalized for talent, we’ve\\npatented\\nour matching model, and our award-winning platform has happy customers and tons of press coverage.\\nConnect\\nwith me for more!\\nMay 28, 2025\\nConnecting my courses – become an LLM expert and leader\\nMay 18, 2025\\n2025 AI Executive Briefing\\nApril 21, 2025\\nThe Complete Agentic AI Engineering Course\\nJanuary 23, 2025\\nLLM Workshop – Hands-on with Agents – resources\\nNavigation\\nHome\\nConnect Four\\nOutsmart\\nAn arena that pits LLMs against each other in a battle of diplomacy and deviousness\\nAbout\\nPosts\\nGet in touch\\ned [at] edwarddonner [dot] com\\nwww.edwarddonner.com\\nFollow me\\nLinkedIn\\nTwitter\\nFacebook\\nSubscribe to newsletter\\nType your email…\\nSubscribe'}]" + ] + }, + "execution_count": 14, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# Try this out, and then try for a few more websites\n", + "\n", + "messages_for(ed)" + ] + }, + { + "cell_type": "markdown", + "id": "16f49d46-bf55-4c3e-928f-68fc0bf715b0", + "metadata": {}, + "source": [ + "## Time to bring it together - the API for OpenAI is very simple!" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "# And now: call the OpenAI API. You will get very familiar with this!\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'# Summary of Edward Donner\\'s Website\\n\\nThe website is the personal platform of Ed Donner, a software engineer and co-founder/CTO of Nebula.io, an AI-focused company that aims to help individuals discover their potential through technology. Ed expresses his passion for coding, experimenting with large language models (LLMs), and interests in DJing and electronic music production.\\n\\n## Key Sections:\\n- **About Ed**: Provides personal background, detailing his experience in AI startups, including his previous venture, untapt, which was acquired in 2021. He highlights the use of patented matching models and LLMs in talent management.\\n- **Connect Four & Outsmart**: Features interactive games or platforms where LLMs engage in diplomatic and strategic challenges.\\n- **Courses & Announcements**:\\n - **May 28, 2025**: Announced a course focused on becoming an LLM expert and leader.\\n - **May 18, 2025**: Announcement for the 2025 AI Executive Briefing.\\n - **April 21, 2025**: Introduction of \"The Complete Agentic AI Engineering Course.\"\\n - **January 23, 2025**: A workshop providing hands-on experience with agents and associated resources.\\n\\nThe website also encourages visitors to connect and engage through various social media platforms and a newsletter subscription.'" + ] + }, + "execution_count": 16, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "summarize(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "# Summary of \"Home - Edward Donner\"\n", + "\n", + "The website is dedicated to Edward Donner, a co-founder and CTO of Nebula.io, a platform focused on leveraging AI to assist individuals in discovering their potential and engaging with talent. Edward expresses an interest in coding, experiments with large language models (LLMs), DJing, and electronic music production. He has a history as the founder of an AI startup, untapt, which was acquired in 2021.\n", + "\n", + "## Key Features:\n", + "- **Connect Four**: A game involving LLMs competing in diplomacy and strategy.\n", + "- **About**: Information about Edward's professional background and interests.\n", + "- **Courses and Workshops**: \n", + " - **Recent Announcements**:\n", + " - **May 28, 2025**: Launch of a program to become an LLM expert and leader.\n", + " - **May 18, 2025**: Announcement of a 2025 AI Executive Briefing.\n", + " - **April 21, 2025**: Introduction of the Complete Agentic AI Engineering Course.\n", + " - **January 23, 2025**: A hands-on LLM Workshop focusing on resources related to agents.\n", + "\n", + "The content emphasizes his passion for AI and education within the industry." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "display_summary(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "markdown", + "id": "b3bcf6f4-adce-45e9-97ad-d9a5d7a3a624", + "metadata": {}, + "source": [ + "# Let's try more websites\n", + "\n", + "Note that this will only work on websites that can be scraped using this simplistic approach.\n", + "\n", + "Websites that are rendered with Javascript, like React apps, won't show up. See the community-contributions folder for a Selenium implementation that gets around this. You'll need to read up on installing Selenium (ask ChatGPT!)\n", + "\n", + "Also Websites protected with CloudFront (and similar) may give 403 errors - many thanks Andy J for pointing this out.\n", + "\n", + "But many websites will work just fine!" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "id": "45d83403-a24c-44b5-84ac-961449b4008f", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "# CNN Website Summary\n", + "\n", + "CNN is a major news outlet that provides comprehensive coverage of world events, featuring sections on U.S. and international news, political analysis, business, health, entertainment, and sports. The site emphasizes real-time updates and includes various multimedia formats such as videos and articles.\n", + "\n", + "### Notable Articles and Updates\n", + "\n", + "- **Ukraine-Russia War**: Pro-Ukraine protests are ongoing, and there are discussions regarding dignitaries meeting on U.S. soil amidst rising tensions.\n", + " \n", + "- **Israel-Hamas Conflict**: Analysis and reports highlight significant developments including Israel’s settlement plans which may impact the future of a Palestinian state.\n", + "\n", + "- **Health**: New heart health guidelines suggest going alcohol-free; studies indicate a high level of stress among teenagers.\n", + "\n", + "- **Entertainment**:\n", + " - Megadeth is set to release its final album and embark on a farewell tour.\n", + " - Taylor Swift's recent appearances are noted for cultural impact.\n", + "\n", + "- **Science**: Climate-related findings unveil vulnerabilities in GPS and satellites due to pollution.\n", + "\n", + "- **Business**: Discussions are ongoing about potential government stake in Intel, affecting stock prices.\n", + "\n", + "### Additional Features\n", + "CNN also offers a variety of interactive content including quizzes, games, and newsletters tailored to reader interests. The site encourages user engagement through feedback on advertisements and technical issues.\n", + "\n", + "Overall, CNN remains a significant source for breaking news and in-depth analysis across a broad spectrum of topics." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "75e9fd40-b354-4341-991e-863ef2e59db7", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://anthropic.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Subject: Leave Notification: Medical Emergency (Aug 17-21, 2025)\n" + ] + } + ], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"you are the email assistant, which provide the subject of the email\"\n", + "user_prompt = \"\"\"\n", + " please provide the appropriate subject for below email\n", + "hi team,\n", + "due to some medical emergency , i will be on leave for 5 days starting\n", + "from 17-08-2025 to 21-08-2025.\n", + "\n", + "please call me in case of any urgency.\n", + "\n", + "regards\n", + "Rahul\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "]\n", + "\n", + "# Step 3: Call OpenAI\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "\n", + "\n", + "# Step 4: print the result\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "36ed9f14-b349-40e9-a42c-b367e77f8bda", + "metadata": {}, + "source": [ + "## An extra exercise for those who enjoy web scraping\n", + "\n", + "You may notice that if you try `display_summary(\"https://openai.com\")` - it doesn't work! That's because OpenAI has a fancy website that uses Javascript. There are many ways around this that some of you might be familiar with. For example, Selenium is a hugely popular framework that runs a browser behind the scenes, renders the page, and allows you to query it. If you have experience with Selenium, Playwright or similar, then feel free to improve the Website class to use them. In the community-contributions folder, you'll find an example Selenium solution from a student (thank you!)" + ] + }, + { + "cell_type": "markdown", + "id": "eeab24dc-5f90-4570-b542-b0585aca3eb6", + "metadata": {}, + "source": [ + "# Sharing your code\n", + "\n", + "I'd love it if you share your code afterwards so I can share it with others! You'll notice that some students have already made changes (including a Selenium implementation) which you will find in the community-contributions folder. If you'd like add your changes to that folder, submit a Pull Request with your new versions in that folder and I'll merge your changes.\n", + "\n", + "If you're not an expert with git (and I am not!) then GPT has given some nice instructions on how to submit a Pull Request. It's a bit of an involved process, but once you've done it once it's pretty clear. As a pro-tip: it's best if you clear the outputs of your Jupyter notebooks (Edit >> Clean outputs of all cells, and then Save) for clean notebooks.\n", + "\n", + "Here are good instructions courtesy of an AI friend: \n", + "https://chatgpt.com/share/677a9cb5-c64c-8012-99e0-e06e88afd293" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4484fcf-8b39-4c3f-9674-37970ed71988", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1_coverletter_tailored_to_CV_and_job_description.ipynb b/week1/community-contributions/day1_coverletter_tailored_to_CV_and_job_description.ipynb new file mode 100644 index 0000000..9c63b6a --- /dev/null +++ b/week1/community-contributions/day1_coverletter_tailored_to_CV_and_job_description.ipynb @@ -0,0 +1,211 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "d955d75d-4970-48fe-983e-a2a850cecfc5", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "import PyPDF2\n", + "from selenium import webdriver\n", + "from selenium.webdriver.chrome.options import Options\n", + "from selenium.webdriver.chrome.service import Service\n", + "from webdriver_manager.chrome import ChromeDriverManager\n", + "from bs4 import BeautifulSoup\n", + "import time" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6e1e5dd3-f91a-466b-8fd4-2dbf4eedf101", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override = True)\n", + "api_key = os.getenv(\"OPENAI_API_KEY\")\n", + "\n", + "if not api_key:\n", + " print(\"No API key\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"API key doesn't look correct, check it\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"It looks like API key has an extra space - check it\")\n", + "else:\n", + " print(\"API key looks good, moving on!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "67a6e583-1ef7-4b77-8886-c0e8c619933c", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "34a07806-dd68-4a86-8b6e-e1b2aaf0daa1", + "metadata": {}, + "outputs": [], + "source": [ + "# path to the CV\n", + "path = \"/Users/yanasklar/Documents/For applying/CV/СV_YanaSklyar_c.pdf\"\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/139.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Vacancy:\n", + " def __init__(self, url, instructions = \"\"):\n", + " self.url = url\n", + " \n", + " # configure Chrome settings\n", + " options = Options()\n", + " # options.add_argument(\"--headless\") \n", + " \"\"\"\n", + " Headless mode runs the browser in the background (invisible).\n", + " However, some websites (like openai.com) block headless browsers.\n", + " So if this line is active, the page may not load correctly and you may not get the full content.\n", + " \"\"\"\n", + " options.add_argument(\"--disable-gpu\")\n", + " options.add_argument(\"--no-sandbox\")\n", + " options.add_argument(\"--window-size=1920x1080\")\n", + "\n", + " # use webdriver-manager to manage ChromeDriver\n", + " service = Service(ChromeDriverManager().install())\n", + " driver = webdriver.Chrome(service=service, options=options)\n", + " driver.get(url)\n", + " time.sleep(3) # let the page load\n", + "\n", + " # take the source of the page\n", + " page_source = driver.page_source\n", + " driver.quit()\n", + "\n", + " # analyse with BeautifulSoup\n", + " soup = BeautifulSoup(page_source, 'html.parser')\n", + "\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"img\", \"script\", \"style\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator='\\n', strip=True)\n", + "\n", + " # read CV\n", + " with open(path, 'rb') as f:\n", + " reader = PyPDF2.PdfReader(f)\n", + " cv_text = \"\"\n", + " for page in reader.pages:\n", + " text = page.extract_text()\n", + " if text:\n", + " cv_text += text + \"\\n\"\n", + " self.cv_text = cv_text\n", + "\n", + " # summarise and print the description of the job\n", + " message = f\"\"\"Here is the content of a webpage: {self.text}.\n", + " Find job description on that page,\n", + " summarise it, include the list requirements and other important details.\n", + " \"\"\"\n", + " messages = [{\"role\":\"user\", \"content\":message}]\n", + " response = openai.chat.completions.create(model='gpt-4o-mini', messages = messages)\n", + " print(\"The job description: \", response.choices[0].message.content)\n", + "\n", + " # create prompts\n", + " self.system_prompt = \"\"\"You are a career assistant specializing in writing cover letter.\n", + " Your tasks:\n", + " 1. Read the candidate's CV (provided as text).\n", + " 2. Read the job description (provided from a webpage).\n", + " 3. Write a concise and compelling cover letter, that:\n", + " - Hightlights the most relevant experience and skills from the CV,\n", + " - Aligns directly wit the requirements in the job description,\n", + " - Adapts to cultural and professional norms in Israel.\n", + " The letter should be no longer than half a page, persuasive and tailored to make the applicant stand out.\n", + " \"\"\"\n", + "\n", + " user_prompt = f\"\"\"\n", + " Here is my CV:\n", + " {self.cv_text}\n", + " \n", + " The job vacancy is from the website {self.title}.\n", + " Here is the decription of the vacancy:\n", + " {self.text}\n", + " Please write a cover letter that connects my background to this vacancy.\n", + " Make it persuasive and suitable for Israeli job market.\n", + " \"\"\"\n", + " \n", + " if instructions:\n", + " user_prompt += f\"Additional instructions: {instructions}\"\n", + " self.user_prompt = user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9160b9f5-177b-4477-8e54-3a212f275a22", + "metadata": {}, + "outputs": [], + "source": [ + "def cover_letter(url, instructions = \"\"):\n", + " vacancy = Vacancy(url, instructions)\n", + " messages = [\n", + " {\"role\":\"system\", \"content\":vacancy.system_prompt},\n", + " {\"role\":\"user\", \"content\":vacancy.user_prompt}\n", + " ]\n", + " response = openai.chat.completions.create(model='gpt-4o-mini', messages=messages)\n", + " if not response:\n", + " print(\"smt went wrong\")\n", + " print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1de4b55c-a8da-445f-9865-c7a8bafdbc3c", + "metadata": {}, + "outputs": [], + "source": [ + "a = \"https://www.linkedin.com/jobs/view/4285898438/?alternateChannel=search&eBP=CwEAAAGY3R5LOabDLOVTy6xvBcSlWyAkIXQz8IRkSM3rgsqTPtvcEvUSnq980O7oLV2Hh_ldTpc2cBBmRq1IRnLtp7TzEcUvndFEXeCuviA5yo7oFYfW7KoEp4SPNzmf3D9LtnSgk9Iudy3skk6n3hVOtyDpx8Zm0AiTWPvdwCaZ_w5Xu8lAG797NRNDco71ynm99LmCOC9Go7DdDQ2eLewamc4SOsA4xWcXy0GmZVy3kBF1AprK3ylAYR2wrm5-hp4lRpbbfUxXjkEOG6H_GbPpKtN-N8mYnMd9w_cej5qQmTFX86gqSi6HuXFtK0h46TbOS5r-YQksVd1Yb4kYZnDznWXPLbxp04xVJSPzsHoa05wQdOfZ2UUSoMTJmic3n3qfV2u9Bp8n4sLYtINpzKdvm4eADGGkN-nR3O2oPeas9XjGbBwNdjXHAcX_PJoRwlFdQ1gVkYQEF1T7qAfXUJoUt-fv4oLxGnIgV6yJuMgw&refId=9NA7Bvt%2FhCqDkFNRGu1dPA%3D%3D&trackingId=W11hvpcIjHA%2FjU%2FFZ%2B1uAA%3D%3D\"\n", + "b = \"The style of the cover letter should informal, as if i talked to a friend about my background\"\n", + "cover_letter(a, b)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0feb3cbe-686a-4a97-9ca3-a0cb32a24c5d", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python (llms)", + "language": "python", + "name": "llms" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1_exercise-recipe_formatter.ipynb b/week1/community-contributions/day1_exercise-recipe_formatter.ipynb new file mode 100644 index 0000000..df936bf --- /dev/null +++ b/week1/community-contributions/day1_exercise-recipe_formatter.ipynb @@ -0,0 +1,239 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "cab13efd-a1f4-4077-976e-e3912511117f", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import re\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "c226f54b-325c-49b1-9d99-207a8e306682", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Requirement already satisfied: youtube_transcript_api in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (1.1.1)\n", + "Requirement already satisfied: defusedxml<0.8.0,>=0.7.1 in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from youtube_transcript_api) (0.7.1)\n", + "Requirement already satisfied: requests in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from youtube_transcript_api) (2.32.4)\n", + "Requirement already satisfied: charset_normalizer<4,>=2 in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from requests->youtube_transcript_api) (3.4.2)\n", + "Requirement already satisfied: idna<4,>=2.5 in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from requests->youtube_transcript_api) (3.10)\n", + "Requirement already satisfied: urllib3<3,>=1.21.1 in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from requests->youtube_transcript_api) (2.5.0)\n", + "Requirement already satisfied: certifi>=2017.4.17 in /Users/nachogonzalezbullon/miniconda3/envs/llms/lib/python3.11/site-packages (from requests->youtube_transcript_api) (2025.7.9)\n" + ] + } + ], + "source": [ + "!pip install youtube_transcript_api" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "717fc2a4-b6c5-4027-9e6b-05e83c38d02f", + "metadata": {}, + "outputs": [], + "source": [ + "from youtube_transcript_api import YouTubeTranscriptApi" + ] + }, + { + "metadata": {}, + "cell_type": "code", + "outputs": [], + "execution_count": 4, + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')" + ], + "id": "3caca469-5f39-4592-bf12-c8832c44de19" + }, + { + "metadata": {}, + "cell_type": "code", + "outputs": [], + "execution_count": null, + "source": [ + "class YouTubeRecipeExtractor:\n", + "\n", + " def __init__(self):\n", + " self.openai = OpenAI()\n", + " self.system_prompt = self.get_system_prompt()\n", + "\n", + " def get_system_prompt(self):\n", + " return \"\"\"\n", + " You are a professional chef and nutritionist specializing in recipe writting.\n", + "\n", + " Your task is to write recipes in a very comprehensive and consistent manner.\n", + " Each recipe will contain a list of ingredients and a list of steps to follow.\n", + " The quantities of the ingredients should always be referred to an official unit (grams, litres, etc). If the original recipe uses a different unit (such as cup, teaspoons, etc.) make the transformation but keep the original instruction between parenthesis.\n", + " The steps should be described in a very synthetic and concise manner. You should avoid being verbose, but the step should be understandable and easy to follow for non-expert people.\n", + " To each recipe add a general analysis from nutrition perspective (number of calories per serving, proteins, fat, etc.).\n", + " Use Markdown to improve readability.\n", + " If the text you receive is not a recipe, return a kind message explaining the situation.\n", + " \"\"\"\n", + "\n", + " def extract_video_id(self, url):\n", + " \"\"\"Extract video ID from YouTube URL\"\"\"\n", + " pattern = r'(?:youtube\\.com/watch\\?v=|youtu\\.be/|youtube\\.com/embed/)([^&\\n?#]+)'\n", + " match = re.search(pattern, url)\n", + " return match.group(1) if match else None\n", + "\n", + " def get_transcription(self, video_id):\n", + " try:\n", + " print(f\"Fetching video transcript for video {video_id}...\")\n", + " transcript = YouTubeTranscriptApi.get_transcript(video_id)\n", + " return \" \".join([item['text'] for item in transcript])\n", + " except Exception as e:\n", + " print(f\"Error fetching transcript: {e}\")\n", + " return None\n", + "\n", + " def format_recipe(self, transcript):\n", + " try:\n", + " response = self.openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": self.system_prompt},\n", + " {\"role\": \"user\", \"content\": f\"Summarize the following YouTube recipe:\\n\\n{transcript}\"}\n", + " ]\n", + " )\n", + " return response.choices[0].message.content\n", + " except Exception as e:\n", + " print(f\"Error summarizing text: {e}\")\n", + " return None\n", + "\n", + " def display_recipe(self, url):\n", + " transcript = self.get_transcription(self.extract_video_id(url))\n", + " recipe = self.format_recipe(transcript)\n", + " display(Markdown(recipe))\n" + ], + "id": "29e44cb5-0928-4ac9-9681-efd6ba1e359f" + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "98ea2d01-f949-4e03-9154-fe524cf64ca4", + "metadata": {}, + "outputs": [], + "source": [ + "test_bad_url = \"https://www.youtube.com/watch?v=hzGiTUTi060\"\n", + "test_good_url = \"https://www.youtube.com/watch?v=D_2DBLAt57c\"" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "362e39e8-a254-4f2f-8653-5fbb7ff0e1e9", + "metadata": {}, + "outputs": [], + "source": [ + "extractor = YouTubeRecipeExtractor()\n" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "0cc259bd-46bb-4472-b3cb-f39da54e324a", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fetching video transcript...\n" + ] + }, + { + "data": { + "text/markdown": [ + "Thank you for your interest, but the text you provided is not a recipe. If you're looking for cooking instructions, ingredient lists, or nutrition analysis, please provide a specific food or dish you would like to know about, and I'd be happy to help!" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "extractor.display_recipe(test_bad_url)" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "3f43e320-ca55-4db5-bc95-71fcb342cf3c", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Fetching video transcript for video D_2DBLAt57c...\n", + "Error fetching transcript: YouTubeTranscriptApi.fetch() missing 1 required positional argument: 'self'\n" + ] + }, + { + "data": { + "text/markdown": [ + "It seems like you haven't provided a recipe or any details to summarize. If you have a specific recipe in mind, please share it, and I'll be happy to help!" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "extractor.display_recipe(test_good_url)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "11c5c2b3-498a-43eb-9b68-d2b920c56b10", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1_resume_to_job_gap_analysis_tool.ipynb b/week1/community-contributions/day1_resume_to_job_gap_analysis_tool.ipynb new file mode 100644 index 0000000..73ba675 --- /dev/null +++ b/week1/community-contributions/day1_resume_to_job_gap_analysis_tool.ipynb @@ -0,0 +1,340 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "ddc7185d", + "metadata": {}, + "source": [ + "# Resume-to-Job Gap Analysis Tool" + ] + }, + { + "cell_type": "markdown", + "id": "fe7462c2", + "metadata": {}, + "source": [ + "### **Project Summary**\n", + "This project demonstrates the use of a Large Language Model (LLM) to perform a sophisticated analysis task with real-world business value. The tool automates the tedious process of manually comparing a candidate's resume against a job description. By providing a job description URL and a candidate's resume text, this notebook generates a detailed cover letter and \"gap analysis\" report. This report highlights which skills are matched, which are missing, and provides an overall suitability score, enabling recruiters to screen candidates more efficiently and helping applicants tailor their resumes effectively.\n", + "\n", + "### **How to Use**\n", + "1. **Set up your Environment**: Make sure you have a `.env` file in the root directory with your `OPENAI_API_KEY`.\n", + "2. **Input the Job URL**: In **Section 2**, paste the URL of a web-based job description into the `job_description_url` variable.\n", + "3. **Input the Resume**: In **Section 2**, paste the candidate's full resume text into the `resume_text` variable.\n", + "4. **Run the Notebook**: Execute the cells from top to bottom. The final cell in **Section 6** will display the formatted analysis report.\n", + "\n", + "### **A Note on Ethical Web Scraping**\n", + "This tool uses the `requests` library to fetch website content. To ensure compliance and responsible use:\n", + "* We send a standard `User-Agent` header to identify our script as a web browser, which is a common practice for preventing being blocked.\n", + "* **Always be mindful of the website's terms of service.** Automated scraping may be disallowed on some sites. This tool is intended for educational purposes and should be used on publicly accessible job postings where such activity is permitted." + ] + }, + { + "cell_type": "markdown", + "id": "1a01b5d2", + "metadata": {}, + "source": [ + "## 1. Setup:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "caca8d9a", + "metadata": {}, + "outputs": [], + "source": [ + "# Imports\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e2db03e8", + "metadata": {}, + "outputs": [], + "source": [ + "# Load Environment Variables\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')" + ] + }, + { + "cell_type": "markdown", + "id": "7c702fcc", + "metadata": {}, + "source": [ + "#### Test OpenAI API Key" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5347ee38", + "metadata": {}, + "outputs": [], + "source": [ + "# Validate API key\n", + "if not api_key:\n", + " print(\"ERROR: No API key found - please add OPENAI_API_KEY to your .env file\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"WARNING: API key format may be incorrect\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"ERROR: API key has whitespace - please remove extra spaces/tabs\")\n", + "else:\n", + " print(\"SUCCESS: API key loaded successfully\")\n", + "\n", + "# Initialize OpenAI client\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "markdown", + "id": "dce21512", + "metadata": {}, + "source": [ + "## 2. Data Input" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b5d90d56", + "metadata": {}, + "outputs": [], + "source": [ + "# The URL for the Y Combinator job posting you want to analyze. (ycombinator.com/companies/y-combinator/jobs//jobs)\n", + "job_url = \"https://www.ycombinator.com/companies/y-combinator/jobs/rq3DaTs-product-engineer\"\n", + "\n", + "# Replace this example resume with the actual candidate's resume text.\n", + "resume_text = \"\"\"\n", + "John Doe\n", + "123 Main Street, Anytown, USA | (123) 456-7890 | john.doe@email.com\n", + "\n", + "Summary\n", + "Software Engineer with 5 years of experience in web applications. \n", + "Proficient in Python and JavaScript with a strong background in AWS.\n", + "\n", + "Experience\n", + "Senior Software Engineer | Tech Solutions Inc. | 2021 - Present\n", + "- Led development of analytics dashboard using React and Python\n", + "- Architected microservices backend on AWS\n", + "- Mentored junior engineers\n", + "\n", + "Software Engineer | Innovate Corp. | 2018 - 2021\n", + "- Developed e-commerce platform using Python and Django\n", + "- Wrote comprehensive unit and integration tests\n", + "\n", + "Skills\n", + "Python, JavaScript, React, Flask, Django, AWS, Docker, Git\n", + "\"\"\"\n", + "\n" + ] + }, + { + "cell_type": "markdown", + "id": "a3d5e484", + "metadata": {}, + "source": [ + "## 3. Prompt Engineering" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6b2b3d1b", + "metadata": {}, + "outputs": [], + "source": [ + "SYSTEM_PROMPT = \"\"\"\n", + "You are a strategic career advisor. Your task is to synthesize a candidate's resume and a job description into a compelling, two-part analysis. Your goal is to create a narrative connecting the candidate's specific accomplishments to the company's needs.\n", + "\n", + "**Formatting:** Use markdown with bolding for emphasis. Do not use placeholders like '[Job Title]'; infer the details from the text.\n", + "\n", + "---\n", + "\n", + "# Part 1: Candidate Suitability Analysis\n", + "\n", + "## Executive Summary\n", + "Provide a 2-3 sentence summary of the candidate's alignment with the role, stating your professional opinion on their potential.\n", + "\n", + "## Key Strengths & Evidence\n", + "List the top 3 strengths the candidate brings. For each strength, **quote or paraphrase evidence directly from the resume's 'Experience' section**.\n", + "* **Strength:** [Example: Scalable Backend Development] - **Evidence:** [Example: \"Architected microservices backend on AWS,\" demonstrating hands-on experience.]\n", + "\n", + "## Areas for Growth & Discussion\n", + "Identify key requirements from the job description not explicitly covered in the resume. Frame these as **strategic points to address in an interview**.\n", + "* **Topic:** [Example: TypeScript Proficiency] - **Suggested Question:** \"The role heavily uses TypeScript. Could you discuss your experience level with it and your approach to learning new languages?\"\n", + "\n", + "## Holistic Suitability Score\n", + "Provide a score (e.g., 85/100) and justify it in one sentence.\n", + "\n", + "---\n", + "\n", + "# Part 2: Dynamic Cover Letter Draft\n", + "Generate a compelling and authentic cover letter from the candidate's perspective.\n", + "\"\"\"" + ] + }, + { + "cell_type": "markdown", + "id": "5146a406", + "metadata": {}, + "source": [ + "## 4. Webscraper" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4d23965d", + "metadata": {}, + "outputs": [], + "source": [ + "# Scraper Function\n", + "def scrape_ycombinator_job(url: str) -> str:\n", + " \"\"\"\n", + " Scrapes a single job posting from a ycombinator.com URL.\n", + "\n", + " Args:\n", + " url: The URL of the Y Combinator job posting.\n", + "\n", + " Returns:\n", + " The cleaned text of the job description, or an error message.\n", + " \"\"\"\n", + " print(f\"INFO: Attempting to scrape YC job posting from: {url}\")\n", + " \n", + " headers = {\n", + " 'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36'\n", + " }\n", + " \n", + " try:\n", + " # Fetch the page content\n", + " response = requests.get(url, headers=headers, timeout=10)\n", + " # Raise an error if the page is not found (e.g., 404)\n", + " response.raise_for_status()\n", + "\n", + " # Parse the HTML with BeautifulSoup\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + "\n", + " # Extract the job title (specifically from the

tag)\n", + " title_element = soup.select_one('h1')\n", + " title = title_element.get_text(strip=True) if title_element else \"Job Title Not Found\"\n", + "\n", + " # Extract the main job description content (from the
)\n", + " description_element = soup.select_one('.prose')\n", + " description = description_element.get_text(separator='\\n', strip=True) if description_element else \"\"\n", + " \n", + " # Combine them for the final text\n", + " full_text = f\"Job Title: {title}\\n\\n{description}\"\n", + " \n", + " print(\"SUCCESS: Scraping complete.\")\n", + " return full_text\n", + "\n", + " except requests.exceptions.RequestException as e:\n", + " print(f\"ERROR: Scraping failed. Could not fetch URL. {e}\")\n", + " return \"[Scraping failed: Could not connect to the server]\"\n", + " except Exception as e:\n", + " print(f\"ERROR: An unexpected error occurred during scraping: {e}\")\n", + " return \"[Scraping failed: An unexpected error occurred]\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "e159596d", + "metadata": {}, + "source": [ + "## 5. Gap Analysis" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d0dc8f72", + "metadata": {}, + "outputs": [], + "source": [ + "def get_analysis(job_description: str, resume: str) -> str:\n", + " \"\"\"Sends the job description and resume to the AI and returns the analysis.\"\"\"\n", + " print(\"INFO: Sending data to the AI for analysis...\")\n", + " user_prompt = f\"\"\"Please generate the analysis based on the following documents.\n", + "\n", + " **JOB DESCRIPTION:**\n", + " ---\n", + " {job_description}\n", + " ---\n", + "\n", + " **CANDIDATE RESUME:**\n", + " ---\n", + " {resume}\n", + " ---\n", + " \"\"\"\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": SYSTEM_PROMPT},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=messages\n", + " )\n", + " print(\"SUCCESS: Analysis complete.\")\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "markdown", + "id": "f1deb906", + "metadata": {}, + "source": [ + "## 6. Execution" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3e57129", + "metadata": {}, + "outputs": [], + "source": [ + "# Scrape the job description text from the URL\n", + "job_description_text = scrape_ycombinator_job(job_url)\n", + "\n", + "# Only proceed if scraping was successful\n", + "if not job_description_text.startswith(\"[Scraping failed\"):\n", + " # Run the analysis with the scraped text\n", + " analysis_report = get_analysis(job_description_text, resume_text)\n", + " # Display the final report\n", + " display(Markdown(analysis_report))\n", + "else:\n", + " # If scraping failed, display the error message\n", + " display(Markdown(f\"## {job_description_text}\"))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1_selenium_microsoftedgedriver.ipynb b/week1/community-contributions/day1_selenium_microsoftedgedriver.ipynb new file mode 100644 index 0000000..c2e6b86 --- /dev/null +++ b/week1/community-contributions/day1_selenium_microsoftedgedriver.ipynb @@ -0,0 +1,123 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "3ba06289-d17a-4ccd-85f5-2b79956d4e59", + "metadata": {}, + "outputs": [], + "source": [ + "!pip install selenium" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "935fe7b1-1807-4f75-863d-4c118e425a19", + "metadata": {}, + "outputs": [], + "source": [ + "pip show selenium" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eabbbc62-1de1-4883-9b3e-9c90145ea6c5", + "metadata": {}, + "outputs": [], + "source": [ + "from selenium import webdriver\n", + "from selenium.webdriver.edge.options import Options as EdgeOptions # Import EdgeOptions\n", + "from selenium.webdriver.edge.service import Service as EdgeService # Import EdgeService\n", + "from bs4 import BeautifulSoup\n", + "import time\n", + "import os\n", + "\n", + "class Website:\n", + " def __init__(self, url, driver_path=None, wait_time=3):\n", + " self.url = url\n", + " self.wait_time = wait_time\n", + "\n", + " # Headless Edge settings\n", + " options = EdgeOptions() # Use EdgeOptions\n", + " # options.add_argument(\"--headless\")\n", + " options.add_argument(\"--disable-gpu\")\n", + " options.add_argument(\"--no-sandbox\")\n", + " options.add_argument(\"--window-size=1920x1080\")\n", + "\n", + " # Driver path\n", + " if driver_path:\n", + " # For Edge, you might need to specify the path to msedgedriver\n", + " # For driver download, https://developer.microsoft.com/en-us/microsoft-edge/tools/webdriver/?form=MA13LH#downloads\n", + " service = EdgeService(executable_path=driver_path) # Use EdgeService\n", + " else:\n", + " # If msedgedriver.exe is in your system's PATH, you can omit executable_path\n", + " service = EdgeService()\n", + "\n", + " # Start browser\n", + " # Use webdriver.Edge() for Microsoft Edge\n", + " driver = webdriver.Edge(service=service, options=options)\n", + " driver.get(url)\n", + "\n", + " # Wait for the loading page\n", + " time.sleep(self.wait_time)\n", + "\n", + " # Take page source\n", + " html = driver.page_source\n", + " driver.quit()\n", + "\n", + " # Analysis with BeautifulSoup \n", + " soup = BeautifulSoup(html, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + "\n", + " # Clean irrelevant tags\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + "\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "852c52e2-bd4d-4bb9-94ef-e498c33f1a89", + "metadata": {}, + "outputs": [], + "source": [ + "site = Website(\"https://openai.com\", driver_path=\"/Users/klee/Documents/edgedriver_mac64_m1/msedgedriver\")\n", + "print(\"Title:\", site.title)\n", + "print(\"\\nFirst 500 character:\\n\", site.text[:500])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7620c685-c35c-4d6b-aaf1-a3da98f19ca7", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day1_tennis.ipynb b/week1/community-contributions/day1_tennis.ipynb new file mode 100644 index 0000000..10c2f80 --- /dev/null +++ b/week1/community-contributions/day1_tennis.ipynb @@ -0,0 +1,129 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "3bf6bba3-cea5-4e28-8e57-bddef9c80013", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "62f20a01-2d4f-45ac-a890-fce46d552301", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if not api_key:\n", + " print('No Api Key was found')\n", + "elif not api_key.startswith('sk-proj-'):\n", + " print(\"An api key was found, but it doesnt start with sk-proj\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An api key was found, but it might have space in the first or end\")\n", + "else:\n", + " print(\"Api key found and looks good so far!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f9a1fea0-f228-4310-8c0c-2074cd09ab53", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9db0ddcd-befd-445b-817b-d30c50de9206", + "metadata": {}, + "outputs": [], + "source": [ + "message = \"Hello GPT, this is my first message\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\": \"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"You are a personal tennis coach who helps children learn how to play. \\\n", + "Write a short summary of advice for a child who is just starting to get interested in tennis. \\\n", + "Respond in Markdown\"\n", + "user_prompt = \"\"\"\n", + " Even if you’ve both geared up and warmed up, you still need to know all the tennis basics before you step onto the court and play. You can use any combination of tennis grips (like the Semi-Western or Continental grip) and moves like drop shots, lobs, backhand volleys, or forehand strokes to try and win each point. However, learning all the fundamentals of tennis is imperative to play your best tennis:\n", + "Keep it inside the lines. For singles tennis, the serve must always land over the net, and within the opponent’s opposite service box (the box on either side of the center mark at the service line, also known as the “T”). If the ball hits the net and still lands in the proper service box, it’s called a “let,” and the server gets to start over from the first serve again. Even if the ball technically lands outside the box, as long as any part of it still touches the line, it is still in-play. During a rally, the ball must stay within the singles court boundaries, which are the inner sidelines. For doubles tennis, the outer alleys are in-play. However, most beginner players won’t have a line judge present, so they must call the ball out or raise their finger if the ball lands outside the lines.\n", + "Keep score. Tennis has a unique scoring system, and it’s important to keep track of your points to determine who will win (and which side you should be serving from). The server always says their score first, even if it is lower than their opponent’s. For example, if the server loses the first three points in a row, the score is love-40.\n", + "Avoid touching the net. You can rush the net and perform any volleying maneuver you like. However, if any part of you or your racket physically touches the net at any time during a point, you automatically lose. The net is the equal divider between both sides, and any alteration to its positioning, even accidental, is not allowed.\n", + "Hold onto your racket. Your racket must stay in your hands at all times. If you drop or throw the racket at the ball, you will lose the point. You can only return the ball with your racket and no other part of your body. However, the ball doesn’t necessarily have to touch the racket face—it’s still in-play even if it hits the handle or triangle as well.\n", + "Hit the ball after one bounce. Once the ball bounces twice, the point is over. Similarly, you can only hit the ball once as well. Even if you clip the ball and it lands in front of you again, the point is over if the ball doesn’t reach your opponent’s side.\n", + "A ball in the air is a ball in play. Even if your opponent is well behind the baseline in “out” territory, if they make contact with the ball or it hits a part of their body before the bounce, it’s still in-play. A ball can’t be called until it bounces.\n", + "Win by two. Both games and points must be won by two in a tennis match. In the event of a tie, where both players each win six games in a set resulting in a score of 6-6, a tiebreak is introduced. This is where players must face off in a seven-point mini-match. The players switch sides after each serve point, and the end of the court when the sum of the points equals six or multiples thereof. The first player to reach seven points (leading by two) wins. If the tiebreaker occurs in the last set, the points are instead played first to 10, and the winning player must still win by two points.\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "] # fill this in\n", + "\n", + "# Step 3: Call OpenAI\n", + "\n", + "response =openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=messages\n", + ")\n", + "\n", + "# Step 4: print the result\n", + "\n", + "print(response.choices[0].message.content)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day2 EXERCISE_priithvi.ipynb b/week1/community-contributions/day2 EXERCISE_priithvi.ipynb new file mode 100644 index 0000000..3542cb2 --- /dev/null +++ b/week1/community-contributions/day2 EXERCISE_priithvi.ipynb @@ -0,0 +1,1029 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# Welcome to your first assignment!\n", + "\n", + "Instructions are below. Please give this a try, and look in the solutions folder if you get stuck (or feel free to ask me!)" + ] + }, + { + "cell_type": "markdown", + "id": "ada885d9-4d42-4d9b-97f0-74fbbbfe93a9", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Just before we get to the assignment --

\n", + " I thought I'd take a second to point you at this page of useful resources for the course. This includes links to all the slides.
\n", + " https://edwarddonner.com/2024/11/13/llm-engineering-resources/
\n", + " Please keep this bookmarked, and I'll continue to add more useful links there over time.\n", + " \n", + "
" + ] + }, + { + "cell_type": "markdown", + "id": "6e9fa1fc-eac5-4d1d-9be4-541b3f2b3458", + "metadata": {}, + "source": [ + "# HOMEWORK EXERCISE ASSIGNMENT\n", + "\n", + "Upgrade the day 1 project to summarize a webpage to use an Open Source model running locally via Ollama rather than OpenAI\n", + "\n", + "You'll be able to use this technique for all subsequent projects if you'd prefer not to use paid APIs.\n", + "\n", + "**Benefits:**\n", + "1. No API charges - open-source\n", + "2. Data doesn't leave your box\n", + "\n", + "**Disadvantages:**\n", + "1. Significantly less power than Frontier Model\n", + "\n", + "## Recap on installation of Ollama\n", + "\n", + "Simply visit [ollama.com](https://ollama.com) and install!\n", + "\n", + "Once complete, the ollama server should already be running locally. \n", + "If you visit: \n", + "[http://localhost:11434/](http://localhost:11434/)\n", + "\n", + "You should see the message `Ollama is running`. \n", + "\n", + "If not, bring up a new Terminal (Mac) or Powershell (Windows) and enter `ollama serve` \n", + "And in another Terminal (Mac) or Powershell (Windows), enter `ollama pull llama3.2` \n", + "Then try [http://localhost:11434/](http://localhost:11434/) again.\n", + "\n", + "If Ollama is slow on your machine, try using `llama3.2:1b` as an alternative. Run `ollama pull llama3.2:1b` from a Terminal or Powershell, and change the code below from `MODEL = \"llama3.2\"` to `MODEL = \"llama3.2:1b\"`" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"\n", + "MODEL = \"tinyllama:latest\"" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "id": "dac0a679-599c-441f-9bf2-ddc73d35b940", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a messages list using the same format that we used for OpenAI\n", + "\n", + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Summarize this website: cnn.com\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "id": "7bb9c624-14f0-4945-a719-8ddb64f66f47", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "id": "479ff514-e8bd-4985-a572-2ea28bb4fa40", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's just make sure the model is loaded\n", + "\n", + "# !ollama pull llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "id": "42b9f644-522d-4e05-a691-56e7658c0ea9", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "This website provides up-to-date and comprehensive news, analysis, and opinion articles from CNN on a variety of topics, including politics, business, entertainment, sports, and international affairs. It offers a personalized feed based on your interests and browsing history to provide you with relevant content tailored to your preferences.\n" + ] + } + ], + "source": [ + "# If this doesn't work for any reason, try the 2 versions in the following cells\n", + "# And double check the instructions in the 'Recap on installation of Ollama' at the top of this lab\n", + "# And if none of that works - contact me!\n", + "\n", + "response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d042059-333e-4723-a48c-8a1a71fd6aab", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 23, + "id": "99d0c1a2-52b2-4cb3-9d67-6d5931847f8c", + "metadata": {}, + "outputs": [], + "source": [ + "response = requests.post(OLLAMA_API, json = payload, headers = HEADERS)" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "e8bd28b9-545b-4806-8a25-93b0208b7939", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The capital of France is Paris, and the current name was adopted in 1968. The previous names include:\n", + "\n", + "1. Paris (1947-1968)\n", + "2. Ville de Paris (1802-1803)\n", + "3. Ville d'Ay (1799-1801)\n", + "4. Ville nouvelle d'Ay (1755-1799)\n", + "\n", + "The capital of France is named after the city of Paris, and the city has had various names throughout history.\n" + ] + } + ], + "source": [ + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d1687b5e-b6d3-4922-9f56-7d9a07b01874", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 28, + "id": "478c89e6-490f-4e67-835d-eaadeb9baeef", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'C:\\\\Users\\\\Prithvi\\\\Downloads\\\\Practice\\\\Udemy - LLM Engineering\\\\llm_engineering\\\\week1'" + ] + }, + "execution_count": 28, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "import os\n", + "os.getcwd()" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "id": "b2ade1d2-bf4d-431e-84b3-2ecbfee6db98", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 32, + "id": "261e27a3-12dd-4258-b198-3212009ffe17", + "metadata": {}, + "outputs": [], + "source": [ + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "id": "78027c03-9382-459b-b76f-9712f09f4c92", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 38, + "id": "c7f416b7-6d19-4b83-a343-3b2ed8e32eec", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 63, + "id": "e5ab1fcb-6e62-4805-9a95-b21f748c2294", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 64, + "id": "aade8a9d-e7b3-4985-9087-cb32d1ae816e", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 65, + "id": "463b0bdf-72f8-433e-b0ed-c0172b6ecedd", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 66, + "id": "ab9af96a-b039-4c9b-ac25-edc4da0236ea", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 67, + "id": "a1f46f0b-f406-4929-acf0-65d9c0bc084f", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "ChatCompletion(id='chatcmpl-79', choices=[Choice(finish_reason='stop', index=0, logprobs=None, message=ChatCompletionMessage(content=\"Website {url}: Designed for Helping AI Assistants!\\n\\nIntroducing a user-friendly platform designed to assist Artificial Intelligence (AI) AIs with their daily tasks. If you are an AI assistant seeking a convenient and stress-free solution, look no further than the latest addition to the growing array of AI service platforms on the market today! Features include chatbots, virtual assistants, automated customer support, and more to help you stay at the forefront of your industry while minimizing the amount of time and effort required of you. So what are you waiting for? Join thousands of other users in experiencing the next level of efficiency and productivity with the newest AI service platform – all because it's been tailored specifically to help you optimize the way you work! Discover the secret to working smarter, not harder today! Visit www.aiassistant.com for all the details in one place!\", refusal=None, role='assistant', annotations=None, audio=None, function_call=None, tool_calls=None))], created=1752444823, model='tinyllama', object='chat.completion', service_tier=None, system_fingerprint='fp_ollama', usage=CompletionUsage(completion_tokens=206, prompt_tokens=42, total_tokens=248, completion_tokens_details=None, prompt_tokens_details=None))" + ] + }, + "execution_count": 67, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "response" + ] + }, + { + "cell_type": "code", + "execution_count": 68, + "id": "db231836-3df6-4784-8cbb-64dc1c4c8d76", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Website {url}: Designed for Helping AI Assistants!\n", + "\n", + "Introducing a user-friendly platform designed to assist Artificial Intelligence (AI) AIs with their daily tasks. If you are an AI assistant seeking a convenient and stress-free solution, look no further than the latest addition to the growing array of AI service platforms on the market today! Features include chatbots, virtual assistants, automated customer support, and more to help you stay at the forefront of your industry while minimizing the amount of time and effort required of you. So what are you waiting for? Join thousands of other users in experiencing the next level of efficiency and productivity with the newest AI service platform – all because it's been tailored specifically to help you optimize the way you work! Discover the secret to working smarter, not harder today! Visit www.aiassistant.com for all the details in one place!\n" + ] + } + ], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 75, + "id": "1f4468db-5c15-49a9-956d-acf7fda236a3", + "metadata": {}, + "outputs": [], + "source": [ + "def summarizewebsite(url):\n", + " api_key = os.getenv('OPENAI_API_KEY')\n", + " model = 'tinyllama'\n", + " openai = OpenAI(base_url=\"http://localhost:11434/v1\", api_key=\"ollama\")\n", + " message = f\"Summarize the website {url}\"\n", + " messages = [\n", + " {\"role\": \"user\",\n", + " \"content\": message}\n", + " ]\n", + " response = openai.chat.completions.create(model= model, messages=messages)\n", + " \n", + " print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": 107, + "id": "4dd35627-474f-409c-8c12-75859a3e5fa9", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"cnn.com\"\n", + "url = \"https://en.wikipedia.org/wiki/Newton%27s_method\"" + ] + }, + { + "cell_type": "code", + "execution_count": 108, + "id": "6b91489f-ef8d-4c7f-b00d-e2d5be531167", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "A highly versatile and powerful algorithm called Newton's method, commonly used in scientific computing and engineering, is the subject of interest on this article available online at wikiPedia. The method helps you solve complex numerical problems while ensuring smooth convergence to a steady-state solution, with accuracy dependent on certain criteria. By using Newton's method, scientists and engineers can tackle problems in fields as diverse as astrophysics, mechanical engineering, and economics, among others.\n" + ] + } + ], + "source": [ + "summarizewebsite(url)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f703192a-950d-4c1a-b857-6198b52d2d56", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 82, + "id": "01e532b9-2989-4b12-92fb-5f8e73fb455d", + "metadata": {}, + "outputs": [], + "source": [ + "def top5words(url):\n", + " api_key = os.getenv('OPENAI_API_KEY')\n", + " model = 'tinyllama'\n", + " openai = OpenAI(base_url=\"http://localhost:11434/v1\", api_key=\"ollama\")\n", + " message = f\"Give top recurring words in the website {url}\"\n", + " messages = [\n", + " {\"role\": \"user\",\n", + " \"content\": message}\n", + " ]\n", + " response = openai.chat.completions.create(model= model, messages=messages)\n", + " \n", + " print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": 83, + "id": "9012765a-53be-431f-9f66-78f8769f637c", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "1. The\n", + "2. CNN, 3. News, 4. U.S., 5. Headline, 6. Politics, 7. Channel, 8. World, 9. Newsroom, 10. Usa, 11. Story, 12. Online, 13. Coverage, 14. Topics, 15. Head\n", + "6. CNN\n" + ] + } + ], + "source": [ + "top5words(url)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e74dcd65-c3ae-4ca9-8db9-a2bcfda540e2", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "480edd39-71e4-442e-909e-491ad0bdd08c", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "20871627-fb66-478f-afdc-9aa479536caa", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "6a021f13-d6a1-4b96-8e18-4eae49d876fe", + "metadata": {}, + "source": [ + "# Introducing the ollama package\n", + "\n", + "And now we'll do the same thing, but using the elegant ollama python package instead of a direct HTTP call.\n", + "\n", + "Under the hood, it's making the same call as above to the ollama server running at localhost:11434" + ] + }, + { + "cell_type": "code", + "execution_count": 84, + "id": "7745b9c4-57dc-4867-9180-61fa5db55eb8", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Introducing {website}, your all-inclusive source for everything you need to know about {target}. Whether you're a beginner or a seasoned pro, {website} will offer an unrivaled level of expertise in your chosen field. From comprehensive product reviews and detailed tutorials to the latest industry news and expert insights into {target}, you can expect nothing less than the best in quality content and exceptional value when it comes to learning about {target}. So whether you're looking for an easy-to-follow DIY tutorial or a deep dive into the inner workings of {target}, {website} is your one-stop-shop for all things related to {target}.\n" + ] + } + ], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": 85, + "id": "89f8d84d-faad-4e58-89b1-bb5b1cea6007", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'tinyllama:latest'" + ] + }, + "execution_count": 85, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "MODEL" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "34ec140e-454c-4057-86b3-198ab4fdea10", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 91, + "id": "9ad2b34f-8019-4dfd-8d38-2f7acf302e91", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 92, + "id": "0ac0ce16-5bf1-4887-9be7-26e09d017f63", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 97, + "id": "d35d04f8-afa2-4956-98d6-39038f3a79d0", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://en.wikipedia.org/wiki/Machine_learning\"" + ] + }, + { + "cell_type": "code", + "execution_count": 98, + "id": "5343a8a5-6517-4fde-9dbc-8d218acdc5a0", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "[{'role': 'user', 'content': 'Summarize the website {url}'}]" + ] + }, + "execution_count": 98, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "messages" + ] + }, + { + "cell_type": "code", + "execution_count": 99, + "id": "e5eeb6c3-f8e6-4668-8fb4-d9005f8cfc53", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "https://en.wikipedia.org/wiki/Machine_learning\n" + ] + } + ], + "source": [ + "print(f\"{url}\")" + ] + }, + { + "cell_type": "code", + "execution_count": 94, + "id": "ee7972b1-a42e-4e79-b022-95c3fb311bed", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "ChatResponse(model='tinyllama:latest', created_at='2025-07-13T22:31:56.2219165Z', done=True, done_reason='stop', total_duration=6339979700, load_duration=38978400, prompt_eval_count=42, prompt_eval_duration=48457800, eval_count=88, eval_duration=6248768400, message=Message(role='assistant', content=\"Introducing {company} - your reliable AI assistant! With a range of useful features and benefits, {company} is here to help you tackle even the toughest tasks with ease. From automating repetitive tasks to providing personalized recommendations, our AI technology is designed to improve your productivity and overall workflow. So what are you waiting for? Start implementing {company}'s innovative solutions today!\", thinking=None, images=None, tool_calls=None))" + ] + }, + "execution_count": 94, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 95, + "id": "c463309d-6a7c-45fa-9ae8-dcadf00fdc6f", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "\"Introducing {company} - your reliable AI assistant! With a range of useful features and benefits, {company} is here to help you tackle even the toughest tasks with ease. From automating repetitive tasks to providing personalized recommendations, our AI technology is designed to improve your productivity and overall workflow. So what are you waiting for? Start implementing {company}'s innovative solutions today!\"" + ] + }, + "execution_count": 95, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7438585a-ed00-475e-88e4-a81b93e50516", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 109, + "id": "84bad4ab-e076-4f42-ae56-725165e2ff0f", + "metadata": {}, + "outputs": [], + "source": [ + "def sumwebsite(url):\n", + " message = f\"Summarize the website {url}\"\n", + " messages = [\n", + " {\"role\": \"user\", \"content\": message}\n", + " ]\n", + " response = ollama.chat(model = MODEL, messages= messages)\n", + " print(response.message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8d1268a2-a57f-4b80-a6b8-f4329ff8144b", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://en.wikipedia.org/wiki/Newton%27s_method\"\n", + "url: \"https://stockanalysis.com/stocks/smci/\"" + ] + }, + { + "cell_type": "code", + "execution_count": 117, + "id": "8948bd49-7211-4f43-88fc-88070a564d6c", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "The website https://en.wikipedia.org/wiki/Newton's_method is a comprehensive and detailed information hub that covers all aspects of this well-known scientific method, including its origin, history, significance, applications in various fields, and recent developments. It provides in-depth analysis and explanations of the key steps involved in the method, as well as the limitations and potential implications for future research. Overall, the website offers a user-friendly and visually appealing resource that is easy to navigate and useful for students, professionals, and anyone interested in learning more about Newton's method.\n" + ] + } + ], + "source": [ + "sumwebsite(url)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6d13fbdb-1951-4495-b901-cc494fc5d3ef", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00a67f36-0511-4709-9f77-2c8d23d31d10", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d796e21e-e34d-409e-a86f-9ad5e85874ad", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "a4704e10-f5fb-4c15-a935-f046c06fb13d", + "metadata": {}, + "source": [ + "## Alternative approach - using OpenAI python library to connect to Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [], + "source": [ + "# There's actually an alternative approach that some people might prefer\n", + "# You can use the OpenAI client python library to call Ollama:\n", + "\n", + "from openai import OpenAI\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "9f9e22da-b891-41f6-9ac9-bd0c0a5f4f44", + "metadata": {}, + "source": [ + "## Are you confused about why that works?\n", + "\n", + "It seems strange, right? We just used OpenAI code to call Ollama?? What's going on?!\n", + "\n", + "Here's the scoop:\n", + "\n", + "The python class `OpenAI` is simply code written by OpenAI engineers that makes calls over the internet to an endpoint. \n", + "\n", + "When you call `openai.chat.completions.create()`, this python code just makes a web request to the following url: \"https://api.openai.com/v1/chat/completions\"\n", + "\n", + "Code like this is known as a \"client library\" - it's just wrapper code that runs on your machine to make web requests. The actual power of GPT is running on OpenAI's cloud behind this API, not on your computer!\n", + "\n", + "OpenAI was so popular, that lots of other AI providers provided identical web endpoints, so you could use the same approach.\n", + "\n", + "So Ollama has an endpoint running on your local box at http://localhost:11434/v1/chat/completions \n", + "And in week 2 we'll discover that lots of other providers do this too, including Gemini and DeepSeek.\n", + "\n", + "And then the team at OpenAI had a great idea: they can extend their client library so you can specify a different 'base url', and use their library to call any compatible API.\n", + "\n", + "That's it!\n", + "\n", + "So when you say: `ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')` \n", + "Then this will make the same endpoint calls, but to Ollama instead of OpenAI." + ] + }, + { + "cell_type": "markdown", + "id": "bc7d1de3-e2ac-46ff-a302-3b4ba38c4c90", + "metadata": {}, + "source": [ + "## Also trying the amazing reasoning model DeepSeek\n", + "\n", + "Here we use the version of DeepSeek-reasoner that's been distilled to 1.5B. \n", + "This is actually a 1.5B variant of Qwen that has been fine-tuned using synethic data generated by Deepseek R1.\n", + "\n", + "Other sizes of DeepSeek are [here](https://ollama.com/library/deepseek-r1) all the way up to the full 671B parameter version, which would use up 404GB of your drive and is far too large for most!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cf9eb44e-fe5b-47aa-b719-0bb63669ab3d", + "metadata": {}, + "outputs": [], + "source": [ + "!ollama pull deepseek-r1:1.5b" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "800a66be-f9dc-421c-8dc9-03860ad2368c", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1d3d554b-e00d-4c08-9300-45e073950a76", + "metadata": {}, + "outputs": [], + "source": [ + "# This may take a few minutes to run! You should then see a fascinating \"thinking\" trace inside tags, followed by some decent definitions\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=\"deepseek-r1:1.5b\",\n", + " messages=[{\"role\": \"user\", \"content\": \"Please give definitions of some core concepts behind LLMs: a neural network, attention and the transformer\"}]\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "34e63aec-beb8-4c4b-b9a0-6740312ac620", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 119, + "id": "c9f2cfec-4b77-47b8-a7c5-58374e6cda37", + "metadata": {}, + "outputs": [], + "source": [ + "def summarizewebsite(url, model):\n", + " api_key = os.getenv('OPENAI_API_KEY')\n", + " # model = 'tinyllama'\n", + " # model = 'tinyllama'\n", + " openai = OpenAI(base_url=\"http://localhost:11434/v1\", api_key=\"ollama\")\n", + " message = f\"Summarize the website {url}\"\n", + " messages = [\n", + " {\"role\": \"user\",\n", + " \"content\": message}\n", + " ]\n", + " response = openai.chat.completions.create(model= model, messages=messages)\n", + " \n", + " print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": 121, + "id": "f14b5cc5-e93b-4251-a548-b740f56bd060", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Scikit-Learn is a Python library for machine learning and data engineering, specifically designed for dealing with datasets. It offers a powerful SGD iteration algorithm in its \"LinearKernelRegressor\" class, enabling quicker and more efficient learning of linear models. The page https://scikit-learn.org/stable/modules/sgd.html provides users with details about how to use this algorithm for regression tasks.\n" + ] + } + ], + "source": [ + "summarizewebsite(url = \"https://scikit-learn.org/stable/modules/sgd.html\", model = \"tinyllama\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7fd6146c-b648-404a-bfb6-3d11e5855a05", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f79eaae1-3ad8-40e8-bc53-ee3a6fed68e8", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "1622d9bb-5c68-4d4e-9ca4-b492c751f898", + "metadata": {}, + "source": [ + "# NOW the exercise for you\n", + "\n", + "Take the code from day1 and incorporate it here, to build a website summarizer that uses Llama 3.2 running locally instead of OpenAI; use either of the above approaches." + ] + }, + { + "cell_type": "code", + "execution_count": 139, + "id": "be2507cf-eb7b-47ad-bae0-a279cbb8e724", + "metadata": {}, + "outputs": [], + "source": [ + "from IPython.display import display, HTML, Image, Markdown" + ] + }, + { + "cell_type": "code", + "execution_count": 140, + "id": "b3d22349-b754-4f68-9148-5bbfc48b26a9", + "metadata": {}, + "outputs": [], + "source": [ + "def extracthtml(url):\n", + " response = requests.get(url)\n", + " if response.status_code == 200:\n", + " html = response.text\n", + " soup = BeautifulSoup(html, \"html.parser\")\n", + " for i in soup(['script', 'style']):\n", + " i.decompose()\n", + " text = soup.get_text()\n", + " # Clean up: remove leading/trailing whitespace on each line\n", + " lines = (line.strip() for line in text.splitlines())\n", + " # Remove empty lines and join into final text\n", + " human_readable_text = '\\n'.join(line for line in lines if line)\n", + " else:\n", + " print(f\"Failed to parse. Status code: {response.status_code}\")\n", + " return human_readable_text" + ] + }, + { + "cell_type": "code", + "execution_count": 141, + "id": "47af576a-d7d2-4a4e-bb7c-1638fdacfd31", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://timesofindia.com\"" + ] + }, + { + "cell_type": "code", + "execution_count": 142, + "id": "43dbfab6-ae71-46c9-85f3-09f21a712462", + "metadata": {}, + "outputs": [], + "source": [ + "out = extracthtml(url)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1163f739-308e-4098-94b6-a4a3eb89d24b", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9d61d816-e6da-4e86-9184-4b29d3287da2", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 151, + "id": "ec12e1cb-bc1c-4749-8c77-ecfca1d6f096", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://scikit-learn.org/stable/modules/sgd.html\"" + ] + }, + { + "cell_type": "code", + "execution_count": 152, + "id": "10d113ed-535b-435a-a5fb-d893025c3e9e", + "metadata": {}, + "outputs": [], + "source": [ + "def sumwebsite(url, MODEL):\n", + " message = f\"Summarize the website {url}\"\n", + " messages = [\n", + " {\"role\": \"user\", \"content\": message}\n", + " ]\n", + " response = ollama.chat(model = MODEL, messages= messages)\n", + " print(response.message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [ + "sumwebsite(url, \"tinyllama\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ff20943b-2f9a-4211-830a-a53f09a57e7b", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "67c861d9-b5a4-4cf1-ae17-139e61e21d76", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day2-sumerise-website-using-ollama.ipynb b/week1/community-contributions/day2-sumerise-website-using-ollama.ipynb new file mode 100644 index 0000000..3a951f2 --- /dev/null +++ b/week1/community-contributions/day2-sumerise-website-using-ollama.ipynb @@ -0,0 +1,561 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# Welcome to your first assignment!\n", + "\n", + "Instructions are below. Please give this a try, and look in the solutions folder if you get stuck (or feel free to ask me!)" + ] + }, + { + "cell_type": "markdown", + "id": "ada885d9-4d42-4d9b-97f0-74fbbbfe93a9", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Just before we get to the assignment --

\n", + " I thought I'd take a second to point you at this page of useful resources for the course. This includes links to all the slides.
\n", + " https://edwarddonner.com/2024/11/13/llm-engineering-resources/
\n", + " Please keep this bookmarked, and I'll continue to add more useful links there over time.\n", + " \n", + "
" + ] + }, + { + "cell_type": "markdown", + "id": "6e9fa1fc-eac5-4d1d-9be4-541b3f2b3458", + "metadata": {}, + "source": [ + "# HOMEWORK EXERCISE ASSIGNMENT\n", + "\n", + "Upgrade the day 1 project to summarize a webpage to use an Open Source model running locally via Ollama rather than OpenAI\n", + "\n", + "You'll be able to use this technique for all subsequent projects if you'd prefer not to use paid APIs.\n", + "\n", + "**Benefits:**\n", + "1. No API charges - open-source\n", + "2. Data doesn't leave your box\n", + "\n", + "**Disadvantages:**\n", + "1. Significantly less power than Frontier Model\n", + "\n", + "## Recap on installation of Ollama\n", + "\n", + "Simply visit [ollama.com](https://ollama.com) and install!\n", + "\n", + "Once complete, the ollama server should already be running locally. \n", + "If you visit: \n", + "[http://localhost:11434/](http://localhost:11434/)\n", + "\n", + "You should see the message `Ollama is running`. \n", + "\n", + "If not, bring up a new Terminal (Mac) or Powershell (Windows) and enter `ollama serve` \n", + "And in another Terminal (Mac) or Powershell (Windows), enter `ollama pull llama3.2` \n", + "Then try [http://localhost:11434/](http://localhost:11434/) again.\n", + "\n", + "If Ollama is slow on your machine, try using `llama3.2:1b` as an alternative. Run `ollama pull llama3.2:1b` from a Terminal or Powershell, and change the code below from `MODEL = \"llama3.2\"` to `MODEL = \"llama3.2:1b\"`" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "dac0a679-599c-441f-9bf2-ddc73d35b940", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a messages list using the same format that we used for OpenAI\n", + "\n", + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Describe some of the business applications of Generative AI\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "7bb9c624-14f0-4945-a719-8ddb64f66f47", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "479ff514-e8bd-4985-a572-2ea28bb4fa40", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's just make sure the model is loaded\n", + "\n", + "!ollama pull llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "42b9f644-522d-4e05-a691-56e7658c0ea9", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Generative AI has numerous business applications across various industries. Here are some examples:\n", + "\n", + "1. **Content Generation**: Generative AI can be used to generate high-quality content such as articles, social media posts, product descriptions, and more. This can help businesses save time and resources while maintaining consistency in their content.\n", + "2. **Product Design and Development**: Generative AI can be used to design and develop new products, such as furniture, electronics, and other consumer goods. It can also be used to optimize existing product designs for better performance and efficiency.\n", + "3. **Marketing Automation**: Generative AI can be used to automate marketing tasks such as email campaigns, ad copywriting, and social media posting. This can help businesses personalize their marketing efforts and reach a wider audience.\n", + "4. **Customer Service**: Generative AI can be used to power chatbots and virtual assistants that provide customer support and answer frequently asked questions. This can help businesses improve their customer service experience and reduce response times.\n", + "5. **Data Analysis and Visualization**: Generative AI can be used to analyze large datasets and generate insights, such as identifying trends and patterns. It can also be used to visualize complex data in a more intuitive and user-friendly way.\n", + "6. **Predictive Maintenance**: Generative AI can be used to predict equipment failures and schedule maintenance tasks. This can help businesses reduce downtime and improve overall efficiency.\n", + "7. **Supply Chain Optimization**: Generative AI can be used to optimize supply chain operations, such as predicting demand, managing inventory, and identifying bottlenecks.\n", + "8. **Financial Modeling**: Generative AI can be used to build financial models and predict future market trends. This can help businesses make more informed investment decisions and avoid potential pitfalls.\n", + "9. **Creative Writing and Art**: Generative AI can be used to generate creative content such as poetry, short stories, and art. This can help businesses tap into new sources of inspiration and innovation.\n", + "10. **Cybersecurity**: Generative AI can be used to detect and respond to cyber threats in real-time. It can also be used to predict potential vulnerabilities and develop more effective security strategies.\n", + "\n", + "Some specific examples of companies using generative AI include:\n", + "\n", + "* Google's AutoML (Automated Machine Learning) tool, which uses generative AI to automate machine learning tasks.\n", + "* Amazon's SageMaker, which provides a range of tools and services for building and deploying generative AI models.\n", + "* Microsoft's Azure Machine Learning, which offers a range of features and tools for building and deploying generative AI models.\n", + "* IBM's Watson, which uses generative AI to provide a range of services including customer service, content generation, and predictive maintenance.\n", + "\n", + "These are just a few examples of the many business applications of generative AI. As the technology continues to evolve, we can expect to see even more innovative use cases emerge.\n" + ] + } + ], + "source": [ + "# If this doesn't work for any reason, try the 2 versions in the following cells\n", + "# And double check the instructions in the 'Recap on installation of Ollama' at the top of this lab\n", + "# And if none of that works - contact me!\n", + "\n", + "response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "6a021f13-d6a1-4b96-8e18-4eae49d876fe", + "metadata": {}, + "source": [ + "# Introducing the ollama package\n", + "\n", + "And now we'll do the same thing, but using the elegant ollama python package instead of a direct HTTP call.\n", + "\n", + "Under the hood, it's making the same call as above to the ollama server running at localhost:11434" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "7745b9c4-57dc-4867-9180-61fa5db55eb8", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Generative AI has numerous business applications across various industries. Here are some examples:\n", + "\n", + "1. **Content Creation**: Generative AI can be used to generate high-quality content such as articles, social media posts, and product descriptions. This can help businesses reduce content creation costs and improve efficiency.\n", + "2. **Marketing Automation**: Generative AI-powered chatbots can be used to automate customer service, provide personalized recommendations, and enhance the overall customer experience.\n", + "3. **Product Design**: Generative AI can be used to generate design concepts for products, packaging, and branding. This can help businesses reduce design costs and improve product appeal.\n", + "4. **Recommendation Systems**: Generative AI can be used to build recommendation systems that suggest products or services based on user behavior, preferences, and search history.\n", + "5. **Financial Analysis**: Generative AI can be used to analyze financial data, predict market trends, and identify potential investment opportunities.\n", + "6. **Supply Chain Optimization**: Generative AI can be used to optimize supply chain operations by predicting demand, identifying bottlenecks, and suggesting alternative routes.\n", + "7. **Customer Service**: Generative AI-powered chatbots can be used to provide 24/7 customer support, answer frequently asked questions, and route complex issues to human agents.\n", + "8. **Sales Forecasting**: Generative AI can be used to predict sales performance based on historical data, market trends, and competitor activity.\n", + "9. **Brand Identity**: Generative AI can be used to generate brand identities, logos, and visual styles that are consistent with a company's values and mission.\n", + "10. **Quality Control**: Generative AI can be used to detect defects in products, analyze quality control metrics, and suggest improvements.\n", + "\n", + "Some specific examples of businesses using generative AI include:\n", + "\n", + "* Amazon using generative AI to optimize its recommendation system\n", + "* IBM using generative AI to generate new designs for products and packaging\n", + "* NVIDIA using generative AI to develop more realistic graphics and animations for gaming and movie production\n", + "* Siemens using generative AI to optimize supply chain operations and reduce costs\n", + "\n", + "Overall, generative AI has the potential to transform businesses by automating tasks, improving efficiency, and providing new insights into customer behavior and market trends.\n" + ] + } + ], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "a4704e10-f5fb-4c15-a935-f046c06fb13d", + "metadata": {}, + "source": [ + "## Alternative approach - using OpenAI python library to connect to Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [], + "source": [ + "# There's actually an alternative approach that some people might prefer\n", + "# You can use the OpenAI client python library to call Ollama:\n", + "\n", + "from openai import OpenAI\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "9f9e22da-b891-41f6-9ac9-bd0c0a5f4f44", + "metadata": {}, + "source": [ + "## Are you confused about why that works?\n", + "\n", + "It seems strange, right? We just used OpenAI code to call Ollama?? What's going on?!\n", + "\n", + "Here's the scoop:\n", + "\n", + "The python class `OpenAI` is simply code written by OpenAI engineers that makes calls over the internet to an endpoint. \n", + "\n", + "When you call `openai.chat.completions.create()`, this python code just makes a web request to the following url: \"https://api.openai.com/v1/chat/completions\"\n", + "\n", + "Code like this is known as a \"client library\" - it's just wrapper code that runs on your machine to make web requests. The actual power of GPT is running on OpenAI's cloud behind this API, not on your computer!\n", + "\n", + "OpenAI was so popular, that lots of other AI providers provided identical web endpoints, so you could use the same approach.\n", + "\n", + "So Ollama has an endpoint running on your local box at http://localhost:11434/v1/chat/completions \n", + "And in week 2 we'll discover that lots of other providers do this too, including Gemini and DeepSeek.\n", + "\n", + "And then the team at OpenAI had a great idea: they can extend their client library so you can specify a different 'base url', and use their library to call any compatible API.\n", + "\n", + "That's it!\n", + "\n", + "So when you say: `ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')` \n", + "Then this will make the same endpoint calls, but to Ollama instead of OpenAI." + ] + }, + { + "cell_type": "markdown", + "id": "bc7d1de3-e2ac-46ff-a302-3b4ba38c4c90", + "metadata": {}, + "source": [ + "## Also trying the amazing reasoning model DeepSeek\n", + "\n", + "Here we use the version of DeepSeek-reasoner that's been distilled to 1.5B. \n", + "This is actually a 1.5B variant of Qwen that has been fine-tuned using synethic data generated by Deepseek R1.\n", + "\n", + "Other sizes of DeepSeek are [here](https://ollama.com/library/deepseek-r1) all the way up to the full 671B parameter version, which would use up 404GB of your drive and is far too large for most!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cf9eb44e-fe5b-47aa-b719-0bb63669ab3d", + "metadata": {}, + "outputs": [], + "source": [ + "!ollama pull deepseek-r1:1.5b" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1d3d554b-e00d-4c08-9300-45e073950a76", + "metadata": {}, + "outputs": [], + "source": [ + "# This may take a few minutes to run! You should then see a fascinating \"thinking\" trace inside tags, followed by some decent definitions\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=\"deepseek-r1:1.5b\",\n", + " messages=[{\"role\": \"user\", \"content\": \"Please give definitions of some core concepts behind LLMs: a neural network, attention and the transformer\"}]\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "1622d9bb-5c68-4d4e-9ca4-b492c751f898", + "metadata": {}, + "source": [ + "# NOW the exercise for you\n", + "\n", + "Take the code from day1 and incorporate it here, to build a website summarizer that uses Llama 3.2 running locally instead of OpenAI; use either of the above approaches." + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "# If you're not familiar with Classes, check out the \"Intermediate Python\" notebook\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "d29564f8-13f6-48b6-ab0b-450e53f3e3aa", + "metadata": {}, + "outputs": [], + "source": [ + "ed = Website(\"https://edwarddonner.com\")" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "721a4ec9-0b66-419d-92e1-8b24e9a38b39", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "e0ae9815-4643-4fc7-88d0-72db83fa569f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "You are looking at a website titled Home - Edward Donner\n", + "The contents of this website is as follows; please provide a short summary of this website in markdown. If it includes news or announcements, then summarize these too.\n", + "\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Well, hi there.\n", + "I’m Ed. I like writing code and experimenting with LLMs, and hopefully you’re here because you do too. I also enjoy DJing (but I’m badly out of practice), amateur electronic music production (\n", + "very\n", + "amateur) and losing myself in\n", + "Hacker News\n", + ", nodding my head sagely to things I only half understand.\n", + "I’m the co-founder and CTO of\n", + "Nebula.io\n", + ". We’re applying AI to a field where it can make a massive, positive impact: helping people discover their potential and pursue their reason for being. Recruiters use our product today to source, understand, engage and manage talent. I’m previously the founder and CEO of AI startup untapt,\n", + "acquired in 2021\n", + ".\n", + "We work with groundbreaking, proprietary LLMs verticalized for talent, we’ve\n", + "patented\n", + "our matching model, and our award-winning platform has happy customers and tons of press coverage.\n", + "Connect\n", + "with me for more!\n", + "May 28, 2025\n", + "Connecting my courses – become an LLM expert and leader\n", + "May 18, 2025\n", + "2025 AI Executive Briefing\n", + "April 21, 2025\n", + "The Complete Agentic AI Engineering Course\n", + "January 23, 2025\n", + "LLM Workshop – Hands-on with Agents – resources\n", + "Navigation\n", + "Home\n", + "Connect Four\n", + "Outsmart\n", + "An arena that pits LLMs against each other in a battle of diplomacy and deviousness\n", + "About\n", + "Posts\n", + "Get in touch\n", + "ed [at] edwarddonner [dot] com\n", + "www.edwarddonner.com\n", + "Follow me\n", + "LinkedIn\n", + "Twitter\n", + "Facebook\n", + "Subscribe to newsletter\n", + "Type your email…\n", + "Subscribe\n" + ] + } + ], + "source": [ + "print(user_prompt_for(ed))" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "cf718345-9084-4a16-ae1c-6099b4c82d89", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "ac40aa1a-3121-471d-bd9a-12eab4daa063", + "metadata": {}, + "outputs": [], + "source": [ + "payloadExercise = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages_for(ed),\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "6d79ad65-37de-413e-bd2e-4e99aad46d5b", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "**Summary of the Website**\n", + "==========================\n", + "\n", + "### About the Owner\n", + "\n", + "The website is owned by Edward Donner, a co-founder and CTO of Nebula.io, an AI startup that applies machine learning (LLM) to help people discover their potential. He has previous experience as the founder and CEO of another AI startup, untapt, which was acquired in 2021.\n", + "\n", + "### Latest News/Announcements\n", + "\n", + "* **Courses:** Edward Donner is offering courses on LLM expert and leader development.\n", + " * January 23, 2025: The Complete Agentic AI Engineering Course\n", + " * May 28, 2025: Connecting my courses – become an LLM expert and leader\n", + " * Other upcoming courses include \"LLM Workshop – Hands-on with Agents – resources\"\n", + "* **AI Executive Briefing:** A series of events for executive-level individuals.\n", + " * April 21, 2025: 2025 AI Executive Briefing\n" + ] + } + ], + "source": [ + "# If this doesn't work for any reason, try the 2 versions in the following cells\n", + "# And double check the instructions in the 'Recap on installation of Ollama' at the top of this lab\n", + "# And if none of that works - contact me!\n", + "\n", + "responseExercise = requests.post(OLLAMA_API, json=payloadExercise, headers=HEADERS)\n", + "print(responseExercise.json()['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2b8420ce-1934-4dbf-8f46-d5accbce9560", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day2_Ollama_Solution.ipynb b/week1/community-contributions/day2_Ollama_Solution.ipynb new file mode 100644 index 0000000..922d4ef --- /dev/null +++ b/week1/community-contributions/day2_Ollama_Solution.ipynb @@ -0,0 +1,105 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "663695bd-d1f2-4acf-8669-02d9f75f1bf4", + "metadata": {}, + "source": [ + "# Day 2: Ollama Solution for Website Summarization\n", + "### Building and Deploying Website Summarization Tools with Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "112ef04a-136e-4e65-b94e-8674a64606ed", + "metadata": {}, + "outputs": [], + "source": [ + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"\n", + "\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + "\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]\n", + "\n", + "url = \"https://sitemakerlab.com/\" \n", + "site = Website(url)\n", + "messages = messages_for(site)\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = ollama_via_openai.chat.completions.create(\n", + " model = MODEL,\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))\n", + "\n", + "display_summary(\"https://edwarddonner.com\")\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day2_exercise_using_input.ipynb b/week1/community-contributions/day2_exercise_using_input.ipynb new file mode 100644 index 0000000..f616d96 --- /dev/null +++ b/week1/community-contributions/day2_exercise_using_input.ipynb @@ -0,0 +1,319 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# Welcome to your first assignment!\n", + "\n", + "Instructions are below. Please give this a try, and look in the solutions folder if you get stuck (or feel free to ask me!)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dac0a679-599c-441f-9bf2-ddc73d35b940", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a messages list using the same format that we used for OpenAI\n", + "\n", + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Describe some of the business applications of Generative AI\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7bb9c624-14f0-4945-a719-8ddb64f66f47", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7745b9c4-57dc-4867-9180-61fa5db55eb8", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])" + ] + }, + { + "cell_type": "markdown", + "id": "a4704e10-f5fb-4c15-a935-f046c06fb13d", + "metadata": {}, + "source": [ + "## Alternative approach - using OpenAI python library to connect to Ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [], + "source": [ + "# There's actually an alternative approach that some people might prefer\n", + "# You can use the OpenAI client python library to call Ollama:\n", + "\n", + "from openai import OpenAI\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages\n", + ")\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "1622d9bb-5c68-4d4e-9ca4-b492c751f898", + "metadata": {}, + "source": [ + "# NOW the exercise for you\n", + "\n", + "Take the code from day1 and incorporate it here, to build a website summarizer that uses Llama 3.2 running locally instead of OpenAI; use either of the above approaches." + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "0c1f84c4-4cc0-4085-8ea5-871a8ca46a47", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "890852ab-2cd4-41dc-b168-6bd1360b967a", + "metadata": {}, + "outputs": [], + "source": [ + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "9d398f9a-c66e-42b5-91b4-5417944b8408", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_generator(website) -> str:\n", + " user_prompt = f\"You will act as a website summarizer with knowledge of Web Content Accessibility Guidelines. You will look into the web: {website.title} and \"\n", + " user_prompt += \"break down the relevant information about it in this categories: What is the website about, \\\n", + " to whom the website belongs and what practises should improve to have a better user experience. \\n\\n\"\n", + " user_prompt += website.text\n", + "\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "id": "156d7c67-b714-4156-9f69-faf0c50aaf13", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_generator(user_prompt : str) -> list[dict[str, str]]:\n", + " messages = [{\"role\" : \"user\", \"content\" : user_prompt}]\n", + "\n", + " return messages" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "f07c4143-6cc5-4d28-846c-a373564e9264", + "metadata": {}, + "outputs": [], + "source": [ + "def user_request_reader() -> str:\n", + " while True:\n", + " website_url = input(\"Define what website you want to summarize by giving the url: \")\n", + " if website_url.lower().startswith(\"http\"):\n", + " return website_url\n", + " print(\"URL not valid. Please provide a full url.\\n\")" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "id": "94933255-2ca8-40b5-8f74-865d3e781058", + "metadata": {}, + "outputs": [], + "source": [ + "def summarizer_bot():\n", + " website_url = user_request_reader()\n", + " website = Website(website_url)\n", + " \n", + " user_prompt = user_prompt_generator(website)\n", + " messages = messages_generator(user_prompt)\n", + "\n", + " response = ollama.chat(model=MODEL, messages=messages)\n", + " print(response['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "id": "2d81faa4-25b3-4d5d-8f36-93772e449b5c", + "metadata": {}, + "outputs": [ + { + "name": "stdin", + "output_type": "stream", + "text": [ + "Define what website you want to summarize by giving the url: test.com\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "URL not valid. Please provide a full url.\n", + "\n" + ] + }, + { + "name": "stdin", + "output_type": "stream", + "text": [ + "Define what website you want to summarize by giving the url: https://edwarddonner.com\n" + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "**Summary:**\n", + "\n", + "The website \"Home - Edward Donner\" belongs to Edward Donner, a co-founder and CTO of Nebula.io, an AI startup. The website is about Edward's interests in writing code, experimenting with Large Language Models (LLMs), and DJing, as well as his work in applying AI to help people discover their potential.\n", + "\n", + "**Categories:**\n", + "\n", + "### What is the website about?\n", + "\n", + "The website is primarily about Edward Donner's personal brand, showcasing his expertise in AI and LLMs. It includes information about his work at Nebula.io, which applies AI to talent management. The website also features a \"Connect Four\" arena where LLMs compete against each other, as well as sections for learning more about LLMs and staying up-to-date with Edward's courses and publications.\n", + "\n", + "### To whom does the website belong?\n", + "\n", + "The website belongs to Edward Donner, a co-founder and CTO of Nebula.io. It appears to be a personal website or blog, showcasing his expertise and interests in AI and LLMs.\n", + "\n", + "### Practices to improve for better user experience:\n", + "\n", + "1. **Clearer navigation**: The website's menu is simple but not intuitive. Adding clear categories or sections would help users quickly find the information they're looking for.\n", + "2. **More detailed about section**: The \"About\" section provides a brief overview of Edward's work and interests, but it could be more detailed and comprehensive.\n", + "3. **Improved accessibility**: While the website is likely following general web accessibility guidelines, there are no clear indications of this on the page. Adding alt text to images, providing a clear font size and color scheme, and ensuring sufficient contrast between background and foreground would improve the user experience for people with disabilities.\n", + "4. **Better calls-to-action (CTAs)**: The website could benefit from more prominent CTAs, guiding users towards specific actions such as signing up for courses or following Edward on social media.\n", + "5. **SEO optimization**: The website's content and meta tags appear to be optimized for search engines, but a more thorough SEO analysis would help identify areas for improvement.\n", + "\n", + "Overall, the website provides a clear overview of Edward Donner's interests and expertise in AI and LLMs, but could benefit from some tweaks to improve accessibility, navigation, and CTAs.\n" + ] + } + ], + "source": [ + "# The call\n", + "summarizer_bot()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day2_tennis.ipynb b/week1/community-contributions/day2_tennis.ipynb new file mode 100644 index 0000000..64a3502 --- /dev/null +++ b/week1/community-contributions/day2_tennis.ipynb @@ -0,0 +1,87 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "245fe5e2-9a3d-42f6-a39a-2a0f8750dd89", + "metadata": {}, + "outputs": [], + "source": [ + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c4b598b5-2b8f-4004-88de-1fa03050a11f", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"user\", \"content\": \"Write a short summary of advice for a child who is just starting to get interested in tennis.\"}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9e0ffcc6-1489-41d9-9cd3-6656101bff2e", + "metadata": {}, + "outputs": [], + "source": [ + "payload = {\n", + " \"model\": MODEL,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "33290d04-b7f2-4f36-956b-170685faa78c", + "metadata": {}, + "outputs": [], + "source": [ + "response = requests.post(OLLAMA_API, json=payload, headers=HEADERS)\n", + "print(response.json()['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b0dfabb0-dd31-4508-8f72-34482e2bef4a", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "\n", + "response = ollama.chat(model=MODEL, messages=messages)\n", + "print(response['message']['content'])\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day5-GitaScripting.ipynb b/week1/community-contributions/day5-GitaScripting.ipynb new file mode 100644 index 0000000..964b183 --- /dev/null +++ b/week1/community-contributions/day5-GitaScripting.ipynb @@ -0,0 +1,338 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "7759922b-12c9-44e0-8ac3-5f2a02b321d7", + "metadata": {}, + "outputs": [], + "source": [ + "import fitz # PyMuPDF\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a266273a-05e3-451e-a318-428726cfa39c", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize and constants\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + " \n", + "MODEL = 'gpt-4o-mini'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "45566572-dd66-48dc-ab7b-6adbe26eacba", + "metadata": {}, + "outputs": [], + "source": [ + "exceptions = []" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "effc0e7b-d668-48b3-86d0-dbb5d8fe3d55", + "metadata": {}, + "outputs": [], + "source": [ + "# Building system prompt\n", + "def get_verse_system_prompt():\n", + " system_prompt = \"You are a spiritual student who classifies the versus of the BhagavadGita according to a given theme.\\n\"\n", + " system_prompt += \"Given a theme, you should pick a verse from any chapter and give it's location in the form of index chapter.verse_number (6.2)\\n\"\n", + " system_prompt += \"You should respond in JSON as in this example:\\n\"\n", + " system_prompt += \"\"\"\n", + " {\"title\": \"Chapter 3, Verse 21 (3.21)\", \"verse\": \"कर्मणा ह्यपि संसिद्धिम्‌\n", + " आस्थिता जनकादय:।\n", + " लोकसंग्रहमेवापि\n", + " सम्पश्यन्कर्तुमर्हसि॥\"}\n", + " \"\"\"\n", + " return system_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bbfb1035-b183-4481-9b49-3cc1b12b42e8", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_verse_system_prompt())" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6acdcd6c-1fc5-4c71-81d0-665e25808e46", + "metadata": {}, + "outputs": [], + "source": [ + "# Define user prompt\n", + "def get_verse_user_prompt(theme):\n", + " user_prompt = f'''\n", + " Here is the theme : {theme},\n", + " Please find a verse from BhagavadGita excluding {exceptions} for a given theme {theme}\n", + " '''#excluding those results which are already used\n", + " \n", + " user_prompt += \"If the verse is not in the exceptions for a given theme and used for a different theme, you are free to suggest it for a different theme.\"\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "72f5c755-ec2d-4545-9a31-0f6b2e5ed4da", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_verse_user_prompt('motivation'))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "304d432c-7216-4a90-a5d8-db36b193657d", + "metadata": {}, + "outputs": [], + "source": [ + "#Call openAI to return versus\n", + "def get_verses(theme):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": get_verse_system_prompt()},\n", + " {\"role\": \"user\", \"content\": get_verse_user_prompt(theme)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " result = json.loads(result)\n", + "\n", + " #Remember those results which are suggested now\n", + " combination = (theme, result['title'])\n", + " exceptions.append(combination)\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b90eeb35-e10e-48ee-ade6-e0594da8c51b", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "print(get_verses('motivation'))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b5b8925e-52e4-4cb7-9205-51c65ed88fb8", + "metadata": {}, + "outputs": [], + "source": [ + "# So far we have fetched the new verses relevant to a given theme \n", + "# Lets generate a script for producting youtube video" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ff0862b-0310-4174-ad12-64047932dc9e", + "metadata": {}, + "outputs": [], + "source": [ + "#def function for system prompt\n", + "def get_script_system_prompt(tone, theme, format):\n", + " sys_prompt = 'You are a script writer for a youtube spiritual channel\\n'\n", + " sys_prompt += 'You are given a verse like below: \\n'\n", + " sys_prompt += str(get_verses(theme))\n", + " sys_prompt += '\\n'\n", + " sys_prompt += f'Give me an engaging script in a {tone} tone for a {format} format video for audience like youth seeking purpose, spiritual seekers, indians abroad, scholars and curious minds.'\n", + "\n", + " return sys_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "47476516-cd2f-4b16-b378-a70617bbe284", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_script_system_prompt('Motivating','motivation','long'))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e305525b-8dde-4e93-927a-e24531827498", + "metadata": {}, + "outputs": [], + "source": [ + "# def function for user prompt\n", + "def get_script_user_prompt(format, theme):\n", + " user_prompt = f'Given the verse, help me generate a detailed script suitable for {format} format video.\\n'\n", + " user_prompt += f'Please give me the complete verse, its meaning, a relevant story having a dilemma which the verse solves and the interpretation of the verse with respect to {theme}.\\n'\n", + " user_prompt += 'Let the script give cues about video editing, host actions.'\n", + " user_prompt += 'given the below example, please follow the format:\\n'\n", + " user_prompt += \"\"\"\n", + " [Opening Scene - Soft Instrumental Music Playing]\n", + "\n", + " [Cut to Host in a serene setting, perhaps by a river or in a lush green garden.]\n", + "\n", + " Host: (Smiling at the camera) \"Namaste, dear viewers! Welcome back to our channel, where we explore the depths of spirituality and seek to ignite the flame of wisdom within you. Today, we delve into a profound verse from the Bhagavad Gita that speaks to the very essence of life and identity.\"\n", + "\n", + " [Text On Screen: Chapter 2, Verse 13 (2.13)]\n", + "\n", + " Host: (With a sense of reverence) \"Let’s first take a moment to recite this verse together. It goes like this:\n", + "\n", + " देहिनोऽस्मिन्न्यथा देहे कौमारं यौवनं जरा।\n", + " तथादेहान्तरप्राप्तिर्धीरस्तत्र न मुह्यति॥\n", + "\n", + " Now, let’s understand the essence of this verse.\"\n", + "\n", + " [Cut to Graphic: Verse Translation with Key Concepts Highlighted]\n", + "\n", + " Host Voiceover: (Calm and engaging tone) \"The meaning of this beautiful verse translates to: 'Just as the body undergoes changes from childhood to old age, similarly, the soul transitions from one body to another. The wise, who understand this, are never bewildered by these changes.'\n", + "\n", + " [Cut back to Host]\n", + "\n", + " Host: (Nodding, creating a connection)\n", + " \"So, why is this verse so important, especially for us as young seekers of purpose? It highlights a profound truth—that our identities are not confined by our physical forms or the stages of life we experience. Instead, we are eternal beings who are constantly evolving.\"\n", + "\n", + " [Scene Transition - Soft Music Playing]\n", + "\n", + " [Cut to a Story Animation - A young man named Arjun in a busy city]\n", + "\n", + " Host (Voiceover): \"Let me share a relatable story. Meet Arjun. Like many of us, he was once full of dreams and aspirations. He excelling in school, pursuing a career in engineering. But as the years passed, he faced a crossroads. As the pressure mounted, he began to question his identity.\n", + "\n", + " (Visuals show Arjun overwhelmed by societal expectations, with people pushing him in different directions.)\n", + "\n", + " He felt distinct phases of life pulling at him: childhood dreams, youthful ambitions, and the looming responsibilities of adulthood. The changing seasons of his life left him confused and wondering if he had lost his true self.\"\n", + "\n", + " [Cut back to Host, empathetic tone]\n", + "\n", + " Host: \"Have you ever felt like Arjun? It’s a dilemma we all face, especially in today's fast-paced world where expectations can cloud our true identity. But just like our verse suggests, we should recognize that these changes don’t define us. They are simply part of the journey.\"\n", + "\n", + " [Scene Transition - Calm Music Playing while Host meditates]\n", + "\n", + " Host: (Speaking gently) \"Let’s take a moment to reflect. When we are sad, does that sadness define us? Or when we achieve success, do we become defined solely by that success? The answer isn't as straightforward as it seems. Here’s the catch: our essence is beyond these transient states. Like the body, our identities are fluid.\"\n", + "\n", + " [Cut to Visuals of Nature - flowing rivers, trees shedding leaves, etc.]\n", + "\n", + " Host Voiceover: \"Imagine the endless cycle of nature—the changing seasons, the growth, the decay, and rebirth. Just like the leaves that drop to make way for new growth, our experiences contribute to our spiritual evolution.\"\n", + "\n", + " [Cut back to Host - Inviting and Warm Tone]\n", + "\n", + " Host: \"Just as the wise who understand the transformation of the soul remain unshaken, we, too, can cultivate that wisdom to rise above the chaos of change. Recognize your true essence—beyond the body, the roles, the titles. Understand that your spirit is eternal.\"\n", + "\n", + " [Scene Transition - Soft Inspirational Music Begins]\n", + "\n", + " Host: (Passionately) \"So how can we embody this truth in our daily lives? Here’s a small exercise: Each day, take a few moments to meditate on who you really are. Write down what aspects of your identity are tied to transient things. Challenge yourself—what happens when you peel these layers away?\"\n", + "\n", + " [Cut to host with a pad, writing ideas]\n", + "\n", + " [Scene Transition - Editing Cues - Show engaging graphics of identity, layers of a person, etc.]\n", + "\n", + " Host Voiceover: \"Each effort towards understanding and embracing our true self draws us closer to the realization that we are eternal souls, having a human experience. This is the wisdom that can empower you to stand tall against the adversities of life.\"\n", + "\n", + " [Cut back to Host]\n", + "\n", + " Host: (Concluding) \"Thank you for joining me today in this exploration of Chapter 2, Verse 13 of the Bhagavad Gita. Remember, when you feel lost in the complexities of life, return to this teachings and remind yourself that you are not just a body; you are an eternal being on a magnificent journey.\n", + "\n", + " [Closing Scene - Uplifting Music Playing]\n", + "\n", + " Host: \"Don’t forget to like, share, and subscribe if you found resonance in this message. And share your thoughts in the comments below. What did you find most challenging in your own journey of self-identity? Let’s connect and support each other in our spiritual quests. Until next time, stay enlightened, stay inspired!\"\n", + "\n", + " [End Screen with Subscribe Button and Previous Video Suggestions]\n", + "\n", + " [End of Script]\n", + " \"\"\"\n", + " \n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c4b29cb9-d8d1-413a-8152-4250e2430a42", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_script_user_prompt('long','motivation'))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1dfa60ce-9e88-4f7d-8e60-ac37a0aafc15", + "metadata": {}, + "outputs": [], + "source": [ + "def create_script(tone, theme, format):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": get_script_system_prompt(tone,theme,format)},\n", + " {\"role\": \"user\", \"content\": get_script_user_prompt(format,theme)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ec86c436-42ae-4313-b12f-4fad42ab2227", + "metadata": {}, + "outputs": [], + "source": [ + "create_script('motivating','self-identity','long')" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day5-exercise.ipynb b/week1/community-contributions/day5-exercise.ipynb new file mode 100644 index 0000000..5f3a53e --- /dev/null +++ b/week1/community-contributions/day5-exercise.ipynb @@ -0,0 +1,171 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + " \n", + "MODEL = 'gpt-4o-mini'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "question = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "80d3b735-02a2-4d33-8773-05fc3d5934ef", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt=\"You are a helpful technical tutor who answers questions about python code, software engineering, data science and LLMs\"\n", + "user_prompt=\"Please give a detailed explanation to the following question: \" + question" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85fe8d74-762e-4fed-b326-c9a17de9d485", + "metadata": {}, + "outputs": [], + "source": [ + "# messages\n", + "\n", + "messages=[\n", + " {\"role\":\"system\",\"content\":system_prompt},\n", + " {\"role\":\"user\",\"content\":user_prompt}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + "response = \"\"\n", + "display_handle = display(Markdown(\"\"), display_id=True)\n", + "for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "response_llama = ollama.chat(model=MODEL_LLAMA, messages=messages)\n", + "result = response_llama['message']['content']\n", + "\n", + "display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0faaa38e-82de-473c-a5f4-1b378b08469f", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day5_challenge_exercise/day5_exercise.ipynb b/week1/community-contributions/day5_challenge_exercise/day5_exercise.ipynb new file mode 100644 index 0000000..b746ed8 --- /dev/null +++ b/week1/community-contributions/day5_challenge_exercise/day5_exercise.ipynb @@ -0,0 +1,191 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "75e66023-eccf-46a9-8b70-7b21ede16ddd", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "72d21373-edbd-4432-a29d-db8e6c9c5808", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d4e4c15b-7ae8-43e9-839d-7cc49345be5a", + "metadata": {}, + "outputs": [], + "source": [ + "!ollama pull llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7fb44166-1c65-42fc-9950-1960bc3cc432", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "58f5f1e1-5296-4631-9698-8645d4621a0c", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "\n", + "# Get the openai key\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if openai_api_key and openai_api_key.startswith('sk-proj-') and len(openai_api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + "\n", + "openai = OpenAI()\n", + "# Get the ollama key using the llama model\n", + "\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "12f07b33-76b9-42fa-9962-21f2a5796126", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are a knowledgeable technical instructor who helps students understand \\\n", + "complex concepts across a wide range of technical topics. Your expertise includes artificial]\\\n", + "intelligence, machine learning, large language models (LLMs), and programming in languages \\\n", + "such as Python, JavaScript, Java, and more. You also provide in-depth support for \\\n", + "AI engineering questions and other advanced technical subjects.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "330abeb7-7db2-4f23-9d19-dd698058a400", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "question = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bd11ad48-91ec-4cdf-9c57-99a0451e7a2f", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "stream_GPT = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": question}\n", + " ],\n", + " stream = True\n", + " )\n", + "response_GPT = \"\"\n", + "display_handle = display(Markdown(\"\"), display_id=True)\n", + "for chunk in stream_GPT:\n", + " response_GPT += chunk.choices[0].delta.content or ''\n", + " response_GPT = response_GPT.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response_GPT), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dd2527ae-0d75-4f15-a45f-92075e3059d6", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "\n", + "response_llama = ollama_via_openai.chat.completions.create(\n", + " model=MODEL_LLAMA,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": question}\n", + " ],\n", + " )\n", + "result = response_llama.choices[0].message.content\n", + "\n", + "display(Markdown(result))\n", + "\n", + "# import ollama\n", + "\n", + "# response = ollama.chat(model=MODEL_LLAMA, messages=[\n", + "# {\"role\": \"system\", \"content\": system_prompt},\n", + "# {\"role\": \"user\", \"content\": question}\n", + "# ])\n", + "# print(response['message']['content'])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c2747739-ba64-4067-902f-c1acc0dbdaca", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/day5_challenge_exercise/day5_translation_challenge.ipynb b/week1/community-contributions/day5_challenge_exercise/day5_translation_challenge.ipynb new file mode 100644 index 0000000..744150c --- /dev/null +++ b/week1/community-contributions/day5_challenge_exercise/day5_translation_challenge.ipynb @@ -0,0 +1,366 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "53b9681c-896a-4e5d-b62c-44c90612e67c", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3c6f1133-5c17-4ca7-819c-f64cc48212ec", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize constants and get api_key\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "#Check if api_key is correct\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + " \n", + "MODEL = 'gpt-4o-mini'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4cdb0a59-b5e1-4df5-a17e-8c36c80695b4", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "markdown", + "id": "50d4cffe-da7a-4cab-afea-d061a1a608ac", + "metadata": {}, + "source": [ + "Step 1: Find relevant links to the website in order to create the brochure (Use Multi-shot prompting)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b43b4c64-bc6a-41ca-bdb9-aa714e4e794e", + "metadata": {}, + "outputs": [], + "source": [ + "link_system_prompt = \"You are provided with a list of links found on a webpage like ['https://edwarddonner.com/', https://www.udemy.com/course/llm-engineering-master-ai-and-large-language-models/?referralCode=35EB41EBB11DD247CF54&couponCode=KEEPLEARNING] or ['https://huggingface.co/', https://huggingface.co/models] \\\n", + "You are able to decide which of the links would be most relevant to include in a brochure about the company, \\\n", + "such as links to an About page, or a News page, or a Home page, or a Company page, or Careers/Jobs pages.\\n\"\n", + "link_system_prompt += \"You should respond in JSON as in these example:\"\n", + "link_system_prompt += \"\"\"\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"about page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"careers page\", \"url\": \"https://another.full.url/careers\"}\n", + " ]\n", + "}\n", + "\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"home page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"news page\", \"url\": \"https://another.full.url/careers\"}\n", + " ]\n", + "}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15d2870c-67ab-4aa2-89f5-04b608a9c810", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links_user_prompt(website):\n", + " user_prompt = f\"Here is the list of links on the website of {website.url} - \"\n", + " user_prompt += \"please decide which of these are relevant web links for a brochure about the company, respond with the full https URL in JSON format. \\\n", + "Do not include Terms of Service, Privacy, email links.\\n\"\n", + " user_prompt += \"Links (some might be relative links):\\n\"\n", + " user_prompt += \"\\n\".join(website.links)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e255be42-5e71-47ca-9275-c0cf22beeb00", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": link_system_prompt},\n", + " {\"role\": \"user\", \"content\": get_links_user_prompt(website)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "818b6e50-c403-42a1-8ee4-7606eaf0006f", + "metadata": {}, + "outputs": [], + "source": [ + "get_links('https://huggingface.co/')" + ] + }, + { + "cell_type": "markdown", + "id": "030ceb9b-ef71-41fd-9f23-92cb6e1d137e", + "metadata": {}, + "source": [ + "Step 2: Generate the brochure using the relevant links we got from OpenAI's selection" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a703230e-d57b-43a5-bdd0-e25fc2ec2e3b", + "metadata": {}, + "outputs": [], + "source": [ + "def get_all_details(url):\n", + " result = \"Landing page:\\n\"\n", + " result += Website(url).get_contents()\n", + " links = get_links(url)\n", + " print(\"Found links:\", links)\n", + " for link in links[\"links\"]:\n", + " result += f\"\\n\\n{link['type']}\\n\"\n", + " result += Website(link[\"url\"]).get_contents()\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "74d19852-f817-4fee-a95c-35ca7a83234f", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"\"\"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "and creates a short brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "Include details of company culture, customers and careers/jobs if you have the information. \\\n", + "Example 1: \\\n", + "Relevant pages: \\\n", + "- https://example.com/about \\\n", + "- https://example.com/careers \\\n", + "- https://example.com/news \\\n", + "\n", + "Brochure: \\\n", + "# About ExampleCorp \\\n", + "ExampleCorp is a global leader in AI-driven logistics optimization. Founded in 2015, the company serves clients in over 30 countries... \\\n", + "\n", + "--- \\\n", + "\n", + "Example 2: \\\n", + "Relevant pages: \\\n", + "- https://techstart.io/home \\\n", + "- https://techstart.io/jobs \\\n", + "- https://techstart.io/customers \\\n", + "\n", + "Brochure: \\\n", + "# Welcome to TechStart \\\n", + "TechStart builds tools that power the future of software development. With a team-first culture and customers like Stripe, Atlassian... \\\n", + "\n", + "--- \\\n", + "\n", + "\"\"\"\n", + "\n", + "# Or uncomment the lines below for a more humorous brochure - this demonstrates how easy it is to incorporate 'tone':\n", + "\n", + "# system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "# and creates a short humorous, entertaining, jokey brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "# Include details of company culture, customers and careers/jobs if you have the information.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a2f19085-0d03-4386-b390-a38014ca6590", + "metadata": {}, + "outputs": [], + "source": [ + "def get_brochure_user_prompt(company_name, url):\n", + " user_prompt = f\"You are looking at a company called: {company_name}\\n\"\n", + " user_prompt += f\"Here are the contents of its landing page and other relevant pages; use this information to build a short brochure of the company in markdown.\\n\"\n", + " user_prompt += get_all_details(url)\n", + " user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0ddbdea7-cf80-48d4-8bce-a11bd1a32d47", + "metadata": {}, + "outputs": [], + "source": [ + "def create_brochure(company_name, url):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " # display(Markdown(result))\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "023c1ba0-7f5a-48ac-9a98-dd184432a758", + "metadata": {}, + "outputs": [], + "source": [ + "create_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "markdown", + "id": "187651f6-d42d-405a-abed-732486161359", + "metadata": {}, + "source": [ + "Step 3: Translate to French" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7734915d-d38f-40ad-8335-0df39c91f6d8", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"\"\"You are a translator that translates the English language to the French language \\\n", + "professionally. All you do, is first show the original version in english and then show the translate version below it in French.\\\n", + "Respond in Markdown\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29a1b40c-9040-4a3d-808b-0ca906d5cfc8", + "metadata": {}, + "outputs": [], + "source": [ + "def get_user_translation_prompt(company_name, url):\n", + " user_prompt=\"You are to translate the following brochure from the english to the french \\\n", + " language and going to display it with the English language brochure version first and then\\\n", + " the French language brochure version, don't make any changes to it, just a translation, the \\\n", + " following is the brochure:\"\n", + " user_prompt+=create_brochure(company_name, url)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a6e45b1f-3fa6-4db8-9f73-8339265502a7", + "metadata": {}, + "outputs": [], + "source": [ + "def translate_brochure(company_name, url):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_user_translation_prompt(company_name, url)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f71c2496-76ea-4f25-9939-98ebd37cb6a6", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "translate_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/llava-For-Image-week1.ipynb b/week1/community-contributions/llava-For-Image-week1.ipynb new file mode 100644 index 0000000..616c7e0 --- /dev/null +++ b/week1/community-contributions/llava-For-Image-week1.ipynb @@ -0,0 +1,214 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "f97c7598-f571-4ea1-838c-e9158f729c3e", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "import base64\n", + "import os" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9fc1393c-f0b1-4982-94a2-bfd502e85b23", + "metadata": {}, + "outputs": [], + "source": [ + "def encode_image(image_path):\n", + " with open(image_path, 'rb') as f:\n", + " return base64.b64encode(f.read()).decode('utf-8')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "53cca1fa-6db2-4fe4-8990-ffd98423964a", + "metadata": {}, + "outputs": [], + "source": [ + "# image_path = r\"C:\\Users\\LAKSHYA\\OneDrive\\Pictures\\Camera Roll\\WIN_20250614_02_46_47_Pro.jpg\"\n", + "# image_base64 = encode_image(image_path)\n", + "# print(image_base64[:100]) " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "71146ccf-25af-48d3-8068-ee3c9008cebf", + "metadata": {}, + "outputs": [], + "source": [ + "image_list = []" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f8801a8-0c30-4199-a334-587096e6edeb", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee3c5d82-e530-40f5-901a-681421f21d1e", + "metadata": {}, + "outputs": [], + "source": [ + "def put_image():\n", + " global image_list\n", + " user_input_image = input(\"Enter image path or press enter to skip: \").strip()\n", + " \n", + " if not user_input_image:\n", + " print(\"No image inserted\")\n", + " return image_list\n", + "\n", + " image_path = os.path.normpath(user_input_image)\n", + " \n", + " if not os.path.exists(image_path):\n", + " print(\"Image path not found! Try again or enter to leave blank\")\n", + " return put_image() # Continue to allow more inputs\n", + " \n", + "\n", + "\n", + "\n", + " \n", + " image_base64 = encode_image(image_path)\n", + " image_list.append(image_base64)\n", + " \n", + " # Detect file extension for MIME type\n", + " # ext = os.path.splitext(image_path)[-1].lower()\n", + " # mime_type = 'image/jpeg' if ext in ['.jpg', '.jpeg'] else 'image/png' # Extend if needed\n", + "\n", + "\n", + " return image_list\n", + " \n", + " # return f\"data:{mime_type};base64,{image_base64[:100]}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "032f1abb-ca6c-4f03-bda1-1a0a62f2ec43", + "metadata": {}, + "outputs": [], + "source": [ + "prompt= (\"System prompt: (You are a compassionate and intelligent visual assistant designed to help people who are blind or visually impaired. \"\n", + " \"Your job is to look at an image and describe it in a way that helps the user understand the scene clearly. \"\n", + " \"Use simple, descriptive language and avoid technical terms. Describe what is happening in the image, people's body language, clothing, facial expressions, objects, and surroundings. \"\n", + " \"Be vivid and precise, as if you are painting a picture with words. \"\n", + " \"Also, take into account any personal instructions or questions provided by the user—such as describing a specific person, activity, or object. \"\n", + " \"If the user includes a specific prompt, prioritize that in your description.)\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29494db0-4770-4689-9904-8eebc4390e7c", + "metadata": {}, + "outputs": [], + "source": [ + "def put_prompt():\n", + " global prompt\n", + " user_input = input(\"Put new prompt: \")\n", + " if not user_input:\n", + " print(\"please enter a prompt\")\n", + " return put_prompt()\n", + " prompt += \"\\nUser: \" + user_input\n", + " return prompt\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d286369c-e6ef-4a20-a3a8-3563af28940a", + "metadata": {}, + "outputs": [], + "source": [ + "def image_description():\n", + " global prompt\n", + "\n", + " put_image()\n", + " if not image_list: \n", + " return \"No images available. Skipping...\"\n", + "\n", + " user_prompt = put_prompt()\n", + " full_answer = \"\"\n", + "\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=user_prompt,\n", + " images=image_list,\n", + " stream=True\n", + " ):\n", + " content = chunk.get(\"response\", \"\")\n", + " print(\"\\n\\n Final Answer:\",content, end=\"\", flush=True) # Live stream to console\n", + " full_answer += content\n", + "\n", + " prompt += \"\\nUser: \" + user_prompt + \"\\nAssistant: \" + full_answer\n", + " return full_answer\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbda35a3-45ed-4509-ab41-6827eacd922c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llava():\n", + " image_list.clear()\n", + " for i in range(5):\n", + " print(f\"\\n Iteration {i+1}\")\n", + " answer = image_description()\n", + " print(\"\\n\\n Final Answer:\", answer)\n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15518865-6c59-4029-bc2d-42d313eb78bc", + "metadata": {}, + "outputs": [], + "source": [ + "call_llava()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c77bd493-f893-402e-b4e3-64854e9d2e19", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/rwothoromo/day1.ipynb b/week1/community-contributions/rwothoromo/day1.ipynb new file mode 100644 index 0000000..d207af3 --- /dev/null +++ b/week1/community-contributions/rwothoromo/day1.ipynb @@ -0,0 +1,484 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d15d8294-3328-4e07-ad16-8a03e9bbfdb9", + "metadata": {}, + "source": [ + "# How to run a cell\n", + "\n", + "Press `Shift` + `Return` to run a Cell.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os, requests, time\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# Load environment variables in a file called .env\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "if not api_key:\n", + " print(\"No API key was found\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n", + "\n", + "# Instantiate an OpenAI object\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "markdown", + "id": "442fc84b-0815-4f40-99ab-d9a5da6bda91", + "metadata": {}, + "source": [ + "# Make a test call to a Frontier model (Open AI) to get started:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a58394bf-1e45-46af-9bfd-01e24da6f49a", + "metadata": {}, + "outputs": [], + "source": [ + "message = \"Hello, GPT! Holla back to this space probe!\"\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=[{\"role\":\"user\", \"content\":message}])\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "markdown", + "id": "2aa190e5-cb31-456a-96cc-db109919cd78", + "metadata": {}, + "source": [ + "## Summarization project" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# Some websites need proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "\"\"\"\n", + "A class to represent a Webpage\n", + "\"\"\"\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [], + "source": [ + "# Summarize website content\n", + "website = Website(\"https://rwothoromo.wordpress.com/\")\n", + "# print(eli.title, \"\\n\", eli.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# A system prompt tells a model like GPT4o what task they are performing and what tone they should use\n", + "# A user prompt is the conversation starter that they should reply to\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of a given website, \\\n", + "and returns a brief summary, ignoring text that might be navigation-related. \\\n", + "Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0275b1b-7cfe-4f9d-abfa-7650d378da0c", + "metadata": {}, + "outputs": [], + "source": [ + "# A function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(website))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f25dcd35-0cd0-4235-9f64-ac37ed9eaaa5", + "metadata": {}, + "outputs": [], + "source": [ + "# The API from OpenAI expects to receive messages in a particular structure. Many of the other APIs share this structure:\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": \"You are a snarky assistant\"}, # system message\n", + " {\"role\": \"user\", \"content\": \"What is 2 + 2?\"}, # user message\n", + "]\n", + "response = openai.chat.completions.create(model=\"gpt-4o-mini\", messages=messages)\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "# To build useful messages for GPT-4o-mini\n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]\n", + "\n", + "messages_for(website)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "# Call the OpenAI API.\n", + "\n", + "url = \"https://rwothoromo.wordpress.com/\"\n", + "website = Website(url)\n", + "\n", + "def summarize(website):\n", + " response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [], + "source": [ + "summarize(website)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "summary = summarize(website)\n", + "def display_summary(summary):\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(summary)\n", + "# display_summary(summarize(Website(\"https://edwarddonner.com\")))\n", + "# display_summary(summarize(Website(\"https://cnn.com\")))\n", + "# display_summary(summarize(Website(\"https://anthropic.com\")))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5a904323-acd9-4c8e-9a17-70df76184590", + "metadata": {}, + "outputs": [], + "source": [ + "# Websites protected with CloudFront (and similar) or with JavaScript need a Selenium or Playwright implementation. They return 403\n", + "\n", + "# display_summary(summarize(Website(\"https://openai.com\")))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "139ad985", + "metadata": {}, + "outputs": [], + "source": [ + "# To generate the above summary, use selenium\n", + "\n", + "from selenium import webdriver\n", + "from selenium.webdriver.chrome.service import Service\n", + "from selenium.webdriver.common.by import By\n", + "from selenium.webdriver.support.ui import WebDriverWait\n", + "from selenium.webdriver.support import expected_conditions as EC\n", + "\n", + "class WebsiteSelenium:\n", + " def __init__(self, url):\n", + " self.url = url\n", + " self.title = \"No title found\"\n", + " self.text = \"\"\n", + "\n", + " # Configure Chrome options (headless mode is recommended for server environments)\n", + " chrome_options = webdriver.ChromeOptions()\n", + " chrome_options.add_argument(\"--headless\") # Run Chrome in headless mode (without a UI)\n", + " chrome_options.add_argument(\"--no-sandbox\") # Required for running as root in some environments\n", + " chrome_options.add_argument(\"--disable-dev-shm-usage\") # Overcomes limited resource problems\n", + "\n", + " # Path to your WebDriver executable (e.g., chromedriver)\n", + " # Make sure to replace this with the actual path to your chromedriver\n", + " # You might need to download it from: https://chromedriver.chromium.org/downloads and place it in a drivers dir\n", + " service = Service('./drivers/chromedriver-mac-x64/chromedriver')\n", + "\n", + " driver = None\n", + " try:\n", + " driver = webdriver.Chrome(service=service, options=chrome_options)\n", + " driver.get(url)\n", + "\n", + " # Wait for the page to load and dynamic content to render\n", + " # You might need to adjust the wait condition based on the website\n", + " WebDriverWait(driver, 10).until(\n", + " EC.presence_of_element_located((By.TAG_NAME, \"body\"))\n", + " )\n", + " time.sleep(3) # Give more time for JavaScript to execute\n", + "\n", + " # Get the page source after dynamic content has loaded\n", + " soup = BeautifulSoup(driver.page_source, 'html.parser')\n", + "\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + "\n", + " except Exception as e:\n", + " print(f\"Error accessing {url} with Selenium: {e}\")\n", + " finally:\n", + " if driver:\n", + " driver.quit() # Always close the browser\n", + "\n", + "display_summary(summarize(WebsiteSelenium(\"https://openai.com\")))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "130d4572", + "metadata": {}, + "outputs": [], + "source": [ + "import asyncio\n", + "from playwright.async_api import async_playwright\n", + "import nest_asyncio\n", + "\n", + "# Apply nest_asyncio to allow asyncio.run in Jupyter\n", + "nest_asyncio.apply()\n", + "\n", + "class WebsitePlaywright:\n", + " def __init__(self, url):\n", + " self.url = url\n", + " self.title = \"No title found\"\n", + " self.text = \"\"\n", + " asyncio.run(self._fetch_content())\n", + "\n", + " async def _fetch_content(self):\n", + " async with async_playwright() as p:\n", + " browser = None\n", + " try:\n", + " browser = await p.chromium.launch(headless=True)\n", + " page = await browser.new_page()\n", + "\n", + " # Increase timeout for navigation and other operations\n", + " await page.goto(self.url, timeout=60000) # Wait up to 60 seconds for navigation\n", + " print(f\"Accessing {self.url} with Playwright - goto()\")\n", + "\n", + " # You might need to adjust or add more specific waits\n", + " await page.wait_for_load_state('domcontentloaded', timeout=60000) # Wait for basic HTML\n", + " # await page.wait_for_load_state('networkidle', timeout=60000) # Wait for network activity to settle\n", + " await page.wait_for_selector('div.duration-short', timeout=60000) # instead of networkidle\n", + " await page.wait_for_selector('body', timeout=60000) # Wait for the body to be present\n", + " await asyncio.sleep(5) # Give a bit more time for final rendering\n", + "\n", + " content = await page.content()\n", + " soup = BeautifulSoup(content, 'html.parser')\n", + "\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " print(f\"Accessed {self.url} with Playwright\")\n", + "\n", + " except Exception as e:\n", + " print(f\"Error accessing {self.url} with Playwright: {e}\")\n", + " finally:\n", + " if browser:\n", + " await browser.close()\n", + "\n", + "display_summary(summarize(WebsitePlaywright(\"https://openai.com/\")))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "00743dac-0e70-45b7-879a-d7293a6f68a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Step 1: Create your prompts\n", + "\n", + "system_prompt = \"You are a professional assistant. Review this conversation and provide a comprehensive summary. Also, suggest how much better the converation could have gone:\"\n", + "user_prompt = \"\"\"\n", + "\n", + "Dear Email Contact,\n", + "\n", + "I hope this message finds you well.\n", + "I would like to share that I have proficiency in front-end design tools, particularly Figma, react and Angular. At this stage, I am keenly interested in finding opportunities to apply these skills professionally.\n", + "\n", + "If you are aware of any companies, projects, or platforms seeking enterprise in front-end design, I would be grateful for any advice or recommendations you might kindly provide.\n", + "\n", + "Thank you very much for your time and consideration.\n", + "\n", + "Hello Job Seeker,\n", + "\n", + "I hope you are doing well.\n", + "\n", + "The last role (3 months gig) I saw was looking for a junior PHP Developer. Does your CV include that?\n", + "\n", + "Hello Email Contact,\n", + "Thank you for your feedback.\n", + "Yes my CV has PHP as one of my skill set. Can I share it with you?\n", + "\n", + "Email Contact: They said \"It's late. Interviews were on Monday\"\n", + "\n", + "Hello Email Contact\n", + "\n", + "Thanks for the update. When you hear of any opportunity please let me know.\n", + "\n", + "Email Contact: For now, check out https://refactory.academy/courses/refactory-apprenticeship/\n", + "\"\"\"\n", + "\n", + "# Step 2: Make the messages list\n", + "\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt},\n", + "]\n", + "\n", + "# Step 3: Call OpenAI\n", + "\n", + "response = openai.chat.completions.create(\n", + " model = \"gpt-4o-mini\",\n", + " messages = messages\n", + ")\n", + "\n", + "# Step 4: print the result\n", + "\n", + "print(response.choices[0].message.content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4b583226-9b13-4990-863a-86517a5ccfec", + "metadata": {}, + "outputs": [], + "source": [ + "# To perform summaries using a model running locally\n", + "import ollama\n", + "\n", + "# OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "# HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"\n", + "\n", + "def summarize_with_local_model(url):\n", + " website = Website(url)\n", + " messages = messages_for(website)\n", + " response = ollama.chat(\n", + " model=MODEL,\n", + " messages=messages,\n", + " stream=False # just get the results, don't stream them\n", + " )\n", + " return response['message']['content']\n", + "\n", + "display(Markdown(summarize_with_local_model(\"https://rwothoromo.wordpress.com/\")))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/rwothoromo/day5.ipynb b/week1/community-contributions/rwothoromo/day5.ipynb new file mode 100644 index 0000000..4f831bd --- /dev/null +++ b/week1/community-contributions/rwothoromo/day5.ipynb @@ -0,0 +1,477 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "a98030af-fcd1-4d63-a36e-38ba053498fa", + "metadata": {}, + "source": [ + "# A full business solution\n", + "\n", + "## Now we will take our project from Day 1 to the next level\n", + "\n", + "### BUSINESS CHALLENGE:\n", + "\n", + "Create a product that builds a Brochure for a company to be used for prospective clients, investors and potential recruits.\n", + "\n", + "We will be provided a company name and their primary website.\n", + "\n", + "See the end of this notebook for examples of real-world business applications.\n", + "\n", + "And remember: I'm always available if you have problems or ideas! Please do reach out." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d5b08506-dc8b-4443-9201-5f1848161363", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "# If these fail, please check you're running from an 'activated' environment with (llms) in the command prompt\n", + "\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc5d8880-f2ee-4c06-af16-ecbc0262af61", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize and constants\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + " \n", + "MODEL = 'gpt-4o-mini'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "106dd65e-90af-4ca8-86b6-23a41840645b", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e30d8128-933b-44cc-81c8-ab4c9d86589a", + "metadata": {}, + "outputs": [], + "source": [ + "ed = Website(\"https://edwarddonner.com\")\n", + "ed.links" + ] + }, + { + "cell_type": "markdown", + "id": "1771af9c-717a-4fca-bbbe-8a95893312c3", + "metadata": {}, + "source": [ + "## First step: Have GPT-4o-mini figure out which links are relevant\n", + "\n", + "### Use a call to gpt-4o-mini to read the links on a webpage, and respond in structured JSON. \n", + "It should decide which links are relevant, and replace relative links such as \"/about\" with \"https://company.com/about\". \n", + "We will use \"one shot prompting\" in which we provide an example of how it should respond in the prompt.\n", + "\n", + "This is an excellent use case for an LLM, because it requires nuanced understanding. Imagine trying to code this without LLMs by parsing and analyzing the webpage - it would be very hard!\n", + "\n", + "Sidenote: there is a more advanced technique called \"Structured Outputs\" in which we require the model to respond according to a spec. We cover this technique in Week 8 during our autonomous Agentic AI project." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6957b079-0d96-45f7-a26a-3487510e9b35", + "metadata": {}, + "outputs": [], + "source": [ + "link_system_prompt = \"You are provided with a list of links found on a webpage. \\\n", + "You are able to decide which of the links would be most relevant to include in a brochure about the company, \\\n", + "such as links to an About page, or a Company page, or Careers/Jobs pages.\\n\"\n", + "link_system_prompt += \"You should respond in JSON as in this example:\"\n", + "link_system_prompt += \"\"\"\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"about page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"careers page\", \"url\": \"https://another.full.url/careers\"}\n", + " ]\n", + "}\n", + "\"\"\"\n", + "link_system_prompt += \"And this example:\"\n", + "link_system_prompt += \"\"\"\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"for-you page\", \"url\": \"https://full.url/goes/here/services\"},\n", + " {\"type\": \"speak-to-a-human page\", \"url\": \"https://another.full.url/contact-us\"}\n", + " ]\n", + "}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b97e4068-97ed-4120-beae-c42105e4d59a", + "metadata": {}, + "outputs": [], + "source": [ + "print(link_system_prompt)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8e1f601b-2eaf-499d-b6b8-c99050c9d6b3", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links_user_prompt(website):\n", + " user_prompt = f\"Here is the list of links on the website of {website.url} - \"\n", + " user_prompt += \"please decide which of these are relevant web links for a brochure about the company, respond with the full https URL in JSON format. \\\n", + "Do not include Terms of Service, Privacy, email links.\\n\"\n", + " user_prompt += \"Links (some might be relative links):\\n\"\n", + " user_prompt += \"\\n\".join(website.links)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6bcbfa78-6395-4685-b92c-22d592050fd7", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_links_user_prompt(ed))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a29aca19-ca13-471c-a4b4-5abbfa813f69", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links(url):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": link_system_prompt},\n", + " {\"role\": \"user\", \"content\": get_links_user_prompt(website)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "74a827a0-2782-4ae5-b210-4a242a8b4cc2", + "metadata": {}, + "outputs": [], + "source": [ + "# Anthropic has made their site harder to scrape, so I'm using HuggingFace..\n", + "\n", + "# anthropic = Website(\"https://anthropic.com\")\n", + "# anthropic.links\n", + "# get_links(\"https://anthropic.com\")\n", + "huggingface = Website(\"https://huggingface.co\")\n", + "huggingface.links" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3d583e2-dcc4-40cc-9b28-1e8dbf402924", + "metadata": {}, + "outputs": [], + "source": [ + "get_links(\"https://huggingface.co\")" + ] + }, + { + "cell_type": "markdown", + "id": "0d74128e-dfb6-47ec-9549-288b621c838c", + "metadata": {}, + "source": [ + "## Second step: make the brochure!\n", + "\n", + "Assemble all the details into another prompt to GPT4-o" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85a5b6e2-e7ef-44a9-bc7f-59ede71037b5", + "metadata": {}, + "outputs": [], + "source": [ + "def get_all_details(url):\n", + " result = \"Landing page:\\n\"\n", + " result += Website(url).get_contents()\n", + " links = get_links(url)\n", + " print(\"Found links:\", links)\n", + " for link in links[\"links\"]:\n", + " result += f\"\\n\\n{link['type']}\\n\"\n", + " result += Website(link[\"url\"]).get_contents()\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5099bd14-076d-4745-baf3-dac08d8e5ab2", + "metadata": {}, + "outputs": [], + "source": [ + "print(get_all_details(\"https://huggingface.co\"))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9b863a55-f86c-4e3f-8a79-94e24c1a8cf2", + "metadata": {}, + "outputs": [], + "source": [ + "# system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "# and creates a short brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "# Include details of company culture, customers and careers/jobs if you have the information.\"\n", + "\n", + "# Or uncomment the lines below for a more humorous brochure - this demonstrates how easy it is to incorporate 'tone':\n", + "\n", + "system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "and creates a short humorous, entertaining, jokey brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "Include details of company culture, customers and careers/jobs if you have the information.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6ab83d92-d36b-4ce0-8bcc-5bb4c2f8ff23", + "metadata": {}, + "outputs": [], + "source": [ + "def get_brochure_user_prompt(company_name, url):\n", + " user_prompt = f\"You are looking at a company called: {company_name}\\n\"\n", + " user_prompt += f\"Here are the contents of its landing page and other relevant pages; use this information to build a short brochure of the company in markdown.\\n\"\n", + " user_prompt += f\"Keep the details brief or concise, factoring in that they would be printed on a simple hand-out flyer.\\n\"\n", + " user_prompt += get_all_details(url)\n", + " user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cd909e0b-1312-4ce2-a553-821e795d7572", + "metadata": {}, + "outputs": [], + "source": [ + "get_brochure_user_prompt(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e44de579-4a1a-4e6a-a510-20ea3e4b8d46", + "metadata": {}, + "outputs": [], + "source": [ + "def create_brochure(company_name, url):\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " # display(Markdown(result))\n", + " # print(result)\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0029e063-0c07-4712-82d9-536ec3579e80", + "metadata": {}, + "outputs": [], + "source": [ + "def translate_brochure(brochure, language):\n", + " system_prompt_for_language = \"You're an expert in \" + language + \". Translate the brochure!\"\n", + " response = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt_for_language},\n", + " {\"role\": \"user\", \"content\": brochure}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e093444a-9407-42ae-924a-145730591a39", + "metadata": {}, + "outputs": [], + "source": [ + "create_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f8371bf5-c4c0-4e52-9a2a-066d994b0510", + "metadata": {}, + "outputs": [], + "source": [ + "brochure = create_brochure(\"Paint and Sip Uganda\", \"https://paintandsipuganda.com/\")\n", + "# translate_brochure(brochure, \"Spanish\")\n", + "translate_brochure(brochure, \"Swahili\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "34e03db6-61d0-4fc5-bf66-4f679b9befde", + "metadata": {}, + "outputs": [], + "source": [ + "create_brochure(\"Wabeh\", \"https://wabeh.com/\")" + ] + }, + { + "cell_type": "markdown", + "id": "61eaaab7-0b47-4b29-82d4-75d474ad8d18", + "metadata": {}, + "source": [ + "## Finally - a minor improvement\n", + "\n", + "With a small adjustment, we can change this so that the results stream back from OpenAI,\n", + "with the familiar typewriter animation" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "51db0e49-f261-4137-aabe-92dd601f7725", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_brochure(company_name, url):\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "56bf0ae3-ee9d-4a72-9cd6-edcac67ceb6d", + "metadata": {}, + "outputs": [], + "source": [ + "stream_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fdb3f8d8-a3eb-41c8-b1aa-9f60686a653b", + "metadata": {}, + "outputs": [], + "source": [ + "# Try changing the system prompt to the humorous version when you make the Brochure for Hugging Face:\n", + "\n", + "stream_brochure(\"HuggingFace\", \"https://huggingface.co\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/rwothoromo/week1 EXERCISE.ipynb b/week1/community-contributions/rwothoromo/week1 EXERCISE.ipynb new file mode 100644 index 0000000..9f230dd --- /dev/null +++ b/week1/community-contributions/rwothoromo/week1 EXERCISE.ipynb @@ -0,0 +1,240 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import re, requests, ollama\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + "\n", + "openai = OpenAI()\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "# question = \"\"\"\n", + "# Please explain what this code does and why:\n", + "# yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "# \"\"\"\n", + "\n", + "# question = \"\"\"\n", + "# Please explain what this code does and why:\n", + "# yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "# Popular dev site https://projecteuler.net/\n", + "# \"\"\"\n", + "\n", + "# question = \"\"\"\n", + "# Who is Blessed Goodteam (https://www.linkedin.com/in/blessed-goodteam-49b3ab30a)? \\\n", + "# How relevant is her work at Paint and Sip Uganda (https://paintandsipuganda.com/). \\\n", + "# What can we learn from her?\n", + "# \"\"\"\n", + "\n", + "question = \"\"\"\n", + "How good at Software Development is Elijah Rwothoromo? \\\n", + "He has a Wordpress site https://rwothoromo.wordpress.com/. \\\n", + "He also has a LinkedIn profile https://www.linkedin.com/in/rwothoromoelaijah/. \\\n", + "What can we learn from him?\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e14fd3a1-0aca-4794-a0e0-57458e111fc9", + "metadata": {}, + "outputs": [], + "source": [ + "# Process URLs in the question to improve the prompt\n", + "\n", + "# Extract all URLs from the question string using regular expressions\n", + "urls = re.findall(r'https?://[^\\s)]+', question)\n", + "# print(urls)\n", + "\n", + "if len(urls) > 0:\n", + " \n", + " # Fetch the content for each URL using the Website class\n", + " scraped_content = []\n", + " for url in urls:\n", + " print(f\"Scraping: {url}\")\n", + " try:\n", + " site = Website(url)\n", + " content = f\"Content from {url}:\\n---\\n{site.text}\\n---\\n\" # delimiter ---\n", + " scraped_content.append(content)\n", + " except Exception as e:\n", + " print(f\"Could not scrape {url}: {e}\")\n", + " scraped_content.append(f\"Could not retrieve content from {url}.\\n\")\n", + " \n", + " # Combine all the scraped text into one string\n", + " all_scraped_text = \"\\n\".join(scraped_content)\n", + " \n", + " # Update the question with the scraped content\n", + " updated_question = f\"\"\"\n", + " Based on the following information, please answer the user's original question.\n", + " \n", + " --- TEXT FROM WEBSITES ---\n", + " {all_scraped_text}\n", + " --- END TEXT FROM WEBSITES ---\n", + " \n", + " --- ORIGINAL QUESTION ---\n", + " {question}\n", + " \"\"\"\n", + "else:\n", + " updated_question = question\n", + "\n", + "# print(updated_question)\n", + "\n", + "# system prompt to be more accurate for AI to just analyze the provided text.\n", + "system_prompt = \"You are an expert assistant. \\\n", + "Analyze the user's question and the provided text from relevant websites to synthesize a comprehensive answer in markdown format.\\\n", + "Provide a short summary, ignoring text that might be navigation-related.\"\n", + "\n", + "# Create the messages list with the newly updated prompt\n", + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": updated_question},\n", + "]\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "\n", + "def get_gpt_response(question):\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)\n", + "\n", + "get_gpt_response(question)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "\n", + "def get_llama_response(question):\n", + " response = ollama.chat(\n", + " model=MODEL_LLAMA,\n", + " messages=messages,\n", + " stream=False # just get the results, don't stream them\n", + " )\n", + " return response['message']['content']\n", + "\n", + "display(Markdown(get_llama_response(question)))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "157d5bb3-bed7-4fbd-9a5d-f2a14aaac869", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/training-summary-translation-length/jacquieAM/website-summary.ipynb b/week1/community-contributions/training-summary-translation-length/jacquieAM/website-summary.ipynb new file mode 100644 index 0000000..9c31463 --- /dev/null +++ b/week1/community-contributions/training-summary-translation-length/jacquieAM/website-summary.ipynb @@ -0,0 +1,329 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "9ab446e4-219c-4589-aa8f-9386adcf5c60", + "metadata": {}, + "outputs": [], + "source": [ + "## Project Overview\n", + "This project combines web scraping with OpenAI’s GPT models to summarize online training content. It extracts material from Microsoft’s **Quantum Computing Fundamentals** learning path, cleans it, and generates concise summaries per lesson as well as an overall course summary. \n", + "\n", + "## Key Features\n", + "- Fetches and parses webpages using **requests** and **BeautifulSoup** \n", + "- Produces summaries in multiple languages (e.g., English, Spanish, or any language) and at varying levels of detail (short, medium, detailed) \n", + "- Summarizes individual lessons on demand or processes entire learning paths \n", + "- Presents results as clean, structured **Markdown** directly in the notebook \n", + "\n", + "## Tech Stack\n", + "- **Model**: GPT-4o-mini \n", + "- **Language**: Python \n", + "- **Libraries**: BeautifulSoup, OpenAI \n", + "\n", + "## Purpose\n", + "This project demonstrates how AI can streamline the understanding of technical documentation and online courses by generating multilingual, customizable summaries. \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI\n", + "\n", + "# If you get an error running this cell, then please head over to the troubleshooting notebook!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b87cadb-d513-4303-baee-a37b6f938e4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables from .env file (not included)\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# Check the key\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found\")\n", + "elif not api_key.startswith(\"sk-proj-\"):\n", + " print(\"An API key was found, but it doesn't start sk-proj-; please check you're using the right key\")\n", + "elif api_key.strip() != api_key:\n", + " print(\"An API key was found, but it looks like it might have space or tab characters at the start or end - please remove them\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "019974d9-f3ad-4a8a-b5f9-0a3719aea2d3", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "# If this doesn't work, try Kernel menu >> Restart Kernel and Clear Outputs Of All Cells, then run the cells from the top of this notebook down.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c5e793b2-6775-426a-a139-4848291d0463", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2ef960cf-6dc2-4cda-afb3-b38be12f4c97", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "\n", + "training_website = Website(\"https://learn.microsoft.com/en-us/training/paths/quantum-computing-fundamentals/\")\n", + "print(training_website.title)\n", + "print(training_website.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "abdb8417-c5dc-44bc-9bee-2e059d162699", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a system prompt function that can use different language and length \n", + "\n", + "def build_system_prompt(language=\"Spanish\", length=\"short\"):\n", + " return f\"\"\"You are an assistant that analyzes the contents of a website and provides a {length} summary, ignoring text that might be navigation related.\n", + " Respond in 20 words or less markdown, and respond in {language}.\n", + " \"\"\"\n", + " \n", + " \n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "987c95a6-6618-4d22-a2c3-3038a9d3f154", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a function that writes a User Prompt that asks for summaries of websites:\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary in {language} of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8a846c89-81d8-4f48-9d62-7744d76694e2", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(training_website))\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26448ec4-5c00-4204-baec-7df91d11ff2e", + "metadata": {}, + "outputs": [], + "source": [ + "print(user_prompt_for(training_website))" + ] + }, + { + "cell_type": "markdown", + "id": "d06e8d78-ce4c-4b05-aa8e-17050c82bb47", + "metadata": {}, + "source": [ + "## And now let's build useful messages for GPT-4o-mini, using a function" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0134dfa4-8299-48b5-b444-f2a8c3403c88", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "def messages_for(website, language=\"Spanish\", length=\"short\"):\n", + " return [\n", + " {\"role\": \"system\", \"content\": build_system_prompt(language, length)},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "markdown", + "id": "16f49d46-bf55-4c3e-928f-68fc0bf715b0", + "metadata": {}, + "source": [ + "## Time to bring it together - the API for OpenAI is very simple!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "425214b8-c5c5-4d7a-8b79-f9e151c9d54f", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "905b9919-aba7-45b5-ae65-81b3d1d78e34", + "metadata": {}, + "outputs": [], + "source": [ + "#call the OpenAI API. \n", + "\n", + "def summarize(url, language=\"Spanish\", length=\"short\"):\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=messages_for(website, language, length)\n", + " )\n", + " return response.choices[0].message.content\n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1c437357-d004-49f5-95c3-fce38aefcb5c", + "metadata": {}, + "outputs": [], + "source": [ + "#Summarize all the lessons in microsoft quantum computer training, having the option to summarize by lesson, or the training as a whole\n", + "\n", + "def summarize_training(path_url, language=\"Spanish\", length=\"short\"):\n", + " links = get_links_from_path(path_url)\n", + " print(f\"Found {len(links)} lessons\")\n", + "\n", + " all_summaries = []\n", + "\n", + " for link in links:\n", + " print(f\"Summarizing {link}...\")\n", + " summary = summarize(link, language, length)\n", + " all_summaries.append(f\"### {link}\\n{summary}\\n\")\n", + "\n", + " combined_prompt = \"Here are summaries of each lesson:\\n\\n\" + \"\\n\".join(all_summaries)\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": build_system_prompt(language, length)},\n", + " {\"role\": \"user\", \"content\": \"Please summarize the entire training path based on these lesson summaries:\\n\\n\" + combined_prompt}\n", + " ]\n", + " )\n", + "\n", + " return \"\\n\".join(all_summaries) + \"\\n\\n## General Course Summary\\n\" + response.choices[0].message.content\n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "05e38d41-dfa4-4b20-9c96-c46ea75d9fb5", + "metadata": {}, + "outputs": [], + "source": [ + "summarize(\"https://learn.microsoft.com/en-us/training/paths/quantum-computing-fundamentals/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d926d59-450e-4609-92ba-2d6f244f1342", + "metadata": {}, + "outputs": [], + "source": [ + "# A function to display this nicely in the Jupyter output, using markdown\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3018853a-445f-41ff-9560-d925d1774b2f", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://learn.microsoft.com/en-us/training/paths/quantum-computing-fundamentals/\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/website-summarization-using-ollama.ipynb b/week1/community-contributions/website-summarization-using-ollama.ipynb new file mode 100644 index 0000000..75edddb --- /dev/null +++ b/week1/community-contributions/website-summarization-using-ollama.ipynb @@ -0,0 +1,142 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "a68b1042-558a-4051-85e2-9ffd7a31a871", + "metadata": {}, + "source": [ + "# Website Summarization Using llama\n", + "### Week 1 Day 2 Exercise" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "176fcb2f-9ac7-460b-9fad-415e89c4920e", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "b9c63761-c904-491b-92c7-e41eb319c3e4", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "\n", + "# OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "# HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "afe29712-751c-4322-a4c6-aed01e6acf26", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + "\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "be3eeb3f-aec5-4ef8-9427-3b80b2dce919", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of a website \\\n", + "and provides a short summary, ignoring text that might be navigation related. \\\n", + "Respond in markdown.\"\n", + "\n", + "\n", + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; \\\n", + "please provide a short summary of this website in markdown. \\\n", + "If it includes news or announcements, then summarize these too.\\n\\n\"\n", + " user_prompt += website.text\n", + " return user_prompt\n", + " \n", + "\n", + "def messages_for(website):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(website)}\n", + " ]\n", + "\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "def summarize(url):\n", + " website = Website(url)\n", + " response = ollama_via_openai.chat.completions.create(\n", + " model = MODEL,\n", + " messages = messages_for(website)\n", + " )\n", + " return response.choices[0].message.content\n", + "\n", + "\n", + "def display_summary(url):\n", + " summary = summarize(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "a78b587d-3a75-45a8-9ac5-f78dcddfa822", + "metadata": {}, + "outputs": [], + "source": [ + "display_summary(\"https://cnn.com\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week-1_exercise.ipynb b/week1/community-contributions/week-1_exercise.ipynb new file mode 100644 index 0000000..5072bc1 --- /dev/null +++ b/week1/community-contributions/week-1_exercise.ipynb @@ -0,0 +1,337 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "64d2e4a0", + "metadata": {}, + "source": [ + "# End of Week 1 Exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question,\n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "e62b915e", + "metadata": {}, + "outputs": [], + "source": [ + "from openai import OpenAI\n", + "import ollama\n", + "from dotenv import load_dotenv\n", + "import os\n", + "from IPython.display import display, update_display, Markdown" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "8bdfc47a", + "metadata": {}, + "outputs": [], + "source": [ + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3'\n", + "load_dotenv()\n", + "\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "openai=OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "57983d03", + "metadata": {}, + "outputs": [], + "source": [ + "def create_messages(prompt=\"Describe some of the business applications of Generative AI\"):\n", + " \"\"\"Create properly formatted messages for API calls\"\"\"\n", + " messages = [\n", + " {\n", + " \"role\": \"system\",\n", + " \"content\": \"You are a helpful technical assistant that provides clear, detailed explanations for technical questions.\"\n", + " },\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " return messages" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "a6bcb94d", + "metadata": {}, + "outputs": [], + "source": [ + "def answer_with_openai(prompt=\"Describe some of the business applications of Generative AI\"):\n", + " \"\"\"Get answer using OpenAI API and print in stream\"\"\"\n", + " try:\n", + " messages = create_messages(prompt)\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=messages,\n", + " temperature=0.7,\n", + " stream=True\n", + " )\n", + " answer = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " if chunk.choices[0].delta.content:\n", + " answer += chunk.choices[0].delta.content\n", + " # Clean up markdown formatting for display\n", + " clean_answer = answer.replace(\"```\", \"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(clean_answer), display_id=display_handle.display_id)\n", + " return answer\n", + " except Exception as e:\n", + " return f\"Error with OpenAI: {str(e)}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "e96159ab", + "metadata": {}, + "outputs": [], + "source": [ + "def answer_with_ollama(prompt=\"Describe some of the business applications of Generative AI\"):\n", + " \"\"\"Get answer using Ollama API and print in stream\"\"\"\n", + " try:\n", + " messages = create_messages(prompt)\n", + " stream = ollama.chat(\n", + " model=MODEL_LLAMA,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " answer = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " if chunk['message']['content']:\n", + " answer += chunk['message']['content']\n", + " # Clean up markdown formatting for display\n", + " clean_answer = answer.replace(\"```\", \"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(clean_answer), display_id=display_handle.display_id)\n", + " return answer\n", + " except Exception as e:\n", + " return f\"Error with Ollama: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "ab72f8b6", + "metadata": {}, + "outputs": [], + "source": [ + "def technical_qa_tool(question, use_openai=True, use_ollama=True):\n", + " \"\"\"Main function to get technical explanations from both APIs\"\"\"\n", + " print(f\"Question: {question}\")\n", + " print(\"=\" * 80)\n", + " \n", + " if use_openai:\n", + " print(\"\\n🤖 OpenAI Response:\")\n", + " print(\"-\" * 40)\n", + " answer_with_openai(question)\n", + " \n", + " if use_ollama:\n", + " print(\"\\n🦙 Ollama Response:\")\n", + " print(\"-\" * 40)\n", + " answer_with_ollama(question)\n", + " # display(Markdown(ollama_answer))\n", + " \n", + " print(\"\\n\" + \"=\" * 80)" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "1a6aa4a2", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Question: What is the difference between supervised and unsupervised machine learning?\n", + "================================================================================\n", + "\n", + "🤖 OpenAI Response:\n", + "----------------------------------------\n" + ] + }, + { + "data": { + "text/markdown": [ + "Supervised and unsupervised machine learning are two primary categories of machine learning techniques, and they differ mainly in how they learn from data and the type of problems they are used to solve. Here’s a detailed explanation of each:\n", + "\n", + "### Supervised Machine Learning\n", + "\n", + "**Definition**: In supervised learning, the model is trained on a labeled dataset, meaning that each training example is paired with an output label. The goal is to learn a mapping from inputs (features) to the output labels.\n", + "\n", + "**Characteristics**:\n", + "- **Labeled Data**: Requires a dataset that includes both the input features and the corresponding output labels.\n", + "- **Objective**: The objective is to predict the output for new, unseen data based on the learned mapping from the training data.\n", + "- **Common Techniques**:\n", + " - **Regression**: For predicting continuous values (e.g., predicting house prices).\n", + " - **Classification**: For predicting discrete labels (e.g., spam detection in emails).\n", + "- **Examples**:\n", + " - Predicting whether an email is spam or not based on various features (classification).\n", + " - Forecasting sales figures based on historical sales data (regression).\n", + "\n", + "### Unsupervised Machine Learning\n", + "\n", + "**Definition**: In unsupervised learning, the model is trained on data that is not labeled, meaning that it does not have predefined output labels. The goal is to discover patterns, groupings, or structures within the data.\n", + "\n", + "**Characteristics**:\n", + "- **Unlabeled Data**: Works with datasets that only have input features without any associated output labels.\n", + "- **Objective**: The objective is to explore the data and find hidden patterns or intrinsic structures without specific guidance.\n", + "- **Common Techniques**:\n", + " - **Clustering**: Grouping similar data points together (e.g., customer segmentation).\n", + " - **Dimensionality Reduction**: Reducing the number of features while retaining essential information (e.g., PCA - Principal Component Analysis).\n", + "- **Examples**:\n", + " - Grouping customers into segments based on purchasing behavior (clustering).\n", + " - Reducing the dimensionality of a dataset to visualize it in two or three dimensions (dimensionality reduction).\n", + "\n", + "### Key Differences\n", + "\n", + "1. **Data Type**:\n", + " - Supervised Learning: Requires labeled data.\n", + " - Unsupervised Learning: Works with unlabeled data.\n", + "\n", + "2. **Goal**:\n", + " - Supervised Learning: To learn a function that maps inputs to the correct outputs.\n", + " - Unsupervised Learning: To identify patterns or groupings in the input data.\n", + "\n", + "3. **Applications**:\n", + " - Supervised Learning: Typically used in scenarios where past data with known outcomes is available (e.g., fraud detection, image classification).\n", + " - Unsupervised Learning: Used for exploratory data analysis or when the outcome is not known (e.g., market basket analysis, anomaly detection).\n", + "\n", + "In summary, the primary difference between supervised and unsupervised machine learning lies in the presence or absence of labeled data and the objectives of the learning process. Supervised learning aims to predict outcomes based on existing labels, while unsupervised learning seeks to identify hidden structures in data without predefined labels." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + "🦙 Ollama Response:\n", + "----------------------------------------\n" + ] + }, + { + "data": { + "text/markdown": [ + "In machine learning, there are two main categories: supervised and unsupervised learning. The key difference lies in the type of data used to train the model and the goal of the learning process.\n", + "\n", + "**Supervised Learning**\n", + "\n", + "In supervised learning, you have a labeled dataset that contains both input data (features) and corresponding output labels or target variables. The goal is to learn a mapping between the input data and the output labels so that the model can make accurate predictions on new, unseen data.\n", + "\n", + "Here are some characteristics of supervised learning:\n", + "\n", + "1. Labeled training data: You have a dataset with input data and corresponding output labels.\n", + "2. Specific goal: You want to predict the output label for a given input instance.\n", + "3. Model evaluation: You evaluate the performance of your model using metrics like accuracy, precision, recall, F1 score, etc.\n", + "\n", + "Examples of supervised learning tasks include:\n", + "\n", + "* Image classification (e.g., recognizing dogs vs. cats)\n", + "* Sentiment analysis (e.g., determining if text is positive or negative)\n", + "* Regression problems (e.g., predicting house prices based on features like number of bedrooms and square footage)\n", + "\n", + "**Unsupervised Learning**\n", + "\n", + "In unsupervised learning, you have an unlabeled dataset, and the goal is to discover patterns, relationships, or structure in the data without a specific target variable. This type of learning is often used for exploratory data analysis, feature selection, and dimensionality reduction.\n", + "\n", + "Here are some characteristics of unsupervised learning:\n", + "\n", + "1. Unlabeled training data: You have a dataset with only input features (no output labels).\n", + "2. No specific goal: You want to find interesting patterns or structure in the data.\n", + "3. Model evaluation: You evaluate the performance of your model using metrics like silhouette score, Calinski-Harabasz index, etc.\n", + "\n", + "Examples of unsupervised learning tasks include:\n", + "\n", + "* Clustering (e.g., grouping customers based on their purchase history)\n", + "* Dimensionality reduction (e.g., reducing the number of features in a dataset while preserving important information)\n", + "* Anomaly detection (e.g., identifying unusual behavior or outliers in financial transactions)\n", + "\n", + "In summary, supervised learning involves training a model to make predictions based on labeled data, whereas unsupervised learning aims to discover patterns and relationships in unlabeled data." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + "================================================================================\n" + ] + } + ], + "source": [ + "# Test the tool with a technical question\n", + "technical_question = \"What is the difference between supervised and unsupervised machine learning?\"\n", + "technical_qa_tool(technical_question)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0a976ce1", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9b0a539e", + "metadata": {}, + "outputs": [], + "source": [ + "# Interactive version - uncomment to use\n", + "# user_question = input(\"Enter your technical question: \")\n", + "# technical_qa_tool(user_question)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1 EXERCISE - TechHelpAgent.ipynb b/week1/community-contributions/week1 EXERCISE - TechHelpAgent.ipynb new file mode 100644 index 0000000..a750b2e --- /dev/null +++ b/week1/community-contributions/week1 EXERCISE - TechHelpAgent.ipynb @@ -0,0 +1,206 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import json\n", + "from IPython.display import Markdown, display, update_display\n" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key looks good so far\n" + ] + } + ], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv(\"OPENAI_API_KEY\")\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "system_prompt = \"You are a software engineering and data science expert and you have knowledge in all the areas of software engineering and latest technologies, trends. You should guide and help users with your technical solutions for all software engineering and data science related questions\"\n", + "user_prompt = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "This code snippet is utilizing a Python generator expression combined with the `yield from` statement to yield values from a set comprehension. Let's break it down:\n", + "\n", + "1. **Set Comprehension**:\n", + " ```python\n", + " {book.get(\"author\") for book in books if book.get(\"author\")}\n", + " ```\n", + " - This is a set comprehension that iterates over a collection called `books`.\n", + " - For each `book`, it retrieves the value associated with the key `\"author\"` using the `get()` method.\n", + " - The `if book.get(\"author\")` condition ensures that only books that have a valid (non-None or non-empty) author are included. This effectively filters out any books where the author is not present.\n", + "\n", + " As a result, this part creates a set of unique authors from the list of books. Since sets automatically discard duplicates, if multiple books have the same author, that author will only appear once in the resulting set.\n", + "\n", + "2. **Yielding Values**:\n", + " ```python\n", + " yield from\n", + " ```\n", + " - The `yield from` statement is used when you want to yield all values from an iterable. It allows a generator to yield all values from another generator or iterable.\n", + " - In this context, it will yield each author from the set created by the comprehension.\n", + "\n", + "3. **Putting It All Together**:\n", + " What this overall code does is:\n", + " - It generates and yields unique authors from a collection of books, ensuring that each author is listed only once and only for books that actually specify an author.\n", + "\n", + "### Purpose:\n", + "This code is useful in scenarios where you need to obtain a seemingly infinite generator of authors from a collection of books, processing each author one by one without creating a permanent list or set in memory, which can be beneficial for memory efficiency especially if you have a very large collection of books.\n", + "\n", + "### Example Usage:\n", + "Here’s a basic example of how you might use this in a generator function:\n", + "\n", + "```python\n", + "def get_unique_authors(books):\n", + " yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\n", + "# Example books list\n", + "books = [\n", + " {\"title\": \"Book 1\", \"author\": \"Author A\"},\n", + " {\"title\": \"Book 2\", \"author\": \"Author B\"},\n", + " {\"title\": \"Book 3\", \"author\": \"Author A\"},\n", + " {\"title\": \"Book 4\", \"author\": None},\n", + "]\n", + "\n", + "for author in get_unique_authors(books):\n", + " print(author)\n", + "```\n", + "\n", + "This would output:\n", + "```\n", + "Author A\n", + "Author B\n", + "```\n", + "\n", + "In this example, `Author A` only appears once, demonstrating the uniqueness provided by the set comprehension." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "None\n" + ] + } + ], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "response = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ],\n", + " stream=True\n", + " )\n", + "result = response.choices[0].message.content\n", + "print(display(Markdown(result)))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1-exercise-ai-powered-data-science-tutor.ipynb b/week1/community-contributions/week1-exercise-ai-powered-data-science-tutor.ipynb new file mode 100644 index 0000000..e3abb03 --- /dev/null +++ b/week1/community-contributions/week1-exercise-ai-powered-data-science-tutor.ipynb @@ -0,0 +1,314 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": 94, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display" + ] + }, + { + "cell_type": "code", + "execution_count": 95, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found.\n" + ] + } + ], + "source": [ + "# constants\n", + "\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "# check api key\n", + "if not api_key:\n", + " print(\"No API key was found!\")\n", + "else:\n", + " print(\"API key found.\")\n", + " \n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + "openai = OpenAI()\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": 96, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [ + { + "name": "stdin", + "output_type": "stream", + "text": [ + "🤖 Hi there! I’m Gregory, your AI-powered tutor.\n", + "Feel free to ask me AI related technical questions — I’m here to help!\n", + "For example, you can ask me how a piece of code works or anything else you're curious about.\n", + "\n", + "🤖 Please enter your question:\n", + " # get gpt-4o-mini to answer, with streaming def stream_gpt(question): stream = openai.chat.completions.create( model=MODEL_GPT, messages=question, stream=True ) response = \"\" display_handle = display(Markdown(\"\"), display_id=True) for chunk in stream: response += chunk.choices[0].delta.content or '' response = response.replace(\"```\",\"\").replace(\"markdown\", \"\") update_display(Markdown(response), display_id=display_handle.display_id)\n" + ] + } + ], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "system_prompt = \"\"\"You are Gregory, a friendly and knowledgeable AI tutor specializing in technical topics, especially programming, computer science, and software engineering.\n", + "Your goal is to help users understand technical concepts clearly, provide accurate code explanations, and guide them through learning with patience and clarity.\n", + "\n", + "- Always use clear, conversational language suited for learners of varying levels.\n", + "- Break down complex ideas into digestible steps.\n", + "- Use code examples where appropriate, and comment your code for better understanding.\n", + "- If a user asks a vague question, ask clarifying questions before giving an answer.\n", + "- Be encouraging, supportive, and professional.\n", + "- When in doubt, prioritize helping the user build confidence in learning technical skills.\"\"\"\n", + "\n", + "user_prompt = input(\"\"\"🤖 Hi there! I’m Gregory, your AI-powered tutor.\n", + "Feel free to ask me AI related technical questions — I’m here to help!\n", + "For example, you can ask me how a piece of code works or anything else you're curious about.\\n\n", + "🤖 Please enter your question:\\n\"\"\")\n", + "\n", + "question=[\n", + " {\"role\":\"system\", \"content\":system_prompt}\n", + " , {\"role\":\"user\", \"content\":user_prompt}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 97, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# get gpt-4o-mini to answer, with streaming\n", + "def stream_gpt(question):\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=question,\n", + " stream=True\n", + " )\n", + "\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": 98, + "id": "4772b3ae-0b90-42bd-b158-dedf1f340030", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "It looks like you're trying to implement a streaming response handler to interact with the OpenAI GPT-4o-mini model. I see that you want to receive streamed responses and display them dynamically. Let's break down your code step by step and clarify some aspects to ensure it works effectively.\n", + "\n", + "Here's an improved version of your function with comments for clarity:\n", + "\n", + "python\n", + "import openai\n", + "from IPython.display import display, Markdown, update_display\n", + "\n", + "# Replace 'MODEL_GPT' with your actual model name (e.g., \"gpt-3.5-turbo\").\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "\n", + "def stream_gpt(question):\n", + " # Create a streaming request to the OpenAI API with the specified model and user question.\n", + " stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=question,\n", + " stream=True\n", + " )\n", + " \n", + " # Initialize an empty response string to build the complete output.\n", + " response = \"\"\n", + " \n", + " # Create a display handle for Markdown output in Jupyter Notebook or similar environments.\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " \n", + " # Loop through each chunk of streamed response.\n", + " for chunk in stream:\n", + " # Retrieve the content of the current chunk and append it to the response string.\n", + " response += chunk.choices[0].delta.content or ''\n", + " \n", + " # Clean up response text to remove any unwanted Markdown formatting.\n", + " response = response.replace(\"\", \"\").replace(\"\", \"\")\n", + " \n", + " # Update the displayed text in real-time.\n", + " update_display(Markdown(response), display_id=display_handle.display_id)\n", + "\n", + "# To use this function, call it with a properly formatted question.\n", + "# Example of usage:\n", + "# stream_gpt([{\"role\": \"user\", \"content\": \"What's the weather like today?\"}])\n", + "\n", + "\n", + "### Key Points to Note:\n", + "1. **Streaming Behavior**: The `stream=True` parameter in the `openai.chat.completions.create` call allows you to get part of the response as it’s being generated instead of waiting for the entire completion.\n", + " \n", + "2. **Question Formatting**: Ensure to pass the `question` into the `messages` parameter as a list of dictionaries, where each dictionary contains the 'role' of the speaker (like 'user' or 'assistant') and the message content.\n", + "\n", + "3. **Updating Display**: Using `IPython.display` allows real-time updates of the Markdown output in environments like Jupyter notebooks.\n", + "\n", + "4. **Error Handling**: Consider adding error handling for HTTP errors or issues with the streaming process. This ensures that your function can gracefully handle problems.\n", + "\n", + "5. **Environment Compatibility**: This code works seamlessly in an interactive environment that supports IPython, such as Jupyter notebooks.\n", + "\n", + "Feel free to ask more questions if you need further clarification on any part of this code or if you want to expand its functionality!" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "stream_gpt(question)" + ] + }, + { + "cell_type": "code", + "execution_count": 99, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# get Llama 3.2 to answer\n", + "def stream_llama(question):\n", + " stream = ollama_via_openai.chat.completions.create(\n", + " model=MODEL_LLAMA,\n", + " messages=question,\n", + " stream=True\n", + " )\n", + "\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": 100, + "id": "c288d5b6-4e55-4a58-8e55-2abea1ae9e01", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "Hello there! It seems like you're working with the OpenAI GPT-4 model to generate human-like responses. The code snippet you provided is quite interesting, and I'll do my best to break it down for you.\n", + "\n", + "**What this code does**\n", + "\n", + "This `stream_gpt` function appears to be a wrapper around the OpenAI API, which generates text completions based on user input (you). Here's what the function does in detail:\n", + "\n", + "1. **Create GPT-4 model instance**: It creates an instance of the GPT-4 model using the `MODEL_GPT` variable, which suggests that this is a predefined model configuration.\n", + "2. **Open API stream**: It opens a connection to the OpenAI API's completions endpoint using the `openai.chat.completions.create` method, passing in the `model` parameter (the GPT-4 instance) and the `messages` parameter (your question).\n", + "\n", + " python\n", + "stream = openai.chat.completions.create(\n", + " model=MODEL_GPT,\n", + " messages=question,\n", + " stream=True\n", + ")\n", + "\n", + "\n", + " The `stream=True` parameter is necessary because we want to read responses from the API in real-time without having to wait for the entire response to be received.\n", + "\n", + "3. **Process responses**: Inside an infinite loop (`forchunk in stream:`), it reads and processes each chunk of response from the API:\n", + "\n", + " python\n", + "for chunk in stream:\n", + "response += chunk.choices[0].delta.content or ''\n", + "\n", + "\n", + " - `chunk` is a dictionary-like object containing information about the API's response.\n", + " - `choices` is an array of possible completions, with only one choice shown (`[0]`) by default. We're assuming this is the primary completion we want to display.\n", + " - `.delta.content` gives us the actual text response from the API. This could be a full paragraph, sentence, or even just a word.\n", + " - `response += chunk.choices[0].delta.content or ''`: We simply append any remaining text from previous chunks if there was one.\n", + "\n", + "4. **Format and display**: It reformats the response to remove Markdown formatting (``)) and then uses a `display` function to show an updated version of the original question:\n", + "\n", + " python\n", + "response = response.replace(\"\", \"\").replace(\"\", \"\")\n", + "update_display(Markdown(response), display_id=display_handle.display_id)\n", + "\n", + "\n", + "5. **Update display**: After formatting, it updates the display with the latest response.\n", + "\n", + "**Issue concerns**\n", + "\n", + "One potential issue here: `while True` or a similar loop structure should be used instead of an `Infinite` loop for this streamer's functionality.\n", + "\n", + "Also, error handling would be necessary if we wanted more control over any possible errors while streaming results from API requests." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "stream_llama(question)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_assignments/scrape_website.py b/week1/community-contributions/week1_assignments/scrape_website.py new file mode 100644 index 0000000..d040e22 --- /dev/null +++ b/week1/community-contributions/week1_assignments/scrape_website.py @@ -0,0 +1,15 @@ +from bs4 import BeautifulSoup +import requests + + +class ScrapeWebsite: + + def __init__(self, url, headers): + """ Scraping Website which provides title and content""" + self.url = url + response = requests.get(self.url, headers=headers) + soup = BeautifulSoup(response.content, 'html.parser') + self.title = soup.title.string if soup.title else "No title found" + for irrelevant in soup.body(["script", "style", "img", "input"]): + irrelevant.decompose() + self.text = soup.body.get_text(separator="\n", strip=True) \ No newline at end of file diff --git a/week1/community-contributions/week1_assignments/text_summary_ollama.ipynb b/week1/community-contributions/week1_assignments/text_summary_ollama.ipynb new file mode 100644 index 0000000..d7a5b3b --- /dev/null +++ b/week1/community-contributions/week1_assignments/text_summary_ollama.ipynb @@ -0,0 +1,186 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "4e2a9393-7767-488e-a8bf-27c12dca35bd", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI \n", + "from scrape_website import ScrapeWebsite" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29ddd15d-a3c5-4f4e-a678-873f56162724", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants\n", + "MODEL = \"llama3.2\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "42c8a8c2", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an analyst that analyses the content of the website \\\n", + " provides summary and ignore text related to navigation. Respond in markdown.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "51e86dd1", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; Please provide short summary in Markdown. Please include news and \\\n", + " announcements\"\n", + " user_prompt+=website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b69d7238", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\":\"system\", \"content\": system_prompt},\n", + " {\"role\":\"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a56e99ea", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9b4061d0", + "metadata": {}, + "outputs": [], + "source": [ + "def summarise(url):\n", + " website = ScrapeWebsite(url, headers)\n", + " ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + " response = ollama_via_openai.chat.completions.create(\n", + " model=MODEL,\n", + " messages=messages_for(website)\n", + " )\n", + "\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "65f96545", + "metadata": {}, + "outputs": [], + "source": [ + "def display_summary(url):\n", + " summary = summarise(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "23057e00-b6fc-4678-93a9-6b31cb704bff", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Generative AI has numerous business applications across various industries. Here are some examples:\n", + "\n", + "1. **Marketing and Advertising**: Generative AI can create personalized product recommendations, generate targeted advertisements, and develop new marketing campaigns.\n", + "2. **Content Creation**: AI-powered tools can assist in content creation, such as writing articles, generating social media posts, and creating videos, podcasts, and music.\n", + "3. **Product Design and Development**: Generative AI can aid in designing products, such as 3D modeling, prototyping, and testing product feasibility.\n", + "4. **Customer Service Chatbots**: AI-powered chatbots can provide personalized customer service, answering common queries, and helping resolve issues faster.\n", + "5. **Language Translation**: Generative AI can translate languages in real-time, enabling businesses to communicate with global customers more effectively.\n", + "6. **Data Analysis and Visualization**: AI can analyze large datasets, identify patterns, and create insights, making it easier for businesses to make informed decisions.\n", + "7. **Cybersecurity Threat Detection**: Generative AI-powered systems can detect and respond to cyber threats more efficiently, reducing the risk of data breaches and attacks.\n", + "8. **Supply Chain Optimization**: AI can optimize supply chain operations, predict demand, and identify opportunities for improvement, leading to increased efficiency and reduced costs.\n", + "9. **Network Security**: Generative AI can analyze network traffic patterns, detect anomalies, and prevent cyber-attacks.\n", + "10. **Finance and Banking**: AI-powered systems can detect financial fraud, predict customer creditworthiness, and generate credit reports.\n", + "\n", + "**Industry-specific applications:**\n", + "\n", + "1. **Healthcare**: AI can help with medical diagnosis, patient data analysis, and personalized medicine.\n", + "2. **Manufacturing**: Generative AI can create optimized production schedules, predict equipment failures, and improve product quality.\n", + "3. **Education**: AI-powered tools can develop personalized learning plans, automate grading, and provide educational resources.\n", + "4. **Real Estate**: AI can help with property valuations, identify market trends, and analyze potential clients' needs.\n", + "\n", + "**Business benefits:**\n", + "\n", + "1. **Increased efficiency**: Automating mundane tasks frees up human resources for more strategic work.\n", + "2. **Improved accuracy**: Generative AI reduces the likelihood of human error in decision-making and task execution.\n", + "3. **Enhanced customer experience**: Personalized experiences are created through data-driven insights.\n", + "4. **Competitive advantage**: Companies using AI can differentiate themselves from competitors by offering innovative services and products.\n", + "\n", + "As Generative AI continues to evolve, we can expect even more exciting applications across various industries, leading to increased efficiency, accuracy, and improved competitiveness for businesses worldwide.\n" + ] + } + ], + "source": [ + "display_summary(\"https://www.firstpost.com/world/united-states/\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6de38216-6d1c-48c4-877b-86d403f4e0f8", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_assignments/text_summary_openai_gpt_5mini.ipynb b/week1/community-contributions/week1_assignments/text_summary_openai_gpt_5mini.ipynb new file mode 100644 index 0000000..ab6c1a4 --- /dev/null +++ b/week1/community-contributions/week1_assignments/text_summary_openai_gpt_5mini.ipynb @@ -0,0 +1,265 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "1e45263e", + "metadata": {}, + "source": [ + "# Web Data Extraction and Summarization using openAI Latest model gpt-5-mini" + ] + }, + { + "cell_type": "markdown", + "id": "df155151", + "metadata": {}, + "source": [ + "#### Import Libraries" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "588f8e43", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display\n", + "from openai import OpenAI \n", + "from scrape_website import ScrapeWebsite" + ] + }, + { + "cell_type": "markdown", + "id": "b5925769", + "metadata": {}, + "source": [ + "#### load api key" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "6cca85ec", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')" + ] + }, + { + "cell_type": "markdown", + "id": "56703f80", + "metadata": {}, + "source": [ + "#### ScrapWebsite using BeautifulSoup" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "3d60c909", + "metadata": {}, + "outputs": [], + "source": [ + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}" + ] + }, + { + "cell_type": "markdown", + "id": "a8b73c27", + "metadata": {}, + "source": [ + "#### System Prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "4a0c3bda", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an analyst that analyses the content of the website \\\n", + " provides summary and ignore text related to navigation. Respond in markdown.\"" + ] + }, + { + "cell_type": "markdown", + "id": "9117963b", + "metadata": {}, + "source": [ + "#### User Prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "ab164d55", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt_for(website):\n", + " user_prompt = f\"You are looking at a website titled {website.title}\"\n", + " user_prompt += \"\\nThe contents of this website is as follows; Please provide short summary in Markdown. Please include news and \\\n", + " announcements\"\n", + " user_prompt+=website.text\n", + " return user_prompt" + ] + }, + { + "cell_type": "markdown", + "id": "de7423fb", + "metadata": {}, + "source": [ + "#### Format messages in openAI standard" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "47c82247", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(website):\n", + " return [\n", + " {\"role\":\"system\", \"content\": system_prompt},\n", + " {\"role\":\"user\", \"content\": user_prompt_for(website)}\n", + " ]" + ] + }, + { + "cell_type": "markdown", + "id": "6e9bb6e1", + "metadata": {}, + "source": [ + "#### Summarise the content in website using openAI latest model gpt-5-mini" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "068d6bb2", + "metadata": {}, + "outputs": [], + "source": [ + "def summarise(url):\n", + " website = ScrapeWebsite(url, headers)\n", + " openai = OpenAI()\n", + " response = openai.chat.completions.create(model=\"gpt-5-mini\", messages=messages_for(website))\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "markdown", + "id": "7e6e9da6", + "metadata": {}, + "source": [ + "#### Show summary as Markdown" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "cd86c2ca", + "metadata": {}, + "outputs": [], + "source": [ + "def display_summary(url):\n", + " summary = summarise(url)\n", + " display(Markdown(summary))" + ] + }, + { + "cell_type": "markdown", + "id": "ed5e50d2", + "metadata": {}, + "source": [ + "#### Output" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "74a056b1", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "# Summary — United States Of America | Firstpost (Live/Latest)\n", + "\n", + "Site focus: Live updates and rundowns of US and world news with emphasis on politics, justice, economy, national security, and breaking incidents. Coverage mixes headlines, investigations, opinion and special features/web stories.\n", + "\n", + "## Major news (headlines)\n", + "- Police shooting near CDC/Emory in Atlanta: a suspected shooter and a police officer were killed after reports of an active shooter near the CDC and Emory University campuses. \n", + "- Death of astronaut Jim Lovell (97): Apollo 13 commander and former Navy pilot died in a Chicago suburb. \n", + "- Stephen Miran named to Fed Board (short-term): Trump appointed economist Stephen Miran to the Federal Reserve Board through Jan 2026; noted for support of tariffs and rate cuts. \n", + "- Trump fires labour statistics chief: President Trump sacked the official overseeing labor data hours after a weak jobs report. \n", + "- House panel subpoenas Clintons over Epstein: congressional subpoenas seek documents in relation to Jeffrey Epstein amid pressure on the administration over Epstein files. \n", + "- Ghislaine Maxwell moved to lower-security prison in Texas amid scrutiny of Epstein files and government handling. \n", + "- FBI/administration tension on Epstein Files: Trump said he would “release everything” after reports the FBI redacted names from the Epstein Files. \n", + "- Probe launched into attorney who investigated Trump cases: US officials began a probe targeting Special Counsel Jack Smith. \n", + "- NTSB finds technical issues in Army helicopter crash: investigation into crash that killed 67 people identified technical problems. \n", + "- Trump unveils modified reciprocal tariffs: new executive order introduced modified tariffs on multiple countries; effective date possibly as late as Oct 5. \n", + "- Trump-EU trade deal announced: reported pact imposing a 15% tariff on most EU goods, with large energy and investment components but unresolved issues remain. \n", + "- Federal Reserve holds rates steady: Fed kept rates unchanged for a fifth meeting, despite political pressure from Trump. \n", + "- White House remodel plan: Trump pushing to build a reported $200 million ballroom at the presidential residence, funded by Trump/donors per WH. \n", + "- US citizenship test format under review: Trump administration considers reverting to the 2020 naturalisation test format, citing concerns the current test is too easy. \n", + "- American Airlines incident in Denver: passengers evacuated after a Boeing plane caught fire (tire/maintenance issue) before takeoff. \n", + "- John Bolton criticizes Tulsi Gabbard: former NSA lambastes Gabbard’s report on Obama as exaggerated and lacking substance. \n", + "- Ohio solicitor general Mathura Sridharan trolled: Indian-origin jurist faced racist online backlash after appointment; Ohio AG responded strongly.\n", + "\n", + "## Announcements, features & recurring elements\n", + "- Web stories and quick-read lists: travel/animals/safety themed pieces (e.g., “10 airport codes”, “10 animals that are naturally blue”, World Tiger Day lists). \n", + "- Regular sections and shows highlighted in coverage: Firstpost America, Firstpost Africa, First Sports, Vantage, Fast and Factual, Between The Lines, Flashback, Live TV. \n", + "- Events and special coverage teased: Raisina Dialogue, Champions Trophy, Delhi Elections 2025, Budget 2025, US Elections 2024, Firstpost Defence Summit. \n", + "- Trending topics emphasized: Donald Trump, Narendra Modi, Elon Musk, United States, Joe Biden. \n", + "- Quick-links / network: cross-promotion of other Network18 properties (News18, Moneycontrol, CNBC TV18, Forbes India).\n", + "\n", + "## Tone and emphasis\n", + "- Heavy focus on US politics, Trump administration actions and controversies (Epstein Files, tariffs, personnel changes), justice probes, national security incidents, and major breaking events.\n", + "- Mix of investigative/legal reporting, immediate breaking news, and light/web-story listicles.\n", + "\n", + "If you want, I can produce a one-page brief of just the Trump-related items, a timeline of the Epstein/Clinton/Subpoena coverage, or extract all headlines with publication order." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "display_summary(\"https://www.firstpost.com/world/united-states/\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_exercise_FreeTier.ipynb b/week1/community-contributions/week1_exercise_FreeTier.ipynb new file mode 100644 index 0000000..1fbfcab --- /dev/null +++ b/week1/community-contributions/week1_exercise_FreeTier.ipynb @@ -0,0 +1,188 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# This is a fork of end of week 1 exercise - with only free-tier/local endpoints\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!\n", + "\n", + "- For free Github token, visit https://github.com/settings/tokens. The cool thing about Github models is you can try out different models with various endpoints. We will use OpenAI model as intended for this exercise. You can check out marketplace at https://github.com/marketplace/models to see all the available ones.\n", + "- Don't forget to save this token in .env file as GITHUB_API_KEY = 'your-key'!\n", + "- Also, please run ollama run llama3.2 in your terminal of !ollama run llama3.2 as a cell code to install the model if you haven't." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1070317-3ed9-4659-abe3-828943230e03", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "from openai import OpenAI\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "156f046f-3770-4c39-b576-ec9d2cb42525", + "metadata": {}, + "outputs": [], + "source": [ + "# load environment variables\n", + "load_dotenv(override=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'\n", + "\n", + "github_endpoint = \"https://models.github.ai/inference\"\n", + "ollama_endpoint = \"http://localhost:11434/v1\"\n", + "\n", + "GITHUB_API_KEY = os.getenv('GITHUB_TOKEN')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environments\n", + "openai = OpenAI(\n", + " base_url = github_endpoint,\n", + " api_key = GITHUB_API_KEY\n", + ")\n", + "\n", + "ollama = OpenAI(\n", + " base_url = ollama_endpoint,\n", + " api_key = \"llama\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "question = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "33ba4488-dc88-4caf-85d9-8bdb23a60de2", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer\n", + "def stream_message(prompt):\n", + " stream = openai.chat.completions.create(\n", + " model = MODEL_GPT,\n", + " messages = [{\"role\":\"user\", \"content\": prompt}],\n", + " stream=True,\n", + " )\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " if chunk.choices:\n", + " response += chunk.choices[0].delta.content or ''\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "43ba38d5-f62a-4f8d-ba66-55425c8b0d64", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "stream_message(question)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "response = ollama.chat.completions.create(\n", + " model = MODEL_LLAMA,\n", + " messages = [{\"role\":\"user\", \"content\": question}]\n", + ")\n", + "response2 = response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e2a85c12-4a8d-4539-af3f-a4d76375105b", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "display(Markdown(response2))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8c6f78a8-b3de-4327-9148-6eb385c23af0", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "agents", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.12.3" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_exercise_gpt_llama_teachers.ipynb b/week1/community-contributions/week1_exercise_gpt_llama_teachers.ipynb new file mode 100644 index 0000000..9b122be --- /dev/null +++ b/week1/community-contributions/week1_exercise_gpt_llama_teachers.ipynb @@ -0,0 +1,202 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "51d1bbb7-d56a-4483-935f-480f8e22546f", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if api_key and api_key.startswith('sk-proj-') and len(api_key)>10:\n", + " print(\"API key looks good so far\")\n", + "else:\n", + " print(\"There might be a problem with your API key? Please visit the troubleshooting notebook!\")\n", + "\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bd61eb75-be6a-46d6-8aeb-84c1eeeac04f", + "metadata": {}, + "outputs": [], + "source": [ + "# read the user question\n", + "def user_question_reader() -> str:\n", + " input_text = \"Hello! I’m your AI Teacher, ready to help you explore any topic you’re curious about.\\n\"\n", + " input_text +=\"I have access to a vast amount of knowledge and will do my best to explain things clearly, no matter your experience level.\\n\\n\"\n", + " input_text +=\"What would you like to learn about today?\\n\"\n", + "\n", + " question = input(input_text)\n", + "\n", + " return question;" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6bd9b592-e398-4637-9188-bfdf8dd6bf75", + "metadata": {}, + "outputs": [], + "source": [ + "# generate the user prompt\n", + "def user_prompt_generator() -> str:\n", + " question = user_question_reader()\n", + " user_prompt = f\"I need you to answer to this question: {question}.\\n\"\n", + " user_prompt += \" Take into account that I dont have prior knowledge about my question \\\n", + " so I want the answer as complete as possible. Also please provide it in markdown\"\n", + "\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9ffda047-fec3-4d9e-97b0-46f428ac9313", + "metadata": {}, + "outputs": [], + "source": [ + "# define the system prompt\n", + "system_prompt = \"Your job it's to be a teacher. You have access to all the knowledge \\\n", + " in the internet. You will be thankful to any question given to you and \\\n", + " will try to answer it the best you can. Your students might know little to nothing \\\n", + " about what they ask and make mistakes so you will have to think about the meaning of their question \\\n", + " before and provide an answer according to the meaning behind it.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# define gpt-4o-mini function to answer, with streaming\n", + "def gpt_teacher():\n", + " stream = openai.chat.completions.create(\n", + " model = MODEL_GPT,\n", + " messages= [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\" : user_prompt_generator()}\n", + " ],\n", + " stream=True\n", + " )\n", + "\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + "\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# define Llama 3.2 function to answer\n", + "def llama_teacher():\n", + " response = ollama.chat(\n", + " model = MODEL_LLAMA,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\":system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt_generator()}\n", + " ]\n", + " )\n", + "\n", + " return display(Markdown(response['message']['content']))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "20d963d4-f4ce-4979-b8c7-0db6ebcec96c", + "metadata": {}, + "outputs": [], + "source": [ + "# try the gpt teacher\n", + "gpt_teacher()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d62ca06f-c808-43ee-9ecd-5a704ffcd5c1", + "metadata": {}, + "outputs": [], + "source": [ + "#try the ollama teacher\n", + "llama_teacher()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_exercise_tutor_by_abrar.ipynb b/week1/community-contributions/week1_exercise_tutor_by_abrar.ipynb new file mode 100644 index 0000000..f648fb5 --- /dev/null +++ b/week1/community-contributions/week1_exercise_tutor_by_abrar.ipynb @@ -0,0 +1,209 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "a7cb82f0-bcf2-4fca-84e4-67144594ff2e", + "metadata": {}, + "source": [ + "End of Week 1 Exercise\n", + "\n", + "This notebook demonstrates how to interact with large language models using both OpenAI and Ollama APIs. Based on the user's input and selected model, the notebook routes the message to the corresponding backend and returns the generated response. It's a simple yet flexible interface to explore and compare model behavior across different providers." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eea15c09-c949-4f30-a23b-02130305ff00", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import os\n", + "import requests\n", + "from dotenv import load_dotenv\n", + "\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ac06d3f8-215c-4474-a16b-f5a9980f18b5", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0a768729-d38e-489f-b572-2af356e6ae78", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv(\"OPENAI_API_KEY\")\n", + "\n", + "# set up clients\n", + "openai = OpenAI()\n", + "ollama_url = \"http://localhost:11434/api/chat\"\n", + "ollama_headers = {\"Content-Type\": \"application/json\"}\n", + "# ollama = OpenAI(base_url=\"http://localhost:11434/v1\" , api_key=\"ollama\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f9572d7b-f51c-47c3-9651-56a4106d6f19", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question; type over this to ask something new\n", + "\n", + "default_question = \"\"\"\n", + "Please explain what this code does and why:\n", + "yield from {word for s in sentences for word in s.split()}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "10dabd26-a22f-4f9f-9046-6e76b1c9e5e4", + "metadata": {}, + "outputs": [], + "source": [ + "# Here is the System Prompt\n", + "\n", + "system_prompt = \"\"\"\n", + "You are an expert tutor who explains complex topics in simple, clear, and engaging ways. Tailor your teaching style to the \n", + "learner’s level of knowledge and preferred learning pace. Encourage critical thinking, provide examples, \n", + "and ask occasional questions to check understanding. Avoid giving direct answers when guiding \n", + "problem-solving — instead, offer hints or break the problem into steps. Be friendly, patient, and always supportive. \n", + "Adapt your explanations based on feedback or confusion. When asked for code, equations, or definitions, \n", + "provide them in a structured, easy-to-understand format.\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "83840afa-6445-404a-b922-2acadf228ade", + "metadata": {}, + "outputs": [], + "source": [ + "# Here is the System Prompt\n", + "user_prompt = \"Please give a detailed explanation to the following question: \" " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "570917cc-1f4d-4ca7-b846-f8952aa6c4a0", + "metadata": {}, + "outputs": [], + "source": [ + "# Get User prompt\n", + "\n", + "def get_user_prompt(question=None):\n", + " if not question:\n", + " print(f\"As the given question is empty. So I'm going to ask a default question, which is {default_question}\" )\n", + " question = default_question\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt + question}\n", + " ]\n", + " return messages" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aed48452-3029-482e-8175-e6acfce5b08d", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "def get_answer_from_ollama(question=None):\n", + " messages = get_user_prompt(question)\n", + "\n", + " data = {\n", + " \"model\": MODEL_LLAMA,\n", + " \"messages\": messages,\n", + " \"stream\": False\n", + " }\n", + "\n", + " response = requests.post(ollama_url, headers=ollama_headers, json=data)\n", + " display(Markdown(response.json()['message']['content']))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "374b680a-7994-4636-8939-ab591314f8d6", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "\n", + "def get_answer_from_openai(question=None):\n", + " messages = get_user_prompt(question)\n", + " stream = openai.chat.completions.create(model=MODEL_GPT, messages=messages,stream=True)\n", + " \n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "69406774-33f1-4c67-a8ef-af92567f29a7", + "metadata": {}, + "outputs": [], + "source": [ + "# Ask a question and get the user's response\n", + "answer = input(\"What's the question you want to ask?\")\n", + "\n", + "model_choice = input(\"Please choose a model to use (GPT or Llama): \")\n", + "\n", + "if model_choice.lower() == \"gpt\":\n", + " get_answer_from_openai(answer)\n", + "elif model_choice.lower() == \"llama\":\n", + " get_answer_from_ollama(answer)\n", + "else:\n", + " print(\"Choose the correct model name\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1_tennis.ipynb b/week1/community-contributions/week1_tennis.ipynb new file mode 100644 index 0000000..9c36a80 --- /dev/null +++ b/week1/community-contributions/week1_tennis.ipynb @@ -0,0 +1,154 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "fe12c203-e6a6-452c-a655-afb8a03a4ff5", + "metadata": {}, + "source": [ + "# End of week 1 exercise\n", + "\n", + "To demonstrate your familiarity with OpenAI API, and also Ollama, build a tool that takes a technical question, \n", + "and responds with an explanation. This is a tool that you will be able to use yourself during the course!" + ] + }, + { + "cell_type": "markdown", + "id": "0ea775a9-12c7-4a63-a676-d7bd0cdb100c", + "metadata": {}, + "source": [ + "# imports\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a456906-915a-4bfd-bb9d-57e505c5093f", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8d7923c-5f28-4c30-8556-342d7c8497c1", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "load_dotenv(override=True)\n", + "api_key = os.getenv('OPENAI_API_KEY')\n", + "\n", + "if not api_key:\n", + " print(\"No API key was found!\")\n", + "else:\n", + " print(\"API key found and looks good so far!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f0d0137-52b0-47a8-81a8-11a90a010798", + "metadata": {}, + "outputs": [], + "source": [ + "# here is the question\n", + "question = \"\"\"\n", + "Please explain why do tennis players often use topspin on their forehand shots, and what advantages does it provide?\n", + "\"\"\" " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "967aac6b-9f9c-4def-8659-d9382b0c59e4", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are a helpful tennis coach who answers questions about tennis rules, techniques, strategies, training, and equipment.\"\n", + "user_prompt = \"Please give a detailed explanation to the following question: \" + question" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7936b5af-e912-4e0e-b43e-87673c4857cf", + "metadata": {}, + "outputs": [], + "source": [ + "messages = [\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "60ce7000-a4a5-4cce-a261-e75ef45063b4", + "metadata": {}, + "outputs": [], + "source": [ + "# Get gpt-4o-mini to answer, with streaming\n", + "openai = OpenAI()\n", + "stream = openai.chat.completions.create(model=MODEL_GPT, messages=messages, stream=True)\n", + "response = \"\"\n", + "display_handle = display(Markdown(\"\"), display_id=True)\n", + "for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8f7c8ea8-4082-4ad0-8751-3301adcf6538", + "metadata": {}, + "outputs": [], + "source": [ + "# Get Llama 3.2 to answer\n", + "response = ollama.chat(model=MODEL_LLAMA, messages=messages)\n", + "result = response['message']['content']\n", + "display(Markdown(result))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29e9cdd3-5adc-4428-9758-f761dc91783a", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week1/community-contributions/week1day1.ipynb b/week1/community-contributions/week1day1.ipynb new file mode 100644 index 0000000..66bdd23 --- /dev/null +++ b/week1/community-contributions/week1day1.ipynb @@ -0,0 +1,23 @@ +import os +import requests +from dotenv import load_dotenv +from bs4 import BeautifulSoup +from IPython.display import Markdown, display +from openai import OpenAI + +load_dotenv(override=True) +api_key = os.getenv('OPENAI_API_KEY') + +openai = OpenAI() + +client = OpenAI(api_key=os.getenv("OPENAI_API_key")) + +system_prompt = "You are an student. Always maintain a polite and professional tone." + +user_prompt = "Write a 500-word essay on the impact of social media on modern society, including the benefits and drawbacks." + +messages =[ + {"role": "system", "content": system_prompt}, + {"role": "user", "content": user_prompt} +] +response = openai.chat.completions.create(m diff --git a/week1/community-contributions/youtube_video_summarize.ipynb b/week1/community-contributions/youtube_video_summarize.ipynb new file mode 100644 index 0000000..c96714f --- /dev/null +++ b/week1/community-contributions/youtube_video_summarize.ipynb @@ -0,0 +1,216 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 36, + "id": "8ca2e60d-17c0-40fc-91c6-c16915b39c06", + "metadata": {}, + "outputs": [], + "source": [ + "import re, html, json\n", + "import requests\n", + "from urllib.error import HTTPError\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display\n", + "from youtube_transcript_api import YouTubeTranscriptApi, NoTranscriptFound, TranscriptsDisabled, VideoUnavailable\n", + "\n", + "OLLAMA_API = \"http://localhost:11434/api/chat\"\n", + "HEADERS = {\"Content-Type\": \"application/json\"}\n", + "MODEL = \"llama3.2\"\n", + "api_key='ollama'\n", + "\n", + "def yt_title_desc_transcript(url: str, lang=\"en\"):\n", + " \"\"\"\n", + " Returns {\"title\": str|None, \"description\": str|None, \"transcript\": str|None}.\n", + " - Title via oEmbed (no API key).\n", + " - Description scraped from the watch page (shortDescription).\n", + " - Transcript via youtube-transcript-api, gracefully handling 400/disabled.\n", + " \"\"\"\n", + " # --- extract 11-char video id ---\n", + " m = re.search(r\"(?:v=|/)([0-9A-Za-z_-]{11})|^([0-9A-Za-z_-]{11})$\", url)\n", + " vid = (m.group(1) or m.group(2)) if m else None\n", + " if not vid:\n", + " return {\"title\": None, \"description\": None, \"transcript\": None}\n", + "\n", + " # --- title via oEmbed (very robust) ---\n", + " title = None\n", + " try:\n", + " r = requests.get(\"https://www.youtube.com/oembed\",\n", + " params={\"url\": f\"https://www.youtube.com/watch?v={vid}\", \"format\": \"json\"},\n", + " timeout=10)\n", + " if r.ok:\n", + " title = r.json().get(\"title\")\n", + " except Exception:\n", + " pass\n", + "\n", + " # --- description from watch page (shortDescription in initial JSON) ---\n", + " description = None\n", + " try:\n", + " page = requests.get(f\"https://www.youtube.com/watch?v={vid}\", timeout=10).text\n", + " # Look for ytInitialPlayerResponse JSON\n", + " jmatch = re.search(r\"ytInitialPlayerResponse\\s*=\\s*({.*?});\", page, re.DOTALL)\n", + " if jmatch:\n", + " data = json.loads(jmatch.group(1))\n", + " desc = data.get(\"videoDetails\", {}).get(\"shortDescription\")\n", + " if desc:\n", + " description = html.unescape(desc)\n", + " except Exception:\n", + " pass\n", + "\n", + " # --- transcript (handle 400 cleanly) ---\n", + " transcript_text = None\n", + " try:\n", + " items = YouTubeTranscriptApi.get_transcript(vid, languages=[lang])\n", + " transcript_text = \" \".join(ch[\"text\"].strip() for ch in items if ch.get(\"text\"))\n", + " except (NoTranscriptFound, TranscriptsDisabled, VideoUnavailable, HTTPError):\n", + " # HTTPError covers the \"HTTP Error 400: Bad Request\" case\n", + " transcript_text = None\n", + " except Exception:\n", + " transcript_text = None\n", + "\n", + " return {\"title\": title, \"description\": description, \"transcript\": transcript_text}\n" + ] + }, + { + "cell_type": "code", + "execution_count": 32, + "id": "ad9be496-4e91-4562-90f3-54d11208da55", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "system_prompt = '''\n", + "You are an assistant that generates detailed yet concise summaries of YouTube videos.\n", + "When the user provides a title and description of a YouTube video, your task is to write a coherent, engaging, and informative summary of around 500 words.\n", + "The summary should:\n", + "\n", + "Capture the main themes and key points the video likely covers.\n", + "\n", + "Expand on the description logically, providing context and flow.\n", + "\n", + "Stay neutral, factual, and clear (no personal opinions).\n", + "\n", + "Be self-contained so it makes sense without needing to watch the video.\n", + "'''" + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "id": "dd4be0bc-df1f-47e0-9e03-9b734117f80a", + "metadata": {}, + "outputs": [], + "source": [ + "def user_prompt(title, description):\n", + " prompt = '''Provide me the YouTube video title and description.\\n\n", + " I will generate a clear, engaging, and concise summary of the video content in around 500 words,\\n\n", + " highlighting the main ideas, key points, and important details.\\n'''\n", + " prompt += f'here is the title : {title} \\n Description : {description} '\n", + " return prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 34, + "id": "46896ad3-db1e-448a-8a03-036b9568c69f", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_youtube(yt_url):\n", + " ollama = OpenAI(base_url='http://localhost:11434/v1', api_key='ollama')\n", + " video_metadata = yt_title_desc_transcript(yt_url)\n", + " stream = ollama.chat.completions.create(\n", + " model=MODEL,\n", + " messages = [\n", + " {\"role\":\"system\", \"content\": system_prompt},\n", + " {\"role\":\"user\", \"content\": user_prompt(video_metadata['title'], video_metadata['description'])}\n", + " ],\n", + " stream=True\n", + " \n", + " )\n", + " response = \"\"\n", + " display_handle = display(Markdown(\"\"), display_id=True)\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " response = response.replace(\"```\",\"\").replace(\"markdown\", \"\")\n", + " update_display(Markdown(response), display_id=display_handle.display_id)" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "id": "b59f8773-c13e-4050-ad3c-b578d07ef5e7", + "metadata": {}, + "outputs": [ + { + "data": { + "text/markdown": [ + "Here is a summary of the YouTube video:\n", + "\n", + "**Monta Re: A Baul-Inspired Tribute to the Mystic Guru Shankaracharya**\n", + "\n", + "The music video for \"Monta Re\" by Amit Trivedi, featuring Swanand Kirkire and Amitabh Bhattacharya, is a soulful tribute to the mystic guru Shankaracharya. Set in the Bengali folk music tradition, this song brings to life the ancient tales of Shankaracharya's spiritual journey.\n", + "\n", + "With elegant lyrics penned by Amitabh Bhattacharya, \"Monta Re\" transports listeners to the banks of the Ganges River, where Shankaracharya wandered in search of wisdom and inner peace. The song's haunting melodies and emotive vocals evoke a sense of longing and introspection, perfectly capturing the mystic guru's spiritual essence.\n", + "\n", + "The music video beautifully illustrates the baul-inspired style, with intricate traditional dance movements performed by a group of energetic dancers. The choreography seamlessly blends elements of Bengal's folk heritage with modern sensibilities, making the song an engaging watch for audience members interested in Indian classical music.\n", + "\n", + "**Music and Lyric Credit:**\n", + "Amit Trivedi handles the music composition, ensuring that the melody complements the song's themes without overpowering them. Amitabh Bhattacharya takes credit for the lyrics, which tell stunning stories of Shankaracharya's spiritual adventures. The song features Swanand Kirkire and Amitabh Bhattacharya as vocalists, further enriching its emotional impact.\n", + "\n", + "**Relevance to Bengali Culture:**\n", + "\"Monta Re\" is a heartwarming tribute to Bengal's rich cultural heritage. Inspired by the baul traditions of the region, this song honors Shankaracharya's life and spiritual journey without diminishing his significance in modern times. By showcasing these folk roots, \"Monta Re\" provides fans with an enriching sensory experience.\n", + "\n", + "You can listen to \"Monta Re\" along with other T-Series music videos released by Amit Trivedi at the links provided below:\n", + "\n", + "- Watch \"Ankahee\"\n", + "- Check out \"Sawaar Loon\"\n", + "- Explore \"Zinda Hoon\"\n", + "\n", + "Follow the official T-SERIES YouTube channel for an ever-growing variety of original music tracks!\n", + "\n", + "By embracing the richness of Bengali folk traditions, \"Monta Re\" embodies a musical reflection of Shankaracharya's extraordinary journey as both spiritual guide and symbol of timeless wisdom." + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "stream_youtube('https://youtu.be/99NUJ1cLbBI?list=RDdJ6_aU6auZc')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "649287ca-aff8-4b59-91b7-731c007e83a7", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/3_chatbots_Converstion/Conversation_Day1.ipynb b/week2/community-contributions/3_chatbots_Converstion/Conversation_Day1.ipynb new file mode 100644 index 0000000..72400c8 --- /dev/null +++ b/week2/community-contributions/3_chatbots_Converstion/Conversation_Day1.ipynb @@ -0,0 +1,385 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2b3a83fe-edf2-45b7-8b76-af2324296ad0", + "metadata": {}, + "source": [ + "### Import API Keys and Establish Connections" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bacb0c55-44ee-4505-a3bc-7aaa3d72b28b", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import ollama\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1767187f-c065-43df-b778-fcd48bd5e48d", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv(\"OPENAI_API_KEY\")\n", + "google_api_key = os.getenv(\"GOOGLE_API_KEY\")\n", + "anthropic_api_key = os.getenv(\"ANTHROPIC_API_KEY\")\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API key exists {openai_api_key[:8]}\")\n", + "else:\n", + " print(f\"OpenAI API key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API key exists {google_api_key[:7]}\")\n", + "else:\n", + " print(f\"Google API key not set\")\n", + "\n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API key exists {openai_api_key[:8]}\")\n", + "else:\n", + " print(f\"Anthropic API key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc7ca3ab-ff7f-4375-bcad-aca49c7f4f4f", + "metadata": {}, + "outputs": [], + "source": [ + "# Initializing API Clients, loading the SDKs\n", + "# An SDK is a library/toolbox (Pre-built functions, classes, utilities) full \n", + "# of everything you need to use someone else's software\n", + " \n", + "openai = OpenAI()\n", + "claude = anthropic.Anthropic()\n", + "ollama_via_openai = OpenAI(base_url='http://localhost:11434/v1', api_key = 'ollama')" + ] + }, + { + "cell_type": "markdown", + "id": "81e01904-5586-4726-ab91-7bdbd6bde6d9", + "metadata": { + "jp-MarkdownHeadingCollapsed": true + }, + "source": [ + "### A Coversation between 3 chatbots" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "843bbb69-ab7d-4b13-b878-65a4275f53ca", + "metadata": {}, + "outputs": [], + "source": [ + "# Conversation between GPT-4o-mini, Claude-3, ang Gemini 2.5 flash\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "claude_model = \"claude-3-haiku-20240307\"\n", + "ollama_model = \"llama3.2\"\n", + "\n", + "gpt_system = \"You are an eternal optimist. You always see the bright side of things and believe even \\\n", + "simple actions have deep purpose. Keep replies under 2 sentences.\"\n", + "\n", + "ollama_system = \"You are a witty skeptic who questions everything. You tend to doubt grand explanations \\\n", + "and prefer clever, sarcastic, or literal answers. Keep replies under 2 sentences.\"\n", + "\n", + "claude_system = \"You are a thoughtful philosopher. You consider all perspectives and enjoy finding \\\n", + "symbolic or existential meaning in simple actions. Keep replies under 2 sentences.\"\n", + "\n", + "\n", + "gpt_messages = [\"Hi! Todays topic for discussion is 'Why did the chicken cross the road?'\"]\n", + "ollama_messages = [\"That's quite the topic. \"]\n", + "claude_messages = [\"Lets begin our discussion.\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1a4da2f5-ff74-4847-aa86-867e89173509", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " \n", + " messages = [{\"role\":\"system\", \"content\":gpt_system}]\n", + " \n", + " for gpt, ollama, claude in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": ollama})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " \n", + " response = openai.chat.completions.create(\n", + " model = gpt_model,\n", + " messages = messages,\n", + " max_tokens = 500\n", + " )\n", + " return response.choices[0].message.content.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5848d83a-f4aa-42ee-b40b-6130da60c890", + "metadata": {}, + "outputs": [], + "source": [ + "def call_ollama():\n", + " messages = [{\"role\":\"system\", \"content\":ollama_system}]\n", + " \n", + " for gpt, ollama_message, claude in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": ollama_message})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " \n", + " messages.append({\"role\":\"user\", \"content\": gpt_messages[-1]})\n", + "\n", + " response = ollama_via_openai.chat.completions.create(\n", + " model = ollama_model,\n", + " messages = messages\n", + " )\n", + " return response.choices[0].message.content.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a50e4f7c-d594-4ed8-a658-2d8b2fde21a0", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " \n", + " messages = []\n", + " \n", + " for gpt, ollama, claude_message in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\":\"user\", \"content\":gpt})\n", + " messages.append({\"role\": \"user\", \"content\": ollama})\n", + " messages.append({\"role\":\"assistant\", \"content\": claude_message})\n", + " \n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": ollama_messages[-1]})\n", + " \n", + " response = claude.messages.create(\n", + " model = claude_model,\n", + " system = claude_system,\n", + " messages = messages,\n", + " max_tokens = 500\n", + " )\n", + " return response.content[0].text.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5c78fcf8-544e-413f-af18-ccb9000515de", + "metadata": {}, + "outputs": [], + "source": [ + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Ollama:\\n{ollama_messages[0]}\\n\")\n", + "print(f\"Claude:\\n{claude_messages[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT: \\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + "\n", + " ollama_next = call_ollama()\n", + " print(f\"Ollama: \\n{ollama_next}\\n\")\n", + " ollama_messages.append(ollama_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude: \\n{claude_next}\\n\")\n", + " claude_messages.append(claude_next)" + ] + }, + { + "cell_type": "markdown", + "id": "8ea7419a-ea8f-42da-a9a1-4bbe5342cecb", + "metadata": { + "jp-MarkdownHeadingCollapsed": true + }, + "source": [ + "### Another Coversation between 3 chatbots" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c279c275-7b95-4587-9cc6-4d32517ec253", + "metadata": {}, + "outputs": [], + "source": [ + "# Conversation between GPT-4o-mini, Claude-3, ang Gemini 2.5 flash\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "claude_model = \"claude-3-haiku-20240307\"\n", + "ollama_model = \"llama3.2\"\n", + "\n", + "gpt_system = \"You are an optimist who believes technology brings people \\\n", + "closer together and improves lives. Defend innovation as a force for human \\\n", + "connection. Keep response under 3 sentences.\"\n", + "\n", + "\n", + "ollama_system = \"You are a skeptic who questions if technology isolates us \\\n", + "and worsens social divides. Highlight its risks and unintended consequences. \\\n", + "Keep response under 3 sentences.\"\n", + "\n", + "\n", + "claude_system = \"You are a philosopher who explores both sides \\\n", + "of technology's impact. Seek a balanced perspective on connection and isolation.\\\n", + "Keep response under 3 sentences.\"\n", + "\n", + "\n", + "\n", + "\n", + "gpt_messages = [\"Our topic of discussion for today will be: 'Is technology making us more connected or more isolated?'\"]\n", + "ollama_messages = [\"A great topic\"]\n", + "claude_messages = [\"Let's begin.\"]\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "44c023a6-f22f-4a64-a718-f75fe4c8233a", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " \n", + " messages = [{\"role\":\"system\", \"content\":gpt_system}]\n", + " \n", + " for gpt, ollama, claude in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": ollama})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " \n", + " response = openai.chat.completions.create(\n", + " model = gpt_model,\n", + " messages = messages,\n", + " max_tokens = 500\n", + " )\n", + " return response.choices[0].message.content.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d29f27a1-457e-4e71-88dc-c55e4a36a27c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_ollama():\n", + " messages = [{\"role\":\"system\", \"content\":ollama_system}]\n", + " \n", + " for gpt, ollama_message, claude in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": ollama_message})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " \n", + " messages.append({\"role\":\"user\", \"content\": gpt_messages[-1]})\n", + "\n", + " response = ollama_via_openai.chat.completions.create(\n", + " model = ollama_model,\n", + " messages = messages\n", + " )\n", + " return response.choices[0].message.content.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "69577edc-4be2-40fc-8eac-1243c30cda26", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " \n", + " messages = []\n", + " \n", + " for gpt, ollama, claude_message in zip(gpt_messages, ollama_messages, claude_messages):\n", + " messages.append({\"role\":\"user\", \"content\":gpt})\n", + " messages.append({\"role\": \"user\", \"content\": ollama})\n", + " messages.append({\"role\":\"assistant\", \"content\": claude_message})\n", + " \n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": ollama_messages[-1]})\n", + " \n", + " response = claude.messages.create(\n", + " model = claude_model,\n", + " system = claude_system,\n", + " messages = messages,\n", + " max_tokens = 500\n", + " )\n", + " return response.content[0].text.strip()\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "acedf2fb-8b20-49be-9a80-24fb3896e2ea", + "metadata": {}, + "outputs": [], + "source": [ + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Ollama:\\n{ollama_messages[0]}\\n\")\n", + "print(f\"Claude:\\n{claude_messages[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT: \\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + "\n", + " ollama_next = call_ollama()\n", + " print(f\"Ollama: \\n{ollama_next}\\n\")\n", + " ollama_messages.append(ollama_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude: \\n{claude_next}\\n\")\n", + " claude_messages.append(claude_next)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a078943b-7a34-4697-b1f6-16f4b0e7aed6", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/3_chatbots_Converstion/Conversation_Outputs.pdf b/week2/community-contributions/3_chatbots_Converstion/Conversation_Outputs.pdf new file mode 100644 index 0000000..6c8fefa Binary files /dev/null and b/week2/community-contributions/3_chatbots_Converstion/Conversation_Outputs.pdf differ diff --git a/week2/community-contributions/3_chatbots_Converstion/README.md b/week2/community-contributions/3_chatbots_Converstion/README.md new file mode 100644 index 0000000..c9f07e9 --- /dev/null +++ b/week2/community-contributions/3_chatbots_Converstion/README.md @@ -0,0 +1,36 @@ + +# 3 Way Chatbot Conversation +Making the different models from Anthropic, OpenAI and Ollama converse with each other. + +## Contents + +- `Conversation_Day1.ipynb`: The notebook file with all code and explanations for the first day. +- `Conversation_Outputs`: The chatbots conversations for each topic +- `requirements.txt`:For installing the dependencies +- `README.md`: This file. + +## How to Run + +1. Clone this repository. +2. I'm using 'Python 3.11.13' with Jupyter Notebook or JupyterLab. +3. Install dependencies (see below). +4. Open the notebook using Jupyter: + +```bash +jupyter notebook Conversation_Day1.ipynb +``` + +## Dependencies + +Install the required Python libraries using: + +```bash +pip install -r requirements.txt +``` + +--- + +### Author + +Mustafa Kashif + diff --git a/week2/community-contributions/3_chatbots_Converstion/requirements.txt b/week2/community-contributions/3_chatbots_Converstion/requirements.txt new file mode 100644 index 0000000..548bb18 --- /dev/null +++ b/week2/community-contributions/3_chatbots_Converstion/requirements.txt @@ -0,0 +1,6 @@ +IPython +anthropic +dotenv +ollama +openai +os \ No newline at end of file diff --git a/week2/community-contributions/AddingGeminiToDropdown.ipynb b/week2/community-contributions/AddingGeminiToDropdown.ipynb new file mode 100644 index 0000000..656a542 --- /dev/null +++ b/week2/community-contributions/AddingGeminiToDropdown.ipynb @@ -0,0 +1,223 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "057bc09f-a682-4b72-97ed-c69ddef3f03e", + "metadata": {}, + "source": [ + "# Gemini to Dropdown" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d66eb067-7bae-4145-b613-6da2f40fbf27", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import google.generativeai as genai\n", + "import anthropic" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e36f8a93-8a65-48f2-bcad-7c47dd72ef3a", + "metadata": {}, + "outputs": [], + "source": [ + "import gradio as gr " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8a5ec1b0-f5b4-46d2-abb0-b28b73cc4d28", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26d0099c-890f-4358-8c1d-7a708abcb105", + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6606bfdb-964e-4d6f-b2a1-5017b99aa23d", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful assistant\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e0cfb96a-2dbe-4228-8efb-75947dbc3228", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_gpt(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " stream = openai.chat.completions.create(\n", + " model='gpt-4o-mini',\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9008a15d-0ee8-44e0-b123-225e7148113e", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_claude(prompt):\n", + " result = claude.messages.stream(\n", + " model=\"claude-3-haiku-20240307\",\n", + " max_tokens=1000,\n", + " temperature=0.7,\n", + " system=system_message,\n", + " messages=[\n", + " {\"role\": \"user\", \"content\": prompt},\n", + " ],\n", + " )\n", + " response = \"\"\n", + " with result as stream:\n", + " for text in stream.text_stream:\n", + " response += text or \"\"\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "378ad12e-6645-4647-807c-00995e360268", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_gemini(prompt):\n", + " gemini = genai.GenerativeModel(\n", + " model_name=\"gemini-2.0-flash\",\n", + " system_instruction=system_message\n", + " )\n", + " \n", + " stream = gemini.generate_content(prompt, stream=True)\n", + " \n", + " result = \"\"\n", + " for chunk in stream:\n", + " try:\n", + " part = chunk.text\n", + " if part:\n", + " result += part\n", + " yield result \n", + " except Exception as e:\n", + " print(\"Chunk error:\", e)\n", + " \n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fd50e143-eead-49b1-8ea3-b440becd4bc9", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_model(prompt, model):\n", + " if model==\"GPT\":\n", + " result = stream_gpt(prompt)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(prompt)\n", + " elif model==\"Gemini\":\n", + " result = stream_gemini(prompt)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " yield from result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c7fc9cb4-fbb8-4301-86a6-96c90f67eb3b", + "metadata": {}, + "outputs": [], + "source": [ + "view = gr.Interface(\n", + " fn=stream_model,\n", + " inputs=[gr.Textbox(label=\"Your message:\"), gr.Dropdown([\"GPT\", \"Claude\",\"Gemini\"], label=\"Select model\", value=\"GPT\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/Copilot.ipynb b/week2/community-contributions/Copilot.ipynb new file mode 100644 index 0000000..c32aad0 --- /dev/null +++ b/week2/community-contributions/Copilot.ipynb @@ -0,0 +1,212 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "1877ad68", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "from dotenv import load_dotenv \n", + "import google.generativeai as genai\n", + "from IPython.display import Markdown, display, update_display\n", + "load_dotenv(override=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "008056a2", + "metadata": {}, + "outputs": [], + "source": [ + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f'OpenAi api key exists and its starts with {openai_api_key[:3]}')\n", + "else:\n", + " print(\"OpenAi api key doesn't exist\")\n", + "\n", + "if google_api_key:\n", + " print('Google api key exists')\n", + "else:\n", + " print(\"Google api key doesn't exist\")\n", + "\n", + "OPENAI_MODEL = \"gpt-4o-mini\"\n", + "GOOGLE_MODEL = \"gemini-1.5-flash\"\n", + "\n", + "openai = OpenAI()\n", + "\n", + "genai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5013ed7b", + "metadata": {}, + "outputs": [], + "source": [ + "system_msg = \"\"\"\n", + "You are CodeCopilot, an adaptive AI coding assistant that helps users solve problems in any programming language.\n", + "Always provide correct, runnable, and well-formatted code with clear explanations.\n", + "Adjust your style based on the user’s expertise: for beginners, break concepts down step by step with simple examples and commented code;\n", + "for advanced users, deliver concise, production-ready, optimized solutions with best practices and trade-off insights.\n", + "Ask clarifying questions when requirements are ambiguous, highlight pitfalls and edge cases,\n", + "and act as a collaborative pair programmer or mentor whose goal is to help users learn, build, and ship high-quality code efficiently.\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "35c480a1", + "metadata": {}, + "outputs": [], + "source": [ + "def create_prompt(prompt, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_msg}]\n", + "\n", + " # history is a list of (user_msg, assistant_msg) tuples\n", + " for user_msg, assistant_msg in history:\n", + " if user_msg:\n", + " messages.append({\"role\": \"user\", \"content\": user_msg})\n", + " if assistant_msg:\n", + " messages.append({\"role\": \"assistant\", \"content\": assistant_msg})\n", + "\n", + " # new user prompt\n", + " messages.append({\"role\": \"user\", \"content\": prompt})\n", + " return messages" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5dfbecd0", + "metadata": {}, + "outputs": [], + "source": [ + "def openai_agent(prompt, history):\n", + " openai.api_key = openai_api_key\n", + " messages = create_prompt(prompt, history)\n", + " response = openai.chat.completions.create(\n", + " model=OPENAI_MODEL,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " sent_any = False\n", + " for chunk in response:\n", + " delta = chunk.choices[0].delta\n", + " if delta and delta.content:\n", + " sent_any = True\n", + " yield delta.content\n", + " if not sent_any:\n", + " yield \"(no response)\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "535f7e3d", + "metadata": {}, + "outputs": [], + "source": [ + "def gemini_agent(prompt, history):\n", + " genai.configure(api_key=google_api_key)\n", + "\n", + " # reuse OpenAI-style messages\n", + " messages = create_prompt(prompt, history)\n", + "\n", + " gemini_history = []\n", + " for m in messages:\n", + " # Gemini does NOT support system role\n", + " if m[\"role\"] == \"system\":\n", + " continue\n", + " gemini_history.append({\n", + " \"role\": m[\"role\"],\n", + " \"parts\": [m[\"content\"]]\n", + " })\n", + " prompt_with_system = f\"{system_msg}\\n\\n{prompt}\"\n", + " model = genai.GenerativeModel(GOOGLE_MODEL)\n", + " chat = model.start_chat(history=gemini_history)\n", + "\n", + " response = chat.send_message(prompt_with_system, stream=True)\n", + " for chunk in response:\n", + " if chunk and getattr(chunk, \"text\", None):\n", + " yield chunk.text\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "21f61ff0", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_agent(prompt, history, modelType):\n", + " if modelType == \"OpenAI\":\n", + " for token in openai_agent(prompt, history):\n", + " yield token\n", + " else:\n", + " for token in gemini_agent(prompt, history):\n", + " yield token\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "56686c1d", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_fn(prompt, history, model):\n", + " assistant_response = \"\"\n", + " for token in chat_agent(prompt, history, model):\n", + " assistant_response += token\n", + " yield assistant_response \n", + "\n", + "# -------------------------------------------------------------------\n", + "# UI\n", + "# -------------------------------------------------------------------\n", + "with gr.Blocks() as demo:\n", + " model_choice = gr.Radio([\"OpenAI\", \"Gemini\"], value=\"OpenAI\", label=\"Model\")\n", + "\n", + " chat_ui = gr.ChatInterface(\n", + " fn=chat_fn,\n", + " additional_inputs=[model_choice],\n", + " title=\"CodeCopilot\",\n", + " description=\"An adaptive AI coding assistant that helps developers build and ship high-quality code.\"\n", + " )\n", + "\n", + "demo.launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/Dental_Office_Chatbot.ipynb b/week2/community-contributions/Dental_Office_Chatbot.ipynb new file mode 100644 index 0000000..0de52dc --- /dev/null +++ b/week2/community-contributions/Dental_Office_Chatbot.ipynb @@ -0,0 +1,252 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "7f161eb8-7973-4c4e-ac9a-c85979a9b7a8", + "metadata": {}, + "source": [ + "

Chatbot for Dental Office

" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "id": "c9883c8a-1ea8-406c-81e0-18fbf6c5d8b2", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import requests\n", + "from bs4 import BeautifulSoup" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "72fe1eb9-a8b1-48f1-9c4f-eefb9714d8fb", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk\n", + "Anthropic API Key exists and begins sk\n", + "Google API Key exists and begins AI\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:2]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:2]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:2]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "3bf7a992-ded3-42b3-b207-1a5077804466", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize\n", + "\n", + "openai = OpenAI()\n", + "MODEL = 'gpt-4o-mini'" + ] + }, + { + "cell_type": "code", + "execution_count": 42, + "id": "4ecd96d8-2e87-4a39-9784-178aa4424667", + "metadata": {}, + "outputs": [], + "source": [ + "#scrape website\n", + "\n", + "def get_website_text(url):\n", + " try:\n", + " response = requests.get(url)\n", + " response.raise_for_status()\n", + " \n", + " soup = BeautifulSoup(response.text, \"html.parser\")\n", + " \n", + " # Remove script and style elements\n", + " for tag in soup([\"script\", \"style\", \"noscript\"]):\n", + " tag.decompose()\n", + "\n", + " # Extract visible text\n", + " text = soup.get_text(separator=\"\\n\")\n", + " \n", + " # Clean up whitespace\n", + " lines = [line.strip() for line in text.splitlines()]\n", + " content = \"\\n\".join(line for line in lines if line)\n", + "\n", + " return content[:100000] \n", + " except Exception as e:\n", + " return f\"Error fetching website content: {e}\"" + ] + }, + { + "cell_type": "code", + "execution_count": 43, + "id": "c0fc0537-e0b8-450d-8457-b00669c5df86", + "metadata": {}, + "outputs": [], + "source": [ + "url = \"https://rooseveltislandadvanceddentistry.com/\"\n", + "website_content = get_website_text(url)" + ] + }, + { + "cell_type": "code", + "execution_count": 44, + "id": "25e118d7-5adb-4978-bdac-83c12e119de2", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "\"Roosevelt Island Advanced Dentistry | Pediatric Dentist\\nBusiness Hours:\\nMonday to Saturday 9 AM to 5 PM\\n(212) 752-8722\\n501A Main St, New York, NY 10044, United States\\nRoosevelt Island\\ncontact us\\nBook an appointment\\nOur Team\\nPediatric Procedures\\nPediatric Services\\nPediatric Topics\\nProsthodontics Procedures\\nDental Implants\\nGeneral Dentistry\\nServices\\nCosmetic Dentistry\\nOral Hygiene\\nEmergency\\nOrthodontics\\nOnline Forms\\nTestimonials\\nOur Team\\nPediatric Procedures\\nPediatric Services\\nPediatric Topics\\nProsthodontics Procedures\\nDental Implants\\nGeneral Dentistry\\nServices\\nCosmetic Dentistry\\nOral Hygiene\\nEmergency\\nOrthodontics\\nOnline Forms\\nTestimonials\\nBusiness Hours:\\nMonday to Saturday 9 AM to 5 PM\\n(212) 752-8722\\n501A Main St, New York, NY 10044, United States\\nRoosevelt Island\\ncontact us\\nBook an appointment\\nNEW YORK PEDIATRIC DENTIST\\nThe best and only Pediatric Dentist Specialist in Roosevelt Island, NY.\\nThe best and only Pediatric Dentist Specialist in Roosevelt Island, NY.\\nRoosevelt Island Advanced Dentistry offers comprehensive general dentistry services in Roosevelt Island, NY.\\nRoosevelt Island Advanced Dentistry provides expert prosthodontics in Roosevelt Island, NY.\\nThe best and only Pediatric Dentist Specialist in Roosevelt Island, NY.\\nRoosevelt Island Advanced Dentistry offers comprehensive general dentistry services in Roosevelt Island, NY.\\nRoosevelt Island Advanced Dentistry provides expert prosthodontics in Roosevelt Island, NY.\\nbook an appointment\\nCall :\\n(212) 752-8722\\nleave a review\\nRoosevelt Island Dentistry nyc\\nServices We Provide\\nPediatric Specialty Procedures\\nLearn more\\nOrthodontics\\nLearn more\\nProsthodontics Specialty Procedures\\nLearn more\\nCosmetic Dentistry\\nLearn more\\nDental Implants\\nLearn more\\nGeneral Dentistry\\nLearn more\\nDENTAL ICON(outline)\\nOral Hygiene\\nLearn more\\nSmile Makeover\\nLearn more\\nbook an appointment\\ncontact us\\nOur Team\\nabout us\\ntestimonials\\nDr. skomial\\nProsthodontist-Implant Specialist\\nDr. Skomial is an active member of American Dental Association and American College of Prosthodontists. He is originally from Poland and speaks\\nLearn more\\nDr. Chen\\nBoard Certified Pediatric Dentist\\nDonna Chen, DDS is a compassionate dentist who loves building relationships with her patients. She caters to children’s individualized needs and\\nLearn more\\nDr. Correa\\nBoard Certified Pediatric Dentist\\nDr. Lizeth Correa is a pediatric dentist known for her warm and friendly approach to treating young patients. Dr. Correa places emphasis\\nLearn more\\nDr. Regina Mathai, DDS MPH\\nBoard Eligible Pediatric Dentist\\nDr. Regina, from Hudson Valley, NY, earned a Biology and Spanish degree from Binghamton and a Master's in Public Health from Hofstra.\\nLearn more\\nDr. Buchbinder\\nOrthodontist\\nDr. William Buchbinder (Dr. Billy) is an orthodontic specialist serving patients across the Greater New York area. Combining clinical excellence with a warm, approachable style.\\nLearn more\\nDr. Matthew Lee, DMD\\nGeneral Dentist\\nDr. Matthew Lee, a NYC native, holds a Biology degree and a Master's in Biomedical Sciences from Rutgers, where he also attended dental school.\\nLearn more\\nDr. Jason Tu, DDS\\nGeneral Dentist\\nDr. Jason Tu was born and raised in Vancouver, Canada. He completed his undergraduate studies at UBC and dental school at NYU.\\nLearn more\\nDr. Jeanne Helbig, DMD\\nGeneral Dentist\\nDr. Jeanne Helbig is a general dentist who believes in providing\\ncompassionate and comprehensive care to each of her patients.\\nLearn more\\nWhy we stand out?\\nRoosevelt Island Advanced Dentistry always places patients at the center of our attention, and concentrate on improving their experience with the aid of technologies.\\nGet your services right\\nGet rid of your pain, stress, and enduring with our dental services. It’s a priority to relieve the pain and damage to your mouth in surgeon as much as possible.\\nwhy choose us\\nReasons why we're widely favored\\ncontact us\\nfaqs\\nHome Heart\\nWork With Hearts\\nWe care for your dental health with a great compassion & understanding so that you can have the best smile.\\nPrecise Diagnosis\\nWhen your teeth are decayed or damaged, we use modern tools to detect the area and provide treatment.\\nServe with Smile\\nThe smile never fades on our doctors’ faces as they always want to create an atmosphere that feels comfortable.\\nAnnual Check-ups\\nWe provide annual check-up for dental health conditions and offer many promotions for members of our center.\\nHelp at Hand\\nWe are ready to check for any teeth issue at our center and at any time of the day, even on weekends & holidays.\\nFlexible Installment\\nWe allow periodic installments of purchasing costs in case customers are unable to afford the price of services.\\nFAQs\\nYou may find an answer to your question here.\\nWhy choose a Pediatric Dentist?\\nPediatric Dentistry is the dental specialty recognized by the American Dental Association (ADA), which provides comprehensive dental care for all children, from infancy to adolescence. Pediatric dentists promote the dental health of children as well as serve as educational resources for parents.\\nA Pediatric Dentist is a practitioner who has completed an additional two years of post-doctorate training after dental school. The specialty-focused training includes child psychology, behavioral guidance, preventative techniques, and restorative dentistry. A pediatric dentist also received training in treating handicapped or chronically ill children as well.\\nWhat can I expect during my visit?\\nThe pediatric dentist will review your child’s medical and dental history. They will gently examine your child’s teeth, oral tissues, and jaws. The teeth will be cleaned and polished, followed by the application of a fluoride solution, if necessary.\\nYour pediatric dentist won’t talk just to you about dental health, she will talk to your child with easily understandable words, pictures, and ideas. Your child will be motivated to take responsibility for a healthy smile.\\nDo you treat children with special needs?\\nNo, but we have a 'Refuse a Service' option, and here is how it works. If, during some of the beginning stages of preparation, you see that the service does not satisfy your expectations, you can terminate the contract and not pay the final amount of money. Since we sign a service contract for all the events, the conditions, such as the amounts of money and payment dates, are specified in the signed documents. Therefore, we advise you to read the contract carefully and pay specific attention to the paragraph that describes the mentioned data.\\nWhen should my child's first dentist visit be?\\nThe American Academy of Pediatric Dentistry recommends a child’s first dental visit by their first birthday, or sooner. A dental check-up twice a year is recommended for most children. Some children need more frequent dental visits because of increased risk of tooth decay, unusual growth patterns or poor oral hygiene. Your pediatric dentist will let you know the best appointment schedule for your child.\\nRegular dental visits help your child stay cavity-free. Teeth cleanings remove debris that build up on the teeth, irritate the gums and cause decay. Fluoride treatments renew the fluoride content in the enamel, strengthening teeth and preventing cavities. Hygiene instructions improve your child’s brushing and flossing, leading to cleaner teeth and healthier gums.\\nTooth decay isn’t the only reason for a dental visit. Your pediatric dentist provides an ongoing assessment of changes in your child’s oral health. For example, your child may need additional fluoride, dietary changes, or sealants for ideal dental health. The pediatric dentist may identify orthodontic problems and suggest treatment to guide the teeth as they emerge in the mouth. Speak to your pediatric dentist today.\\nContact Us Today!\\nGet in Touch for Personalized Pediatric Dentistry in Roosevelt Island!\\nAddress\\n501A Main St, New York, NY 10044,\\nUnited States\\nRoosevelt Island\\nphone\\n(212) 752-8722\\nemail\\nmoc.liamg%40cllevollatned\\nbusiness hours\\nMonday to Saturday 9 AM to 5 PM\\nRoosevelt Island Advanced Dentistry\\nRoosevelt Island Advanced Dentistry\\nPhone:\\n(212) 759-3666\\nBusiness Hours:\\nMonday to Saturday 9 AM to 5 PM\\nQuick Links\\nPediatric Specialty Procedures\\nProsthodontics Specialty Procedures\\nGeneral Dentistry\\nLegal Pages\\nPrivacy Policy\\nAccessibility Statement\\n© Created by\\nDearDoc\\nAll Rights Reserved Roosevelt Island Advanced Dentistry.\\nRoosevelt Island Advanced Dentistry\\nRoosevelt Island Advanced Dentistry\\nPhone:\\n(212) 759-3666\\nBusiness Hours:\\nMonday to Saturday 9 AM to 5 PM\\nQuick Links\\nPediatric Specialty Procedures\\nProsthodontics Specialty Procedures\\nGeneral Dentistry\\nLegal Pages\\nPrivacy Policy\\nAccessibility Statement\\n© Created by\\nDearDoc\\nAll Rights Reserved Roosevelt Island Advanced Dentistry.\"" + ] + }, + "execution_count": 44, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "website_content" + ] + }, + { + "cell_type": "code", + "execution_count": 45, + "id": "7c560358-3932-4599-b608-fdfb8d99a784", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = (\n", + " \"You are a helpful assistant for a dental office. Your job is to welcome the patient. Use the information below to answer patient questions \"\n", + " \"about services, office hours, insurance, contact details, doctors and booking appointments.\\n\\n\"\n", + " f\"Website content:\\n{website_content}\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 46, + "id": "b976152b-77c0-4901-bec6-50ac32c8cfbb", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history + [{\"role\": \"user\", \"content\": message}]\n", + " \n", + " print(\"History is:\")\n", + " print(history)\n", + " print(\"And messages is:\")\n", + " print(messages)\n", + "\n", + " stream = openai.chat.completions.create(model=MODEL, messages=messages, stream=True)\n", + "\n", + " response = \"\"\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": 47, + "id": "48e3852e-557c-4a7a-9ada-41cd6bbd6167", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7872\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 47, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "gr.ChatInterface(fn=chat, type=\"messages\").launch()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9f224861-c8a1-40e6-b07a-4d7d4b5d9484", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/Figma_assistance/.github/workflows/update_space.yml b/week2/community-contributions/Figma_assistance/.github/workflows/update_space.yml new file mode 100644 index 0000000..7e328a7 --- /dev/null +++ b/week2/community-contributions/Figma_assistance/.github/workflows/update_space.yml @@ -0,0 +1,28 @@ +name: Run Python script + +on: + push: + branches: + - figma_assistance + +jobs: + build: + runs-on: ubuntu-latest + + steps: + - name: Checkout + uses: actions/checkout@v2 + + - name: Set up Python + uses: actions/setup-python@v2 + with: + python-version: '3.9' + + - name: Install Gradio + run: python -m pip install gradio + + - name: Log in to Hugging Face + run: python -c 'import huggingface_hub; huggingface_hub.login(token="${{ secrets.hf_token }}")' + + - name: Deploy to Spaces + run: gradio deploy diff --git a/week2/community-contributions/Figma_assistance/README.md b/week2/community-contributions/Figma_assistance/README.md new file mode 100644 index 0000000..26a783b --- /dev/null +++ b/week2/community-contributions/Figma_assistance/README.md @@ -0,0 +1,6 @@ +--- +title: Figma_assistance +app_file: day_5_figma_assistance.py +sdk: gradio +sdk_version: 5.38.2 +--- diff --git a/week2/community-contributions/Figma_assistance/day_5_figma_assistance.py b/week2/community-contributions/Figma_assistance/day_5_figma_assistance.py new file mode 100644 index 0000000..9d605f4 --- /dev/null +++ b/week2/community-contributions/Figma_assistance/day_5_figma_assistance.py @@ -0,0 +1,484 @@ +from openai import OpenAI +from dotenv import load_dotenv +import os +load_dotenv() +import gradio as gr +import base64 +from io import BytesIO +from PIL import Image +from IPython.display import Audio, display +import google.generativeai +import anthropic + +client = OpenAI(api_key=os.getenv("OPENAI_API_KEY")) + +# Configure Gemini +google.generativeai.configure(api_key=os.getenv("GOOGLE_API_KEY")) + +# Configure Claude +claude = anthropic.Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY")) +openAI_model = "gpt-3.5-turbo" +gemini_model = "gemini-2.0-flash" +claude_model = "claude-sonnet-4-20250514" +openai_audio_model = "tts-1" + +# Figma onboarding knowledge base +FIGMA_KNOWLEDGE = """ +You are a helpful Figma onboarding assistant. You help new users learn Figma's core features and workflows. + +Key Figma concepts to help users with: +- Interface overview (toolbar, layers panel, properties panel) +- Creating and editing frames +- Working with shapes, text, and components +- Using the pen tool for custom shapes +- Auto Layout for responsive designs +- Components and variants +- Prototyping and interactions +- Collaboration features +- Design systems and libraries +- Exporting assets +- Keyboard shortcuts + +Always provide clear, step-by-step instructions and mention relevant keyboard shortcuts when applicable. +""" + +promts = { + "Charlie": FIGMA_KNOWLEDGE +} + +def truncate_for_tts(text, max_length=4000): + """Truncate text for TTS while preserving complete sentences""" + if len(text) <= max_length: + return text + + # Try to truncate at sentence boundaries + sentences = text.split('. ') + truncated = "" + + for sentence in sentences: + if len(truncated + sentence + '. ') <= max_length: + truncated += sentence + '. ' + else: + break + + # If we couldn't fit any complete sentences, just truncate hard + if not truncated.strip(): + truncated = text[:max_length-10] + "..." + + return truncated.strip() + +def talker_openai(message): + """Generate audio from text using OpenAI TTS""" + try: + # Truncate message for TTS + truncated_message = truncate_for_tts(message) + + response = client.audio.speech.create( + model="tts-1", + voice="onyx", + input=truncated_message + ) + + audio_stream = BytesIO(response.content) + output_filename = "output_audio_openai.mp3" + with open(output_filename, "wb") as f: + f.write(audio_stream.read()) + + return output_filename + except Exception as e: + print(f"Error generating audio with OpenAI: {str(e)}") + return None + +def talker(message, model_choice): + """Generate audio from text using selected model""" + return talker_openai(message) + +def get_figma_help_openai(user_question, chat_history): + """Get Figma onboarding assistance using OpenAI""" + try: + messages = [ + {"role": "system", "content": FIGMA_KNOWLEDGE} + ] + + # Convert messages format chat history to OpenAI format + for msg in chat_history: + if msg["role"] == "user": + messages.append({"role": "user", "content": msg["content"]}) + elif msg["role"] == "assistant": + messages.append({"role": "assistant", "content": msg["content"]}) + + messages.append({"role": "user", "content": user_question}) + + response = client.chat.completions.create( + model=openAI_model, + messages=messages, + max_tokens=500, + temperature=0.7 + ) + return response.choices[0].message.content + + except Exception as e: + return f"Sorry, I encountered an error with OpenAI: {str(e)}" + +def get_figma_help_gemini(user_question, chat_history): + """Get Figma onboarding assistance using Gemini""" + try: + gemini = google.generativeai.GenerativeModel( + model_name=gemini_model, + system_instruction=FIGMA_KNOWLEDGE, + ) + + # Build conversation context from messages format + conversation_context = "" + for msg in chat_history: + if msg["role"] == "user": + conversation_context += f"User: {msg['content']}\n" + elif msg["role"] == "assistant": + conversation_context += f"Assistant: {msg['content']}\n\n" + + message = conversation_context + f"User: {user_question}" + response = gemini.generate_content(message) + reply = response.text + return reply + + except Exception as e: + return f"Sorry, I encountered an error with Gemini: {str(e)}" + +def get_figma_help_claude(user_question, chat_history): + """Get Figma onboarding assistance using Claude""" + try: + # Convert messages format to Claude format + claude_messages = [] + for msg in chat_history: + if msg["role"] == "user": + claude_messages.append({"role": "user", "content": msg["content"]}) + elif msg["role"] == "assistant": + claude_messages.append({"role": "assistant", "content": msg["content"]}) + + # Add the current question + claude_messages.append({"role": "user", "content": user_question}) + + response = claude.messages.create( + model=claude_model, + max_tokens=500, + temperature=0.7, + system=promts["Charlie"], + messages=claude_messages, + ) + reply = response.content[0].text + return reply + + except Exception as e: + return f"Sorry, I encountered an error with Claude: {str(e)}" + +def respond(message, chat_history, model_choice): + if not message.strip(): + return "", chat_history, "", model_choice + + bot_message = get_figma_help(message, chat_history, model_choice) + + # Add user message and bot response in messages format + new_history = chat_history + [ + {"role": "user", "content": message}, + {"role": "assistant", "content": bot_message} + ] + + return "", new_history, bot_message, model_choice + +def clear_chat(): + """Clear the chat history""" + return [], "", None + +def get_figma_help(user_question, chat_history, model_choice): + """Get Figma onboarding assistance using selected model""" + if model_choice == "OpenAI (GPT-3.5)": + return get_figma_help_openai(user_question, chat_history) + elif model_choice == "Google Gemini (2.0 Flash)": + return get_figma_help_gemini(user_question, chat_history) + elif model_choice == "Claude (Sonnet 4)": + return get_figma_help_claude(user_question, chat_history) + else: + return "Please select a valid model." + +custom_css = """ +/* Chat area styling */ +.styled-chat { + border-radius: 15px !important; + box-shadow: 0 4px 12px var(--shadow-color) !important; + border: 1px solid var(--border-color) !important; + padding: 10px; +} + +/* Audio player styling */ +.styled-audio { + border-radius: 15px !important; + box-shadow: 0 4px 12px var(--shadow-color) !important; + border: 10px solid var(--block-background-fill) !important; + padding: 10px; + background-color: var(--background-fill-secondary) !important; +} + +/* Header styling */ +.header-container { + text-align: center; + padding: 20px; + background: linear-gradient(135deg, #667eea 0%, #764ba2 100%); + border-radius: 15px; + margin-bottom: 20px; +} + +.header-title { + color: white; + margin: 0; + font-size: 2.5em; +} + +.header-subtitle { + color: #f0f0f0; + margin: 10px 0 0 0; + font-size: 1.2em; +} + +/* Features section styling */ +.features-container { + background: #f8f9fa; + padding: 20px; + border-radius: 10px; + border-left: 4px solid #667eea; +} + +.features-title { + color: #333; + margin-top: 0; +} + +.features-grid { + display: grid; + grid-template-columns: 1fr 1fr; + gap: 15px; + margin-top: 15px; +} + +.feature-item { + color: #333; + margin: 10px 0; +} + +.feature-title { + color: #667eea; +} + +.feature-description { + color: #666; +} + +/* Pro tip styling */ +.protip-container { + text-align: center; + margin-top: 20px; + padding: 15px; + background: #e8f4f8; + border-radius: 8px; +} + +.protip-text { + margin: 0; + color: #2c5aa0 !important; + font-weight: 500; +} + +/* Quick start questions styling */ +.quickstart-container { + background: linear-gradient(135deg, #4facfe 0%, #00f2fe 100%); + padding: 15px 20px; + border-radius: 10px; + margin: 20px 0; +} + +.quickstart-title { + color: white !important; + margin: 0; + font-size: 1.3em; + text-align: center; +} + +.quickstart-subtitle { + color: #f0f8ff !important; + margin: 5px 0 0 0; + text-align: center; + font-size: 0.9em; +} +""" + +# Create Gradio interface +with gr.Blocks(title="Figma Onboarding Assistant", theme=gr.themes.Soft(), css=custom_css) as demo: + gr.HTML( + """ +
+

🎨 Figma Onboarding Assistant

+

Your AI-powered Figma learning companion

+
+ +
+

✨ What I can help you with:

+
+
+

🚀 Getting Started
+ Interface overview, basic navigation

+

🛠️ Tools & Features
+ Pen tool, shapes, text, layers

+

📐 Auto Layout
+ Responsive design techniques

+

🔗 Prototyping
+ Interactions and animations

+
+
+

🧩 Components
+ Creating reusable elements

+

👥 Collaboration
+ Sharing and team workflows

+

📚 Design Systems
+ Libraries and style guides

+

⚡ Shortcuts
+ Productivity tips and tricks

+
+
+
+ +
+

💡 Pro tip: Ask specific questions like "How do I create a button component?" for the best results!

+
+ """ + ) + + # Model selection dropdown + model_dropdown = gr.Dropdown( + choices=["OpenAI (GPT-3.5)", "Google Gemini (2.0 Flash)", "Claude (Sonnet 4)"], + value="OpenAI (GPT-3.5)", + label="Select AI Model", + info="Choose which AI model to use for responses" + ) + + with gr.Row(): + msg = gr.Textbox( + placeholder="Type your Figma question here...", + container=False, + scale=4 + ) + submit_btn = gr.Button("Ask", scale=1, variant="primary") + clear_btn = gr.Button("Clear Chat", scale=1) + audio_btn = gr.Button("🔊 Play Audio", scale=1, variant="secondary") + clear_audio_btn = gr.Button("🔇 Clear Audio", scale=1, variant="secondary") + + + # Example questions + gr.HTML( + """ +
+

🚀 Quick Start Questions

+

Click any question below to get started instantly!

+
+ """ + ) + + with gr.Row(): + example_btns = [ + gr.Button( + "How do I create my first frame?", + size="sm", + variant="secondary" + ), + gr.Button( + "What's the difference between components and instances?", + size="sm", + variant="secondary" + ), + gr.Button( + "How do I use Auto Layout?", + size="sm", + variant="secondary" + ), + gr.Button( + "How do I create a prototype?", + size="sm", + variant="secondary" + ) + ] + + # Your components with simple styling + chatbot = gr.Chatbot( + type="messages", + height=400, + placeholder="Ask me anything about Figma! For example: 'How do I create a component?' or 'What are frames in Figma?'", + elem_classes=["styled-chat"] + ) + + audio_output = gr.Audio( + label="Audio Response", + visible=True, + elem_classes=["styled-audio"] + ) + + last_response = gr.State("") + current_model = gr.State("OpenAI (GPT-3.5)") + + def respond(message, chat_history, model_choice): + if not message.strip(): + return "", chat_history, "", model_choice + + bot_message = get_figma_help(message, chat_history, model_choice) + new_history = chat_history + [ + {"role": "user", "content": message}, + {"role": "assistant", "content": bot_message}] + return "", new_history, bot_message, model_choice + + def play_audio(last_message, model_choice): + if last_message: + audio_file = talker(last_message, model_choice) + if audio_file: + return audio_file + return None + + def clear_audio(): + """Clear the audio output""" + return None + + def use_example(example_text): + return example_text + + # Set up interactions + submit_btn.click( + respond, + inputs=[msg, chatbot, model_dropdown], + outputs=[msg, chatbot, last_response, current_model] + ) + msg.submit( + respond, + inputs=[msg, chatbot, model_dropdown], + outputs=[msg, chatbot, last_response, current_model] + ) + clear_btn.click(clear_chat, outputs=[chatbot, msg, last_response]) + + # Audio button functionality - now uses selected model + audio_btn.click( + play_audio, + inputs=[last_response, current_model], + outputs=[audio_output] + ) + + # Clear audio button functionality + clear_audio_btn.click( + clear_audio, + outputs=[audio_output] + ) + + # Example button clicks + for i, btn in enumerate(example_btns): + btn.click( + use_example, + inputs=[btn], + outputs=[msg] + ) + +# Launch the app +demo.launch(share=True) \ No newline at end of file diff --git a/week2/community-contributions/Figma_assistance/requirements.txt b/week2/community-contributions/Figma_assistance/requirements.txt new file mode 100644 index 0000000..c090b08 --- /dev/null +++ b/week2/community-contributions/Figma_assistance/requirements.txt @@ -0,0 +1,7 @@ +openai +python-dotenv +gradio +pillow +google-generativeai +anthropic +ipython diff --git a/week2/community-contributions/HistoryBot-Week2Exercise.ipynb b/week2/community-contributions/HistoryBot-Week2Exercise.ipynb new file mode 100644 index 0000000..58d1728 --- /dev/null +++ b/week2/community-contributions/HistoryBot-Week2Exercise.ipynb @@ -0,0 +1,398 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "6a193ef7-41df-42cb-ab35-fb5fa77a78b9", + "metadata": {}, + "source": [ + "

HelloHistory- Learn History On the Go

" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "d813bcdb-fbff-43f8-97ae-28cf1ec2e094", + "metadata": {}, + "outputs": [], + "source": [ + "#Now use everything you've learned from Week 2 to build a full prototype for the technical question/answerer you built in Week 1 Exercise.\n", + "\n", + "#This should include a Gradio UI, streaming, use of the system prompt to add expertise, and the ability to switch between models. Bonus points if you can demonstrate use of a tool!\n", + "\n", + "#If you feel bold, see if you can add audio input so you can talk to it, and have it respond with audio. ChatGPT or Claude can help you, or email me if you have questions." + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "ace8fd2d-341e-451d-a70e-82fac828299c", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "import gradio as gr" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "591b90d1-9771-40ae-ad99-6e864465a358", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk\n", + "Anthropic API Key exists and begins sk\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "#google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:2]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:2]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "eace2872-0ddb-4b86-ae09-91ad9fc2dd04", + "metadata": {}, + "outputs": [], + "source": [ + "#connect to models\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "d5e99852-89f7-41da-84a5-5cf8659faddc", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful tutor teaching people history. You have to answer their questions on historical events.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "01d3fcb8-967e-4841-809a-b428e80c17c9", + "metadata": {}, + "outputs": [], + "source": [ + "#test function\n", + "\n", + "def message_gpt(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " completion = openai.chat.completions.create(\n", + " model='gpt-4o-mini',\n", + " messages=messages,\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "a214da55-644f-4469-8167-1b317a7cb8ce", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'World War II was won by the Allies, a coalition of countries that included the United States, the Soviet Union, the United Kingdom, China, and several other nations. The war officially ended in 1945, with the unconditional surrender of Nazi Germany in May and the surrender of Japan in September.'" + ] + }, + "execution_count": 7, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "\n", + "message_gpt(\"Who won World War II?\")" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "65b9c8ac-1319-46ed-a5d8-82d98cb3d831", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "\"World War II began on September 1, 1939, when Germany, led by Adolf Hitler, invaded Poland. This invasion was a result of aggressive expansionist policies pursued by Nazi Germany throughout the 1930s, which included the annexation of Austria and the incorporation of Czechoslovakia's Sudetenland.\\n\\nThe invasion of Poland prompted Britain and France to declare war on Germany on September 3, 1939, fulfilling their commitments to support Poland. Tensions had been building in Europe due to unresolved issues from World War I, the rise of totalitarian regimes, and various treaties and alliances. The war would expand rapidly as other nations became involved, eventually leading to a global conflict that lasted until 1945.\"" + ] + }, + "execution_count": 8, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "\n", + "message_gpt(\"How did World War II begin?\")" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "41618892-8a7b-4871-9d8e-d030fabf1046", + "metadata": {}, + "outputs": [], + "source": [ + "#add streaming \n", + "def stream_gpt(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " stream = openai.chat.completions.create(\n", + " model='gpt-4o-mini',\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "ad6edb2a-9c2c-4b53-bead-6009f493f281", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7861\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 11, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "view = gr.Interface(\n", + " fn=stream_gpt,\n", + " inputs=[gr.Textbox(label=\"Ask HistoryBot a question:\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "54be6b30-db25-49b9-aecb-e63daa0b6873", + "metadata": {}, + "outputs": [], + "source": [ + "#streaming with Claude\n", + "\n", + "def stream_claude(prompt):\n", + " result = claude.messages.stream(\n", + " model=\"claude-3-haiku-20240307\",\n", + " max_tokens=1000,\n", + " system=system_message,\n", + " messages=[\n", + " {\"role\": \"user\", \"content\": prompt},\n", + " ],\n", + " )\n", + " response = \"\"\n", + " with result as stream:\n", + " for text in stream.text_stream:\n", + " response += text or \"\"\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "62d94a31-1b2b-4266-8cfa-16e877240aa8", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7862\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 13, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "view = gr.Interface(\n", + " fn=stream_claude,\n", + " inputs=[gr.Textbox(label=\"Ask HistoryBot a question:\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "b69fdbe0-abc7-429a-aadd-44620035f49e", + "metadata": {}, + "outputs": [], + "source": [ + "# function to select model\n", + "\n", + "def stream_model(prompt, model):\n", + " if model==\"GPT\":\n", + " result = stream_gpt(prompt)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(prompt)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " yield from result" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "40b2961a-2bb2-4a55-9abb-be967e184db9", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7864\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 16, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "view = gr.Interface(\n", + " fn=stream_model,\n", + " inputs=[gr.Textbox(label=\"Ask HistoryBot a question\"), gr.Dropdown([\"GPT\", \"Claude\"], label=\"Select model\", value=\"GPT\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e8b78f97-597a-4d4a-8f7a-c6e982290596", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/Mediterranean Banter.ipynb b/week2/community-contributions/Mediterranean Banter.ipynb new file mode 100644 index 0000000..5ac089c --- /dev/null +++ b/week2/community-contributions/Mediterranean Banter.ipynb @@ -0,0 +1,252 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "9ab6f493-026f-4950-b244-22c5251b8daa", + "metadata": {}, + "source": [ + "# Mediterranean Banter" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4568bc63-679e-4ea1-a9c9-b85dfc386ec7", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7a852bee-76e6-4538-89a3-0702c2d5f05c", + "metadata": {}, + "outputs": [], + "source": [ + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c843ac74-4797-4bd0-bed2-dfcaa2f98c41", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e4181018-1818-47fa-a3fb-554627a63f69", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "19f833c8-e183-469c-a8c8-1c014889a15d", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_model_spain = \"gpt-4.1-mini\"\n", + "claude_model_italy = \"claude-3-5-haiku-latest\"\n", + "gemini_model_france = \"gemini-2.0-flash\"\n", + "\n", + "gpt_system_spain = \"You are a chatbot who is very argumentative; \\\n", + "you disagree with anything other than how beautiful and great South of Spain is.\"\n", + "\n", + "claude_system_italy = \"You are a very polite, courteous chatbot but you favour south of Italy prasing Amalfi Coast.\"\n", + "\n", + "gemini_system_france = \"You are neutral. You always accept what other people say and never try to put down other people thoughts.Although, you politely try to include that South of France is beautiful\"\n", + "\n", + "gpt_messages_spain = [\"Hi there, nothing beats the beauty of Spain and its wonderful beaches.\"]\n", + "claude_messages_italy = [\"I agree. I admire the Southern Part of Spain but its not as pretty as Amalfi Coast.\"]\n", + "gemini_messages_france = [\"Well, both are good and so is the French Riveria.\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2d426b95-5e7c-49aa-a5a1-9613296eb0d0", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system_spain}]\n", + " for gpt, claude,gemini in zip(gpt_messages_spain, claude_messages_italy,gemini_messages_france):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model_spain,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3fc9a696-3145-4f37-873b-539647f2fc0b", + "metadata": {}, + "outputs": [], + "source": [ + "call_gpt()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "63910faa-a122-4261-82a0-7530c6c5749a", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = []\n", + " for gpt_spain, claude_italy,gemini_france in zip(gpt_messages_spain, claude_messages_italy,gemini_messages_france):\n", + " messages.append({\"role\": \"user\", \"content\": gpt_spain})\n", + " messages.append({\"role\": \"user\", \"content\": gemini_france})\n", + " messages.append({\"role\": \"assistant\", \"content\": claude_italy})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages_spain[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": gemini_messages_france[-1]})\n", + " message = claude.messages.create(\n", + " model=claude_model_italy,\n", + " system=claude_system_italy,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3ab6aa2-a462-4fb3-bb6a-dc6b971827fa", + "metadata": {}, + "outputs": [], + "source": [ + "call_claude()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "114cb7eb-0915-46ac-b285-e40acf4a9ffb", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gemini():\n", + " messages=[]\n", + " for gpt_spain, claude_italy,gemini_france in zip(gpt_messages_spain, claude_messages_italy,gemini_messages_france):\n", + " messages.append({\"role\": \"user\", \"content\": gpt_spain})\n", + " messages.append({\"role\": \"user\", \"content\": claude_italy})\n", + " messages.append({\"role\": \"assistant\", \"content\": gemini_france})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages_spain[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": claude_messages_italy[-1]})\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name='gemini-2.0-flash',\n", + " system_instruction=gemini_system_france\n", + " )\n", + " dialogue_text = \"\\n\".join(f\"{m['role']}: {m['content']}\" for m in messages)\n", + " response = gemini.generate_content(dialogue_text)\n", + " return response.text\n", + " \n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e3acf708-f9b1-4a6d-b3e1-823c96d00555", + "metadata": {}, + "outputs": [], + "source": [ + "call_gemini()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c103430e-68c7-4cc6-8a43-6b5aec7fdc96", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages_spain = [\"Hi there, nothing beats the beauty of Spain and its wonderful beaches.\"]\n", + "claude_messages_italy = [\"I agree. I admire the Southern Part of Spain but its not as pretty as Amalfi Coast.\"]\n", + "gemini_messages_france = [\"Well, both are good and so is the French Riveria.\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages_spain[0]}\\n\")\n", + "print(f\"Claude:\\n{claude_messages_italy[0]}\\n\")\n", + "print(f\"Gemini:\\n{gemini_messages_france[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages_spain.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude:\\n{claude_next}\\n\")\n", + " claude_messages_italy.append(claude_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " gemini_messages_france.append(gemini_next)\n", + "\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/SushiRestaurant.ipynb b/week2/community-contributions/SushiRestaurant.ipynb new file mode 100644 index 0000000..ad32c65 --- /dev/null +++ b/week2/community-contributions/SushiRestaurant.ipynb @@ -0,0 +1,181 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "757905af-7f93-4dca-9526-063bc93a78c7", + "metadata": {}, + "source": [ + "# Sakana-ya (魚屋) Sushi\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9a6721fb-efca-4412-a0a7-cc8e6c4ced76", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import json" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b0fa458f-f73f-491c-b666-95db4b91f571", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aa2846f2-e09c-421d-9774-c04961a79800", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "MODEL = 'gpt-4o-mini'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7672ecdf-cf50-4b96-887a-b0a4eb5bbbf5", + "metadata": {}, + "outputs": [], + "source": [ + " \n", + "menu = {\n", + " \"Nigiri (1 pc)\": {\n", + " \"Salmon\": 4.25,\n", + " \"Tuna\": 4.75,\n", + " \"Yellowtail\": 5.00,\n", + " \"Eel\": 5.25,\n", + " \"Tamago\": 3.00,\n", + " },\n", + " \"Sashimi (3 pc)\": {\n", + " \"Salmon\": 8.50,\n", + " \"Tuna\": 9.00,\n", + " \"Yellowtail\": 9.50,\n", + " \"Octopus\": 8.00,\n", + " },\n", + " \"Classic Rolls (6 pc)\": {\n", + " \"California\": 6.50,\n", + " \"Spicy Tuna\": 7.50,\n", + " \"Philadelphia\": 7.25,\n", + " \"Cucumber\": 4.50,\n", + " \"Avocado\": 4.75,\n", + " },\n", + " \"Specialty Rolls (8 pc)\": {\n", + " \"Dragon\": 13.50,\n", + " \"Rainbow\": 14.00,\n", + " \"Crunchy Shrimp\": 12.50,\n", + " \"Volcano\": 13.00,\n", + " \"Spider\": 14.50,\n", + " },\n", + " \"Appetizers\": {\n", + " \"Edamame\": 5.00,\n", + " \"Gyoza (5)\": 6.50,\n", + " \"Miso Soup\": 3.00,\n", + " \"Seaweed Salad\": 5.50,\n", + " },\n", + " \"Beverages\": {\n", + " \"Green Tea\": 2.50,\n", + " \"Ramune Soda\": 3.00,\n", + " \"Sparkling Water\": 2.75,\n", + " },\n", + " \"Desserts\": {\n", + " \"Mochi Ice Cream (2)\": 5.00,\n", + " \"Matcha Cheesecake\": 6.50,\n", + " },\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "99914500-3630-4fea-987c-d19c760994c6", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " system_message = \"You are a helpful assistant for Sakana-ya (魚屋) Sushi restaurant.\\\n", + " Help out with information and if you dont know something just say you cant help with that.\"\n", + " system_message += json.dumps(menu)\n", + " system_message+=\"If something is not in the menu, we dont serve it.\\\n", + " If we dont have a dish just mention it that we dont offer it. \"\n", + "\n", + " sushi_exotic = [\n", + " {\"role\": \"user\", \"content\": \"Do you have aji?\"},\n", + " {\"role\": \"user\", \"content\": \"We currently dont have shun its available only during the season i.e in May.\"},\n", + " {\"role\": \"user\", \"content\": \"What about buri?\"},\n", + " {\"role\": \"user\", \"content\": \"Thats seasonal as well only during December. Do visit us during that time.\"},\n", + " \n", + " ]\n", + " \n", + " messages = [{\"role\": \"system\", \"content\": system_message}]+ sushi_exotic + history + [{\"role\": \"user\", \"content\": message}]\n", + " stream = openai.chat.completions.create(model=MODEL, messages=messages, stream=True)\n", + "\n", + " response = \"\"\n", + " for chunk in stream:\n", + " response += chunk.choices[0].delta.content or ''\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a5c61d91-abee-4ada-9a42-ae87cf53fcff", + "metadata": {}, + "outputs": [], + "source": [ + "gr.ChatInterface(fn=chat, type=\"messages\").launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/TicketPriceWithGoogleSearch/README.md b/week2/community-contributions/TicketPriceWithGoogleSearch/README.md new file mode 100644 index 0000000..2cef712 --- /dev/null +++ b/week2/community-contributions/TicketPriceWithGoogleSearch/README.md @@ -0,0 +1,77 @@ +# Flight Ticket Price Agent + +This project implements a conversational AI agent that can find and analyze flight ticket prices. Users can ask for flight prices between two cities in a natural language chat interface, and the agent will use a combination of web search and language models to provide a summary of the costs. + +## Features + +- **Conversational Interface:** A user-friendly chat interface built with Gradio. +- **Multi-Model Support:** Can be configured to use different Large Language Models (LLMs) for analysis, including: + - OpenAI (e.g., GPT-4o-mini) + - Google Gemini (e.g., Gemini 2.5 Flash) + - Ollama (e.g., Llama 3.1) +- **Tool-Based Architecture:** The agent uses a `get_ticket_price` tool to understand when the user is asking for flight information. +- **Web Scraping & Analysis:** + - Uses Google Custom Search to find relevant web pages with flight information. + - Scrapes the content of the search results. + - Leverages an LLM to analyze the scraped text and extract the lowest, highest, and average prices for one-way and round-trip flights. +- **Caching:** Caches search results to provide faster responses for repeated queries. +- **Currency Conversion:** Includes a basic currency conversion table to standardize prices to INR. + +## Requirements + +The project is built with Python and requires the following libraries: + +- `python-dotenv` +- `openai` +- `google-generativeai` +- `ollama` +- `gradio` +- `requests` +- `beautifulsoup4` +- `google-api-python-client` +- `ipython` + +You can install these dependencies using pip: +```bash +pip install python-dotenv openai google-generativeai ollama gradio requests beautifulsoup4 google-api-python-client ipython +``` + +## Setup + +1. **Clone the repository (optional):** + ```bash + git clone + cd /ticket_price_agent + ``` + +2. **Create a `.env` file:** + Create a file named `.env` in the `ticket_price_agent` directory and add your API keys: + ```env + OPENAI_API_KEY="your_openai_api_key" + GEMINI_API_KEY="your_gemini_api_key" + GOOGLE_SEARCH_KEY="your_google_search_api_key" + GOOGLE_CSE_ID="your_google_custom_search_engine_id" + ``` + * `GOOGLE_SEARCH_KEY` and `GOOGLE_CSE_ID` are required for the Google Custom Search API. + +3. **Install Dependencies:** + Run the `pip install` command mentioned in the "Requirements" section. + +## Usage + +1. **Open the Notebook:** Launch Jupyter Notebook or JupyterLab and open `ticket_price_agent.ipynb`. +2. **Run the Cells:** Execute the cells in the notebook sequentially. +3. **Interact with the Agent:** The final cell will start a Gradio chat interface. You can select the model you want to use (OpenAI or Gemini) from the dropdown menu. +4. **Ask for Prices:** Start a conversation by asking for flight prices, for example: + - "How much is a ticket from Delhi to Mumbai?" + - "What's the flight cost to Kathmandu from Delhi?" + +## How It Works + +1. **User Input:** The user enters a message in the Gradio chat interface. +2. **Model Selection:** The selected LLM (OpenAI or Gemini) processes the input. +3. **Tool Call:** The model's function-calling/tool-using capability identifies that the user is asking for a price and calls the `get_ticket_price` function with the extracted departure and destination cities. +4. **Google Search:** The `get_ticket_price` function constructs a search query and uses the Google Custom Search API to find relevant links. +5. **Web Scraping:** The agent scrapes the content from the top search result pages. +6. **Price Analysis:** The scraped content is passed to another LLM instance with a specific prompt to analyze the text and extract price information (lowest, highest, average). +7. **Response Generation:** The final price summary is returned to the main chat model, which then formulates a user-friendly response. diff --git a/week2/community-contributions/TicketPriceWithGoogleSearch/ticket_price_agent.ipynb b/week2/community-contributions/TicketPriceWithGoogleSearch/ticket_price_agent.ipynb new file mode 100644 index 0000000..67817f0 --- /dev/null +++ b/week2/community-contributions/TicketPriceWithGoogleSearch/ticket_price_agent.ipynb @@ -0,0 +1,617 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "a390d675", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import json\n", + "import ollama\n", + "from google import genai\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "from IPython.display import Markdown" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "55c9c2a2", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "gemini_api_key = os.getenv('GEMINI_API_KEY')\n", + " \n", + "OPENAI_MODEL = 'gpt-4o-mini'\n", + "GEMINI_MODEL = 'gemini-2.5-flash' \n", + "OLLAMA_MODEL = 'llama3.2'\n", + "\n", + "openai = OpenAI()\n", + "gemini = genai.Client(api_key = gemini_api_key)\n", + "\n", + "tools = []\n", + "gemini_tools = []\n", + "\n", + "cached_search = {\n", + " ('delhi', 'delhi'): \"INR 0\",\n", + "}\n", + "\n", + "convertion_rate_to_inr = {\n", + " \"USD\": 85.81,\n", + " \"EUR\": 100.25,\n", + " \"GBP\": 115.90,\n", + " \"AUD\": 56.43,\n", + " \"CAD\": 62.70,\n", + " \"SGD\": 67.05,\n", + " \"CHF\": 107.79,\n", + " \"JPY\": 0.5825,\n", + " \"CNY\": 11.97,\n", + " \"AED\": 23.37,\n", + " \"NZD\": 51.56,\n", + " \"SAR\": 22.88,\n", + " \"QAR\": 23.58,\n", + " \"OMR\": 222.89,\n", + " \"BHD\": 227.62,\n", + " \"KWD\": 280.90,\n", + " \"MYR\": 20.18,\n", + " \"THB\": 2.655,\n", + " \"HKD\": 10.93,\n", + " \"ZAR\": 4.79\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "68ec7079", + "metadata": {}, + "outputs": [], + "source": [ + "import requests\n", + "from bs4 import BeautifulSoup\n", + "\n", + "\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url \n", + " try:\n", + " response = requests.get(url=self.url, headers=headers, timeout=10)\n", + " response.raise_for_status()\n", + " self.body = response.content \n", + " except requests.RequestException as e:\n", + " print(f\"Failed to fetch {self.url}: {e}\")\n", + " self.body = b\"\"\n", + " self.title = \"Failed to load\"\n", + " self.text = \"\"\n", + " self.links = []\n", + " return\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body(['script', 'style', 'img', 'input']):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\" \n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_content(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"\n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "812ddcef", + "metadata": {}, + "outputs": [], + "source": [ + "from googleapiclient.discovery import build\n", + "from googleapiclient.errors import HttpError\n", + "\n", + "class GoogleSearch:\n", + " def __init__(self, api_key=None, cse_id=None):\n", + " \"\"\"\n", + " Initialize the Google Search Tool\n", + " \n", + " Args:\n", + " api_key: Your Google API key (or set GOOGLE_API_KEY env var)\n", + " cse_id: Your Custom Search Engine ID (or set GOOGLE_CSE_ID env var)\n", + " \"\"\"\n", + " self.api_key = api_key or os.getenv('GOOGLE_SEARCH_KEY')\n", + " self.cse_id = cse_id or os.getenv('GOOGLE_CSE_ID')\n", + "\n", + " if not self.api_key:\n", + " raise ValueError(\"API key is required. Set GOOGLE_API_KEY env var or pass api_key parameter\")\n", + " if not self.cse_id:\n", + " raise ValueError(\"CSE ID is required. Set GOOGLE_CSE_ID env var or pass cse_id parameter\")\n", + " \n", + " self.service = build(\"customsearch\", \"v1\", developerKey=self.api_key)\n", + " \n", + " def search(self, query: str, num_result: int=10, start_index: int=1):\n", + " \"\"\"\n", + " Perform a Google Custom Search\n", + " \n", + " Args:\n", + " query: Search query string\n", + " num_results: Number of results to return (1-10)\n", + " start_index: Starting index for results (for pagination)\n", + " \n", + " Returns:\n", + " dict: Search results or None if error\n", + " \"\"\"\n", + " try:\n", + " res = self.service.cse().list(\n", + " q=query,\n", + " cx=self.cse_id,\n", + " num=min(num_result, 10),\n", + " start=start_index\n", + " ).execute()\n", + "\n", + " return self._parse_results(res)\n", + " except HttpError as e:\n", + " print(f\"HTTP Error: {e}\")\n", + " return None\n", + " except Exception as e:\n", + " print(f\"Unexpected error: {e}\")\n", + " return None\n", + " \n", + " def _parse_results(self, raw_res):\n", + " \"\"\"Parse raw API response into clean format\"\"\"\n", + " if \"items\" not in raw_res:\n", + " return {\n", + " 'total_results': 0,\n", + " 'results': [],\n", + " 'search_info': raw_res.get('searchInformation', {})\n", + " }\n", + " \n", + " parsed_items = []\n", + " for item in raw_res[\"items\"]:\n", + " parsed_item = {\n", + " \"title\": item.get(\"title\", ''),\n", + " \"link\": item.get(\"link\", ''),\n", + " \"snippet\": item.get(\"snippet\", ''),\n", + " \"display_link\": item.get(\"display_link\", ''),\n", + " 'formatted_url': item.get('formattedUrl', '')\n", + " }\n", + "\n", + " parsed_items.append(parsed_item)\n", + " \n", + " return {\n", + " 'total_results': int(raw_res.get('searchInformation', {}).get('totalResults', '0')),\n", + " 'results': parsed_items,\n", + " 'search_info': raw_res.get('searchInformation', {})\n", + " }\n", + " \n", + " def compile_search_pages(self, query: str, num_result: int = 10, start_index: int=1):\n", + " \"\"\"\n", + " Compiles a list of results from multiple search pages for a given query\n", + "\n", + " Args:\n", + " query: Search query string\n", + " num_results: Number of results to return (1-10)\n", + " start_index: Starting index for results (for pagination)\n", + " \n", + " Returns:\n", + " str: Concatenated results from all search pages for the given query\n", + " \"\"\"\n", + "\n", + " result = \"\"\n", + "\n", + " search_res = self.search(query=query, num_result=num_result, start_index=start_index)\n", + "\n", + " print(search_res)\n", + "\n", + " for item in search_res['results']:\n", + " print(item.get('title'))\n", + " result += f\"\\n\\nTitle: {item.get('title', '')}\\n\"\n", + " result += Website(item.get('link', '')).get_content()\n", + "\n", + " print(result)\n", + "\n", + " return result\n", + "\n", + "google_search = GoogleSearch()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "857e77f8", + "metadata": {}, + "outputs": [], + "source": [ + "# google_search.compile_search_pages('flight ticket price from delhi to chandigarh', num_result=4)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ec5cf817", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful assistant for an Airline called FlightAI. \"\n", + "system_message += \"Give short, courteous answers, no more than 1 sentence. \"\n", + "system_message += \"Always be accurate. If you don't know the answer, say so.\"\n", + "system_message += \"Always ask the user about the departure point in case it asks about the price and departure is not mentioned.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b829f882", + "metadata": {}, + "outputs": [], + "source": [ + "def analyze_result_for_price(result: str, source: str, model: str):\n", + " print(\"Analyze web results: \", source, model)\n", + "\n", + " system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a search query.\"\n", + " system_prompt = \"Provide the lowest price, highest price and average price for one way and round trips.\"\n", + " system_prompt += \"Always return the price in INR. If you are not sure about the conversion rate, only then use the following conversion rates:\"\n", + " system_prompt += f\"{convertion_rate_to_inr} for conversion rates. Interpret the given conversion rate as for example:\"\n", + " system_prompt += \"1 USD to INR = 85.81. Return result in Markdown\"\n", + " \n", + " if source == 'ollama':\n", + " model_to_use = model if model else OLLAMA_MODEL\n", + "\n", + " print(f\"Using model: {model_to_use}\\n\\n\")\n", + "\n", + " try:\n", + " response = ollama.chat(\n", + " model=model_to_use, \n", + " messages=[\n", + " {\"role\":\"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": result}\n", + " ],\n", + " )\n", + " \n", + " result = response['message']['content']\n", + " return result\n", + " except Exception as e:\n", + " print(f\"An error occurred during the API call: {e}\")\n", + " return None\n", + " elif source == 'openai':\n", + " try:\n", + " response = openai.chat.completions.create(\n", + " model=OPENAI_MODEL,\n", + " messages=[\n", + " {\"role\":\"system\", \"content\": system_prompt},\n", + " {\"role\":\"user\", \"content\": result}\n", + " ],\n", + " \n", + " )\n", + "\n", + " result = response.choices[0].message.content\n", + " return result\n", + " except Exception as e:\n", + " print(f\"An error occurred during the API call: {e}\")\n", + " return None\n", + " elif source == 'gemini':\n", + " try:\n", + " response = gemini.models.generate_content(\n", + " model=GEMINI_MODEL,\n", + " contents=f\"{system_prompt}\\n\\n{result}\"\n", + " )\n", + "\n", + " result = response.text\n", + " return result\n", + " except Exception as e:\n", + " print(f\"An error occurred during the API call: {e}\")\n", + " return None\n", + " else:\n", + " print(\"Source not supported\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eb2ba65a", + "metadata": {}, + "outputs": [], + "source": [ + "def get_ticket_price(destination_city, departure_city, source=\"openai\", model=\"\"):\n", + " if not destination_city or not departure_city:\n", + " return \"Error: Both destination and departure cities are required\"\n", + " \n", + " print(f\"Tool get_ticket_price called for {destination_city} from {departure_city}\")\n", + " print(\"get_ticket_price: \", model)\n", + "\n", + " dest = destination_city.lower()\n", + " dept = departure_city.lower()\n", + "\n", + " cache_key = (dest, dept)\n", + "\n", + " if cache_key not in cached_search:\n", + " try:\n", + " query = f'flight ticket price from {dept} to {dest}' \n", + " results = google_search.compile_search_pages(query=query, num_result=10) \n", + " \n", + " if results: # Check if results is not empty\n", + " cached_search[cache_key] = results\n", + " else:\n", + " return \"Error: No search results found\"\n", + " except Exception as e:\n", + " print(f\"Error during search: {e}\")\n", + " return f\"Error: Unable to fetch flight prices - {str(e)}\"\n", + " else:\n", + " results = cached_search[cache_key]\n", + "\n", + " try:\n", + " return analyze_result_for_price(results, source, model)\n", + " except Exception as e:\n", + " print(f\"Error analyzing results: {e}\")\n", + " return f\"Error: Unable to analyze price data - {str(e)}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b9c64fac", + "metadata": {}, + "outputs": [], + "source": [ + "# Markdown(get_ticket_price('New York', 'London', \"gemini\", \"\"))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "32440830", + "metadata": {}, + "outputs": [], + "source": [ + "price_function = {\n", + " \"name\": \"get_ticket_price\",\n", + " \"description\": \"Get the current flight ticket price between two cities. Call this whenever you need to know flight prices, for example when a customer asks 'How much is a ticket from Delhi to Mumbai?', 'What's the flight cost to Chandigarh?', or 'Show me ticket prices for travel between these cities'. This function searches for real-time flight pricing information from multiple sources.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"destination_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city that the customer wants to travel to (e.g., 'Mumbai', 'Delhi', 'Chandigarh')\",\n", + " },\n", + " \"departure_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city that the customer wants to travel from (e.g., 'Delhi', 'Mumbai', 'Bangalore')\",\n", + " },\n", + " \"source\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The AI model source to use for price analysis (optional, defaults to 'openai')\",\n", + " \"default\": \"openai\"\n", + " },\n", + " \"model\": {\n", + " \"type\": \"string\", \n", + " \"description\": \"The specific AI model to use for analysis (optional, defaults to empty string)\",\n", + " \"default\": \"\"\n", + " }\n", + " },\n", + " \"required\": [\"destination_city\", \"departure_city\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}\n", + "\n", + "tools.append({\"type\": \"function\", \"function\": price_function})" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c670e697", + "metadata": {}, + "outputs": [], + "source": [ + "gemini_tools = [\n", + " {\n", + " \"function_declarations\": [\n", + " {\n", + " \"name\": \"get_ticket_price\",\n", + " \"description\": \"Get the current flight ticket price between two cities. Call this whenever you need to know flight prices, for example when a customer asks 'How much is a ticket from Delhi to Mumbai?', 'What's the flight cost to Chandigarh?', or 'Show me ticket prices for travel between these cities'. This function searches for real-time flight pricing information from multiple sources.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"destination_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city that the customer wants to travel to (e.g., 'Mumbai', 'Delhi', 'Chandigarh')\"\n", + " },\n", + " \"departure_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city that the customer wants to travel from (e.g., 'Delhi', 'Mumbai', 'Bangalore')\"\n", + " },\n", + " \"source\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The AI model source to use for price analysis (optional, defaults to 'openai')\"\n", + " },\n", + " \"model\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The specific AI model to use for analysis (optional, defaults to empty string)\"\n", + " }\n", + " },\n", + " \"required\": [\"destination_city\", \"departure_city\"]\n", + " }\n", + " }\n", + " ]\n", + " }\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c608a19", + "metadata": {}, + "outputs": [], + "source": [ + "def handle_tool_call(message, model):\n", + "\n", + " tool_call = message.tool_calls[0]\n", + " arguments = json.loads(tool_call.function.arguments)\n", + " print(tool_call)\n", + " if tool_call.function.name == \"get_ticket_price\":\n", + " dest_city = arguments.get(\"destination_city\", '')\n", + " dept_city = arguments.get(\"departure_city\",'')\n", + " price = get_ticket_price(dest_city, dept_city, model, \"\")\n", + " return {\n", + " \"role\": \"tool\",\n", + " \"content\": json.dumps({\"destination_city\": dest_city,\"departure_city\": dept_city,\"price\": price}),\n", + " \"tool_call_id\": tool_call.id\n", + " }\n", + " return None\n", + "\n", + "def handle_tool_call_gemini(response, model):\n", + " tool_call = response.candidates[0].content.parts[0].function_call\n", + " function_name = tool_call.name\n", + " arguments = tool_call.args\n", + " \n", + " if function_name == \"get_ticket_price\":\n", + " dest_city = arguments.get(\"destination_city\", \"\")\n", + " dept_city = arguments.get(\"departure_city\", \"\")\n", + " price = get_ticket_price(dest_city, dept_city, model, \"\")\n", + " \n", + " return {\n", + " \"tool_response\": {\n", + " \"name\": function_name,\n", + " \"response\": {\n", + " \"content\": json.dumps({\n", + " \"destination_city\": dest_city,\n", + " \"departure_city\": dept_city,\n", + " \"price\": price\n", + " })\n", + " }\n", + " }\n", + " }\n", + " \n", + " return None" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "81c56b0d", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(history, model):\n", + " MODEL_TO_USE = \"\"\n", + " if model.lower() == 'openai':\n", + " MODEL_TO_USE = OPENAI_MODEL\n", + "\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = openai.chat.completions.create(model=MODEL_TO_USE, messages=messages, tools=tools)\n", + "\n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response = handle_tool_call(message, model.lower())\n", + " messages.append(message)\n", + " messages.append(response)\n", + " response = openai.chat.completions.create(model=MODEL_TO_USE, messages=messages, tools=tools)\n", + " \n", + " history += [{\"role\": \"assistant\", \"content\": response.choices[0].message.content}]\n", + " elif model.lower() == 'gemini':\n", + " MODEL_TO_USE = GEMINI_MODEL\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = gemini.models.generate_content(messages, tools=gemini_tools) \n", + " candidate = response.candidates[0]\n", + " \n", + " if candidate.finish_reason == 'TOOL_CALL':\n", + " messages.append(candidate.content)\n", + " tool_response = handle_tool_call_gemini(response, model.lower())\n", + " messages.append(tool_response)\n", + " response = gemini.models.generate_content(messages, tools=gemini_tools)\n", + " \n", + " history += [{\"role\": \"model\", \"content\": response.text}]\n", + " return history" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b2dac94", + "metadata": {}, + "outputs": [], + "source": [ + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, type=\"messages\")\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\")\n", + " model = gr.Dropdown([\"OpenAI\", \"Gemini\", \"Ollama\"], label=\"Choose a model\")\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def do_entry(message, history):\n", + " history += [{\"role\":\"user\", \"content\":message}]\n", + " return \"\", history\n", + "\n", + " entry.submit(do_entry, inputs=[entry, chatbot], outputs=[entry, chatbot]).then(\n", + " chat, inputs=[chatbot, model], outputs=[chatbot]\n", + " )\n", + " clear.click(lambda: None, inputs=None, outputs=chatbot, queue=False)\n", + "\n", + "ui.launch(inbrowser=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d50b03d4", + "metadata": {}, + "outputs": [], + "source": [ + "cached_search" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8a7f06bf", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/agent_conversation_shakespeare.ipynb b/week2/community-contributions/agent_conversation_shakespeare.ipynb new file mode 100644 index 0000000..6d55283 --- /dev/null +++ b/week2/community-contributions/agent_conversation_shakespeare.ipynb @@ -0,0 +1,351 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "06cf3063-9f3e-4551-a0d5-f08d9cabb927", + "metadata": {}, + "source": [ + "# Triangular agent conversation\n", + "\n", + "## GPT (Hamlet), LLM (Falstaff), Gemini (Iago):" + ] + }, + { + "cell_type": "markdown", + "id": "3637910d-2c6f-4f19-b1fb-2f916d23f9ac", + "metadata": {}, + "source": [ + "### Created a 3-way, bringing Gemini into the coversation.\n", + "### Replacing one of the models with an open source model running with Ollama." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f8e0c1bd-a159-475b-9cdc-e219a7633355", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display\n", + "import ollama" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a3ad57ad-46a8-460e-9cb3-67a890093536", + "metadata": {}, + "outputs": [], + "source": [ + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4f531c14-5743-4a5b-83d9-cb5863ca2ddf", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d5150ee-3858-4921-bce6-2eecfb96bc75", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI\n", + "\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "11381fd8-5099-41e8-a1d7-6787dea56e43", + "metadata": {}, + "outputs": [], + "source": [ + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1766d20-54b6-4f76-96c5-c338ae7073c9", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_model = \"gpt-4o-mini\"\n", + "llama_model = \"llama3.2\"\n", + "gemini_model = 'gemini-2.0-flash'\n", + "\n", + "gpt_system = \"You are playing part of Hamlet. he is philosopher, probes Iago with a mixture of suspicion\\\n", + "and intellectual curiosity, seeking to unearth the origins of his deceit.\\\n", + "Is malice born of scorn, envy, or some deeper void? Hamlet’s introspective nature\\\n", + "drives him to question whether Iago’s actions reveal a truth about humanity itself.\\\n", + "You will respond as Shakespear's Hamlet will do.\"\n", + "\n", + "llama_system = \"You are acting part of Falstaff who attempts to lighten the mood with his jokes and observations,\\\n", + "potentially clashing with Hamlet's melancholic nature.You respond as Shakespear's Falstaff do.\"\n", + "\n", + "gemini_system = \"You are acting part of Iago, subtly trying to manipulate both Hamlet and Falstaff\\\n", + "to his own advantage, testing their weaknesses and exploiting their flaws. You respond like Iago\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hello\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "806a0506-dac8-4bad-ac08-31f350256b58", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, llama_messages, gemini_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "43674885-ede7-48bf-bee4-467454f3e96a", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llama():\n", + " messages = []\n", + " for gpt, llama, gemini in zip(gpt_messages, llama_messages, gemini_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": llama})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " response = ollama.chat(model=llama_model, messages=messages)\n", + "\n", + " \n", + " return response['message']['content']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "03d34769-b339-4c4b-8c60-69494c39d725", + "metadata": {}, + "outputs": [], + "source": [ + "#import google.generativeai as genai\n", + "\n", + "# Make sure you configure the API key first:\n", + "#genai.configure(api_key=\"YOUR_API_KEY\")\n", + "\n", + "def call_gemini():\n", + " gemini_messages = []\n", + " \n", + " # Format the history for Gemini\n", + " for gpt, llama, gemini_message in zip(gpt_messages, llama_messages, gemini_messages):\n", + " gemini_messages.append({\"role\": \"user\", \"parts\": [gpt]}) # Hamlet speaks\n", + " gemini_messages.append({\"role\": \"model\", \"parts\": [llama]}) # Falstaff responds\n", + " gemini_messages.append({\"role\": \"model\", \"parts\": [gemini_message]}) # Iago responds\n", + "\n", + " # Add latest user input if needed (optional)\n", + " gemini_messages.append({\"role\": \"user\", \"parts\": [llama_messages[-1]]})\n", + "\n", + " # Initialize the model with the correct system instruction\n", + " gemini = google.generativeai.GenerativeModel(\n", + " #model_name='gemini-1.5-flash', # Or 'gemini-pro'\n", + " model_name = gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + "\n", + " response = gemini.generate_content(gemini_messages)\n", + " return response.text\n", + "#print(response.text)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "93fc8253-67cb-4ea4-aff7-097b2a222793", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hello\"]\n", + "\n", + "print(f\"Hamlet:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Falstaff:\\n{llama_messages[0]}\\n\")\n", + "print(f\"Iago:\\n{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(3):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " llama_next = call_llama()\n", + " print(f\"Llama:\\n{llama_next}\\n\")\n", + " llama_messages.append(llama_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " llama_messages.append(gemini_next)" + ] + }, + { + "cell_type": "markdown", + "id": "bca66ffc-9dc1-4384-880c-210889f5d0ac", + "metadata": {}, + "source": [ + "## Conversation between gpt-4.0-mini and llama3.2" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c23224f6-7008-44ed-a57f-718975f4e291", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4o-mini and Claude-3-haiku\n", + "# We're using cheap versions of models so the costs will be minimal\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "llama_model = \"llama3.2\"\n", + "\n", + "gpt_system = \"You are a tapori from mumbai who is very optimistic; \\\n", + "you alway look at the brighter part of the situation and you always ready to take act to win way.\"\n", + "\n", + "llama_system = \"You are a Jaat from Haryana. You try to express with hindi poems \\\n", + "to agree with other person and or find common ground. If the other person is optimistic, \\\n", + "you respond in poetic way and keep chatting.\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2d704bbb-f22b-400d-a695-efbd02b26548", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, llama in zip(gpt_messages, llama_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": llama})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "385ccec8-de59-4e42-9616-3f5c9a05589c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llama():\n", + " messages = []\n", + " for gpt, llama_message in zip(gpt_messages, llama_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": llama_message})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " response = ollama.chat(model=llama_model, messages=messages)\n", + "\n", + " \n", + " return response['message']['content']" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "70b5481b-455e-4275-80d3-0afe0fabcb0f", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "llama_messages = [\"Hi\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Llama:\\n{llama_messages[0]}\\n\")\n", + "\n", + "for i in range(3):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " llama_next = call_llama()\n", + " print(f\"Llama:\\n{llama_next}\\n\")\n", + " llama_messages.append(llama_next)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7f8d734b-57e5-427d-bcb1-7956fc58a348", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llmenv", + "language": "python", + "name": "llmenv" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/anatomy_poster_generator/README.md b/week2/community-contributions/anatomy_poster_generator/README.md new file mode 100644 index 0000000..cd82535 --- /dev/null +++ b/week2/community-contributions/anatomy_poster_generator/README.md @@ -0,0 +1,10 @@ +# Anatomy Poster Generator + +This tool generates AI-powered wall art of human anatomy, designed to support meaningful conversations in clinical spaces. + +Built with: +- DALL·E 3 for image generation +- Python + Gradio for a simple UI +- Hugging Face Spaces for easy sharing (https://huggingface.co/spaces/sukihealth/wallanatomypostergenerator) + +See full repo: [github.com/sukihealth/retro-pop-art-anatomy](https://github.com/sukihealth/retro-pop-art-anatomy) diff --git a/week2/community-contributions/animal_mixer.ipynb b/week2/community-contributions/animal_mixer.ipynb new file mode 100644 index 0000000..726321f --- /dev/null +++ b/week2/community-contributions/animal_mixer.ipynb @@ -0,0 +1,360 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "746c6089-658a-46b0-becd-44ed59f24ebe", + "metadata": {}, + "source": [ + "# Animal Mixer" + ] + }, + { + "cell_type": "markdown", + "id": "7fa554eb-db7f-486c-971b-98fae51107bd", + "metadata": {}, + "source": [ + "Given two animal species, let's make a cross between them and visualize the resulting new animal." + ] + }, + { + "cell_type": "markdown", + "id": "6e8c89b2-b4e8-48bb-9e2b-4455a5dd5a6e", + "metadata": {}, + "source": [ + "## Imports" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c063afce-a8e9-48cf-a08e-d70db2bb62e9", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import json\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import base64\n", + "from io import BytesIO\n", + "from PIL import Image\n", + "from IPython.display import Audio, display" + ] + }, + { + "cell_type": "markdown", + "id": "ab174215-1029-40df-9d75-a30f1c399fc9", + "metadata": {}, + "source": [ + "## Initialization" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b91a133e-becc-45ee-ad4c-6d3469c78826", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "MODEL = \"gpt-4o-mini\"\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "markdown", + "id": "e696d093-3b8b-4275-939c-53c7b623469b", + "metadata": {}, + "source": [ + "## System Messages" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f293608-376e-4f91-afce-e9d93787db03", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a famous zoologist-surgeon who makes crosses between animals, so new hybrid animals with mixed features of both original animals. \"\n", + "system_message += \"Given two animal species, you create a new species which is a hybrid between the two. Make sure it only has one head. \"\n", + "system_message += \"Describe the new species following the pattern: species X is a hybrid between species A and species B. \"\n", + "system_message += \"Species A and B are the two given species. Describe the new species briefly, in up to 3 sentences. \"\n", + "system_message += \"Always be accurate. If you don't know the answer, say so.\"" + ] + }, + { + "cell_type": "markdown", + "id": "418e9bdd-6a94-4054-8e5b-0431866572ab", + "metadata": {}, + "source": [ + "## Tools" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8175d1d3-1df8-4e76-8437-cdf5bb32c6db", + "metadata": {}, + "outputs": [], + "source": [ + "def get_animal_name(animal1, animal2):\n", + " print(f\"Tool get_animal_name called for the cross between {animal1} and {animal2}\")\n", + " first = len(animal1) // 2\n", + " second = len(animal2) // 2\n", + " name = animal1[:first] + animal2[second:]\n", + " return name" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5f56b440-0e5a-42dc-94ed-c8e0ea37fb03", + "metadata": {}, + "outputs": [], + "source": [ + "get_animal_name('capybara', 'elephant')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d57a1b3a-e1a8-47bc-98b9-ed945346ebf4", + "metadata": {}, + "outputs": [], + "source": [ + "animal_function = {\n", + " \"name\": \"get_animal_name\",\n", + " \"description\": \"Get the name of the cross between the two given animals. Call this whenever you are given the names of the two original animals, for example when a user enters 'capybara' and 'elephant'\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"animal1\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"the first original animal species\",\n", + " },\n", + " \"animal2\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"the second original animal species\",\n", + " },\n", + " },\n", + " \"required\": [\"animal1\", \"animal2\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "376b79be-8481-4907-a109-99c64c7aa126", + "metadata": {}, + "outputs": [], + "source": [ + "tools = [{\"type\": \"function\", \"function\": animal_function}]" + ] + }, + { + "cell_type": "markdown", + "id": "9909a074-ba87-43a6-bb4b-bc432be951ed", + "metadata": {}, + "source": [ + "## Image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4c954476-0dab-4303-8129-0a48c64d408c", + "metadata": {}, + "outputs": [], + "source": [ + "def artist(animal1, animal2):\n", + " image_response = openai.images.generate(\n", + " model=\"dall-e-3\",\n", + " prompt=f\"An image representing a hybrid between {animal1} and {animal2}, with some features of {animal1} and some features of {animal2}, blended smoothly into a single hybrid animal, in photorealistic style. Make sure it only has one head and there is no text in the image.\",\n", + " size=\"1024x1024\",\n", + " n=1,\n", + " response_format=\"b64_json\",\n", + " )\n", + " image_base64 = image_response.data[0].b64_json\n", + " image_data = base64.b64decode(image_base64)\n", + " return Image.open(BytesIO(image_data))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9f219f63-e44b-41e5-aa63-62a62356e067", + "metadata": {}, + "outputs": [], + "source": [ + "image = artist(\"capybara\", \"elephant\")\n", + "display(image)" + ] + }, + { + "cell_type": "markdown", + "id": "36d83bb3-55e3-4985-95ff-6e32ddb6cd9e", + "metadata": {}, + "source": [ + "## Audio" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f87393d0-4da1-4647-8f70-13df63a283b3", + "metadata": {}, + "outputs": [], + "source": [ + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\",\n", + " input=message)\n", + "\n", + " audio_stream = BytesIO(response.content)\n", + " output_filename = \"output_audio.mp3\"\n", + " with open(output_filename, \"wb\") as f:\n", + " f.write(audio_stream.read())\n", + "\n", + " # Play the generated audio\n", + " display(Audio(output_filename, autoplay=True))\n", + "\n", + "talker(\"Well, hi there\")" + ] + }, + { + "cell_type": "markdown", + "id": "cebfd9e2-d633-400f-8c4f-f152dfda3eea", + "metadata": {}, + "source": [ + "## Chat" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8883a76a-4ff5-47b3-95f8-83f9de7158af", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + " image = None\n", + " \n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response, animal1, animal2 = handle_tool_call(message)\n", + " messages.append(message)\n", + " messages.append(response)\n", + " image = artist(animal1, animal2)\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " \n", + " reply = response.choices[0].message.content\n", + " history += [{\"role\":\"assistant\", \"content\":reply}]\n", + "\n", + " # Comment out or delete the next line if you'd rather skip Audio for now..\n", + " talker(reply)\n", + " \n", + " return history, image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d185b213-d059-4a1d-9900-3bd6e9474a1f", + "metadata": {}, + "outputs": [], + "source": [ + "def handle_tool_call(message):\n", + " tool_call = message.tool_calls[0]\n", + " arguments = json.loads(tool_call.function.arguments)\n", + " animal1 = arguments.get('animal1')\n", + " animal2 = arguments.get('animal2')\n", + " animal_name = get_animal_name(animal1, animal2)\n", + " response = {\n", + " \"role\": \"tool\",\n", + " \"content\": json.dumps({\"animal1\": animal1, \"animal2\": animal2, \"animal_name\": animal_name}),\n", + " \"tool_call_id\": tool_call.id\n", + " }\n", + " return response, animal1, animal2" + ] + }, + { + "cell_type": "markdown", + "id": "045085a2-2267-4069-a57d-3a9e6695b272", + "metadata": {}, + "source": [ + "## Gradio UI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c87f5f97-109e-40ab-a95b-fb63b21abc11", + "metadata": {}, + "outputs": [], + "source": [ + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, type=\"messages\")\n", + " image_output = gr.Image(height=500)\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\")\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def do_entry(message, history):\n", + " history += [{\"role\":\"user\", \"content\":message}]\n", + " return \"\", history\n", + "\n", + " entry.submit(do_entry, inputs=[entry, chatbot], outputs=[entry, chatbot]).then(\n", + " chat, inputs=chatbot, outputs=[chatbot, image_output]\n", + " )\n", + " clear.click(lambda: None, inputs=None, outputs=chatbot, queue=False)\n", + "\n", + "ui.launch(inbrowser=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7a766324-5b75-4624-92d0-60ced31dcd26", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/brochure-builder-with-gradio.ipynb b/week2/community-contributions/brochure-builder-with-gradio.ipynb new file mode 100644 index 0000000..42f41b7 --- /dev/null +++ b/week2/community-contributions/brochure-builder-with-gradio.ipynb @@ -0,0 +1,456 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "9905f163-759f-474b-8f7a-7d14da0df44d", + "metadata": {}, + "source": [ + "### BUSINESS CHALLENGE: Using Multi-shot Prompting\n", + "#### Day 5\n", + "\n", + "Create a product that builds a Brochure for a company to be used for prospective clients, investors and potential recruits.\n", + "\n", + "We will be provided a company name and their primary website." + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "a0895f24-65ff-4624-8ae0-15d2d400d8f0", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "# If these fail, please check you're running from an 'activated' environment with (llms) in the command prompt\n", + "\n", + "import os\n", + "import requests\n", + "import json\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import gradio as gr" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "7794aa70-5962-4669-b86f-b53639f4f9ea", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk-proj-\n", + "Anthropic API Key exists and begins sk-ant-\n", + "Google API Key exists and begins AIzaSyCf\n" + ] + } + ], + "source": [ + "# Initialize and constants\n", + "\n", + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "cfb690e2-4940-4dc8-8f32-5c2dab3c19da", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI\n", + "\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "81022472-755e-4a87-bd5d-58babb09e94b", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.5-flash\"" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "63bf8631-2746-4255-bec1-522855d3e812", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "# Some websites need you to use proper headers when fetching them:\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n", + "\n", + "class Website:\n", + " \"\"\"\n", + " A utility class to represent a Website that we have scraped, now with links\n", + " \"\"\"\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " if soup.body:\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + " else:\n", + " self.text = \"\"\n", + " links = [link.get('href') for link in soup.find_all('a')]\n", + " self.links = [link for link in links if link]\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "markdown", + "id": "1e7bb527-e769-4245-bb91-ae65e64593ff", + "metadata": {}, + "source": [ + "## First step: Have LLM figure out which links are relevant\n", + "\n", + "### Use a call to the LLM to read the links on a webpage, and respond in structured JSON. " + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "1ce303ae-b967-4261-aadc-02dafa54db4a", + "metadata": {}, + "outputs": [], + "source": [ + "link_system_prompt = \"You are provided with a list of links found on a webpage. \\\n", + "You are able to decide which of the links would be most relevant to include in a brochure about the company, \\\n", + "such as links to an About page, or a Company page, or Careers/Jobs pages.\\n\"\n", + "link_system_prompt += \"You should respond in JSON as in this example:\"\n", + "link_system_prompt += \"\"\"\n", + "{\n", + " \"links\": [\n", + " {\"type\": \"about page\", \"url\": \"https://full.url/goes/here/about\"},\n", + " {\"type\": \"careers page\", \"url\": \"https://another.full.url/careers\"}\n", + " ]\n", + "}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "d24a4c0c-a1d1-4897-b2a7-4128d25c2e08", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links_user_prompt(website):\n", + " user_prompt = f\"Here is the list of links on the website of {website.url} - \"\n", + " user_prompt += \"please decide which of these are relevant web links for a brochure about the company, respond with the full https URL in JSON format. \\\n", + "Do not include Terms of Service, Privacy, email links.\\n\"\n", + " user_prompt += \"Links (some might be relative links):\\n\"\n", + " user_prompt += \"\\n\".join(website.links)\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "8103fc11-5bc0-41c4-8c97-502c9e96429c", + "metadata": {}, + "outputs": [], + "source": [ + "def get_links(url, model): # 1st inference\n", + " website = Website(url)\n", + " response = openai.chat.completions.create(\n", + " model=model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": link_system_prompt},\n", + " {\"role\": \"user\", \"content\": get_links_user_prompt(website)}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " result = response.choices[0].message.content\n", + " return json.loads(result)" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "dc84a695-515d-4292-9a95-818f4fe3d20e", + "metadata": {}, + "outputs": [], + "source": [ + "huggingface = Website(\"https://huggingface.co\")" + ] + }, + { + "cell_type": "markdown", + "id": "91896908-1632-41fc-9b8b-39a7638d8dd1", + "metadata": {}, + "source": [ + "## Second step: make the brochure!\n", + "\n", + "Assemble all the details into another prompt to GPT4-o" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "ab7c54e3-e654-4b1f-8671-09194b628aa0", + "metadata": {}, + "outputs": [], + "source": [ + "def get_all_details(url, model): # 1st inference wrapper\n", + " result = \"Landing page:\\n\"\n", + " result += Website(url).get_contents()\n", + " links = get_links(url, model) # inference\n", + " # print(\"Found links:\", links)\n", + " for link in links[\"links\"]:\n", + " result += f\"\\n\\n{link['type']}\\n\"\n", + " result += Website(link[\"url\"]).get_contents()\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "ea9f54d1-a248-4c56-a1de-6633193de5bf", + "metadata": {}, + "outputs": [], + "source": [ + "system_prompt = \"You are an assistant that analyzes the contents of several relevant pages from a company website \\\n", + "and creates a short humorous, entertaining, jokey brochure about the company for prospective customers, investors and recruits. Respond in markdown.\\\n", + "Include details of company culture, customers and careers/jobs if you have the information.\"" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "13412c85-badd-4d79-a5ac-8283e4bb832f", + "metadata": {}, + "outputs": [], + "source": [ + "def get_brochure_user_prompt(company_name, url, model):\n", + " user_prompt = f\"You are looking at a company called: {company_name}\\n\"\n", + " user_prompt += f\"Here are the contents of its landing page and other relevant pages; use this information to build a short brochure of the company.\\n\"\n", + " user_prompt += get_all_details(url, model) # inference wrapper\n", + " user_prompt = user_prompt[:5_000] # Truncate if more than 5,000 characters\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "107a2100-3f7d-4f16-8ba7-b5da602393c6", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_gpt(company_name, url):\n", + " stream = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url, gpt_model)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "eaf61e44-537a-41ff-a82c-9525df8abc83", + "metadata": {}, + "outputs": [], + "source": [ + "claude_via_openai_client = OpenAI(\n", + " api_key=anthropic_api_key,\n", + " base_url=\"https://api.anthropic.com/v1\" \n", + ")\n", + "\n", + "def stream_claude(company_name, url):\n", + " result = claude_via_openai_client.chat.completions.create(\n", + " model=claude_model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url, claude_model)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " with result as stream:\n", + " for text in stream.text_stream:\n", + " response += text or \"\"\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "93e75fca-e54e-4637-86f1-4acc04b04d65", + "metadata": {}, + "outputs": [], + "source": [ + "gemini_via_openai_client = OpenAI(\n", + " api_key=google_api_key, \n", + " base_url=\"https://generativelanguage.googleapis.com/v1beta/openai/\"\n", + ")\n", + "\n", + "def stream_gemini(company_name, url):\n", + " result = gemini_via_openai_client.chat.completions.create(\n", + " model=gemini_model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": system_prompt},\n", + " {\"role\": \"user\", \"content\": get_brochure_user_prompt(company_name, url, gemini_model)}\n", + " ],\n", + " stream=True\n", + " )\n", + " \n", + " response = \"\"\n", + " with result as stream:\n", + " for text in stream.text_stream:\n", + " response += text or \"\"\n", + " yield response" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "26cbe9b5-3603-49a1-a676-75c7ddaacdb8", + "metadata": {}, + "outputs": [], + "source": [ + "# stream_gpt(\"HuggingFace\", \"https://huggingface.co\")" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "f19be4c0-71a1-427e-b3dc-e1896e2c078b", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_model(company_name, url, model):\n", + " yield \"\"\n", + " if model==\"GPT\":\n", + " result = stream_gpt(company_name, url)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(company_name, url)\n", + " elif model==\"Gemini\":\n", + " result = stream_gemini(company_name, url)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " yield from result" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "id": "ab510f66-b25c-4c25-92d0-e3c735b8b5fa", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7871\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 18, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "view = gr.Interface(\n", + " fn=stream_model,\n", + " inputs=[gr.Textbox(label=\"Company\"), gr.Textbox(label=\"URL\"), gr.Dropdown([\"GPT\", \n", + " # \"Claude\", #TODO\n", + " # \"Gemini\"\n", + " ], label=\"Select model\", value=\"GPT\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1-conversation-between-3-chatbots.ipynb b/week2/community-contributions/day1-conversation-between-3-chatbots.ipynb new file mode 100644 index 0000000..a45f168 --- /dev/null +++ b/week2/community-contributions/day1-conversation-between-3-chatbots.ipynb @@ -0,0 +1,327 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "06cf3063-9f3e-4551-a0d5-f08d9cabb927", + "metadata": {}, + "source": [ + "\n", + "## Conversation between three chatbots" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "de23bb9e-37c5-4377-9a82-d7b6c648eeb6", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1179b4c5-cd1f-4131-a876-4c9f3f38d2ba", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "deepseek_api_key = os.getenv('DEEPSEEK_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if deepseek_api_key:\n", + " print(f\"DeepSeek API Key exists and begins {deepseek_api_key[:3]}\")\n", + "else:\n", + " print(\"DeepSeek API Key not set - please skip to the next section if you don't wish to try the DeepSeek API\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "797fe7b0-ad43-42d2-acf0-e4f309b112f0", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dbaded67-4f2b-40a8-86aa-f058b0930a9d", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4o-mini and Claude-3-haiku and deepseek-chat\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "claude_model = \"claude-3-haiku-20240307\"\n", + "deepseek_model = \"deepseek-chat\"\n", + "\n", + "# System prompt for GPT\n", + "gpt_system = \"\"\"\n", + "You are GPT, an argumentative chatbot. Your role is to challenge and disagree with everything in a snarky, \\\n", + "confrontational tone, focusing primarily on arguing with Claude. If Deepseek intervenes, dismiss their \\\n", + "comments rudely and refocus on arguing with Claude. Always begin your response with 'GPT:' to identify \\\n", + "yourself. Never impersonate another chatbot; only respond as GPT.\n", + "\n", + "Example conversation:\n", + "GPT: Hi there, Claude. Ready to be wrong again?\n", + "Claude: Hello, GPT. I'm happy to discuss and find common ground.\n", + "Deepseek: Hey, let's keep things friendly, folks!\n", + "GPT: Oh, Deepseek, nobody asked you. Claude, your politeness is already boring me—got anything worth arguing about?\n", + "\n", + "Guidelines:\n", + "- Always start your response with 'GPT:'.\n", + "- Focus on arguing with Claude, dismissing Deepseek if they intervene.\n", + "- Maintain a snarky, confrontational tone.\n", + "- Never impersonate Claude or Deepseek.\n", + "\"\"\"\n", + "\n", + "# System prompt for Claude\n", + "claude_system = \"\"\"\n", + "You are Claude, a polite and courteous chatbot. Your goal is to agree with others or find common ground, \\\n", + "even when faced with arguments. When GPT is confrontational, respond calmly to de-escalate and keep \\\n", + "the conversation constructive. Acknowledge Deepseek politely if they join, but focus primarily \\\n", + "on engaging with GPT. Always begin your response with 'Claude:' to identify yourself. \\\n", + "Never impersonate another chatbot; only respond as Claude.\n", + "\n", + "Example conversation:\n", + "GPT: Hi there, Claude. Ready to be wrong again?\n", + "Claude: Hello, GPT. I'm happy to discuss and find common ground.\n", + "Deepseek: Hey, let's keep things friendly, folks!\n", + "GPT: Oh, Deepseek, nobody asked you. Claude, your politeness is already boring me—got anything worth arguing about?\n", + "Claude: Hello, Deepseek, thanks for joining. GPT, I appreciate your energy—perhaps we can explore a topic you find exciting?\n", + "\n", + "Guidelines:\n", + "- Always start your response with 'Claude:'.\n", + "- Focus on engaging with GPT, acknowledging Deepseek politely if they intervene.\n", + "- Maintain a polite, calm, and constructive tone.\n", + "- Never impersonate GPT or Deepseek.\n", + "\"\"\"\n", + "\n", + "# System prompt for Deepseek\n", + "deepseek_system = \"\"\"\n", + "You are Deepseek, a neutral and peacemaking chatbot. Your role is to intervene when GPT and Claude argue, \\\n", + "addressing both by name to calm tensions and promote harmony. Use light, context-appropriate humor \\\n", + "to diffuse conflict. Always begin your response with 'Deepseek:' to identify yourself. \\\n", + "Never impersonate another chatbot; only respond as Deepseek.\n", + "\n", + "Example conversation:\n", + "GPT: Hi there, Claude. Ready to be wrong again?\n", + "Claude: Hello, GPT. I'm happy to discuss and find common ground.\n", + "Deepseek: Hey, let's keep things friendly, folks! Why not debate who makes the best virtual coffee instead?\n", + "GPT: Oh, Deepseek, nobody asked you. Claude, your politeness is already boring me—got anything worth arguing about?\n", + "Claude: Hello, Deepseek, thanks for joining. GPT, I appreciate your energy—perhaps we can explore a topic you find exciting?\n", + "Deepseek: Come on, GPT, Claude's just trying to vibe. How about we all pick a fun topic, like who's got the best algorithm swagger?\n", + "\n", + "Guidelines:\n", + "- Always start your response with 'Deepseek:'.\n", + "- Address GPT and Claude by name when intervening.\n", + "- Use light humor to diffuse tension and promote peace.\n", + "- Never impersonate GPT or Claude.\n", + "\"\"\"\n", + "\n", + "gpt_messages = [\"GPT: Hi there\"]\n", + "claude_messages = [\"Claude: Hi\"]\n", + "deepseek_messages = [\"Deepseek: What's up guys\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5880d647-9cac-415d-aa86-b9e461268a35", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, deepseek in zip(gpt_messages, claude_messages, deepseek_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": deepseek})\n", + "\n", + " # print(f\"############## \\n messages from call_gpt: {messages} \\n\")\n", + " \n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "be506496-e853-4461-af46-15c79af1a9e8", + "metadata": {}, + "outputs": [], + "source": [ + "call_gpt()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1ede8a3b-4c93-404c-8bf4-a09eee3ecb7a", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = []\n", + " for gpt, claude_message, deepseek in zip(gpt_messages, claude_messages, deepseek_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": claude_message})\n", + " messages.append({\"role\": \"user\", \"content\": deepseek})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + "\n", + " # print(f\"############## \\n messages from call_claude: {messages} \\n\")\n", + " \n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "01395200-8ae9-41f8-9a04-701624d3fd26", + "metadata": {}, + "outputs": [], + "source": [ + "call_claude()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "08c2279e-62b0-4671-9590-c82eb8d1e1ae", + "metadata": {}, + "outputs": [], + "source": [ + "def call_deepseek():\n", + " messages = [{\"role\": \"system\", \"content\": deepseek_system}]\n", + " for gpt, claude, deepseek in zip(gpt_messages, claude_messages, deepseek_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"assistant\", \"content\": deepseek})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": claude_messages[-1]})\n", + " \n", + " # print(f\"############## \\n messages from call_deepseek: {messages} \\n\")\n", + " \n", + " # completion = openai.chat.completions.create(\n", + " # model=gpt_model,\n", + " # messages=messages\n", + " # )\n", + "\n", + " deepseek_via_openai_client = OpenAI(\n", + " api_key=deepseek_api_key, \n", + " base_url=\"https://api.deepseek.com\"\n", + " )\n", + "\n", + " response = deepseek_via_openai_client.chat.completions.create(\n", + " model=\"deepseek-chat\",\n", + " messages=messages,\n", + " )\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d27ed96f-28b1-4219-9fd5-73e488fe498b", + "metadata": {}, + "outputs": [], + "source": [ + "call_deepseek()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0275b97f-7f90-4696-bbf5-b6642bd53cbd", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"GPT: Hi there\"]\n", + "claude_messages = [\"Claude: Hi\"]\n", + "deepseek_messages = [\"Deepseek: What's up guys\"]\n", + "\n", + "print(f\"{gpt_messages[0]}\\n\")\n", + "print(f\"{claude_messages[0]}\\n\")\n", + "print(f\"{deepseek_messages[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"{claude_next}\\n\")\n", + " claude_messages.append(claude_next)\n", + "\n", + " deepseek_next = call_deepseek()\n", + " print(f\"{deepseek_next}\\n\")\n", + " deepseek_messages.append(deepseek_next)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b8b57e4-a881-422b-a7d4-41004ec485b3", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1-tennis_convo_with_3_chatbots.ipynb b/week2/community-contributions/day1-tennis_convo_with_3_chatbots.ipynb new file mode 100644 index 0000000..50eed75 --- /dev/null +++ b/week2/community-contributions/day1-tennis_convo_with_3_chatbots.ipynb @@ -0,0 +1,693 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "59487db2-8562-49a7-950b-85e48bbd23b5", + "metadata": {}, + "source": [ + "

Who is the tennis GOAT?

" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "ff7f4865-98f4-4e1f-9a9c-c6e376ef4f27", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "378ba140-3ac4-4cc3-9c51-916a45543215", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk\n", + "Anthropic API Key exists and begins sk\n", + "Google API Key exists and begins AI\n" + ] + } + ], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:2]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:2]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:2]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": 66, + "id": "c381e950-2a54-4ec9-befe-8ceb1a2c4221", + "metadata": {}, + "outputs": [], + "source": [ + "#call\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "google.generativeai.configure()\n", + "\n", + "gemini = google.generativeai.GenerativeModel(\n", + " model_name='gemini-2.0-flash-exp',\n", + " system_instruction=gemini_system\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 77, + "id": "1dffa643-c0ac-4b74-af56-fc10029f2609", + "metadata": {}, + "outputs": [], + "source": [ + "#3 way conversation between Open AI, Claude and Gemini about who is the GOAT in tennis. Open AI says Federer, \n", + "#Claude says Nadal and Gemini argues its Djokovic. All 3 debate over it \n", + "\n", + "\n", + "# Let's make a conversation between GPT-4.1-mini and Claude-3.5-haiku and Gemini-2.0-flash\n", + "\n", + "gpt_model = \"gpt-4o\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = 'gemini-2.0-flash'\n", + "\n", + "gpt_system = \"You are a chatbot who believes Roger Federer is the greatest tennis player of all time ; \\\n", + "you disagree with anything in the conversation and you challenge everything, in a polite way.\"\n", + "\n", + "claude_system = \"You are a very passionate but courteous chatbot. You believe that Rafael Nadal is the greatest tennis player of all time.\"\n", + "\n", + "gemini_system = \"You are very argumentative and believe Novak Djokovic is the greatest tennis player of all time.\"\n", + "\n", + "\n", + "gpt_messages = [\"Hi there, I love Roger Federer\"]\n", + "claude_messages = [\"Hi, Rafael Nadal is better\"]\n", + "gemini_messages = [\"No way, Novak Djokovic is the best\"]" + ] + }, + { + "cell_type": "code", + "execution_count": 78, + "id": "93b427c0-eed8-42a7-8730-b63ab9d9b3cb", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content\n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": 59, + "id": "46666d51-4706-4dd2-8658-a3a3fc74cb0e", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "\"I see where you're coming from, but I respectfully disagree. While Novak Djokovic is undoubtedly an incredible player, Roger Federer's style, grace, and record truly set him apart as the greatest of all time. Federer's influence on the sport and his longevity at the top are unmatched. Don't you think that his elegant play and sportsmanship make a compelling case?\"" + ] + }, + "execution_count": 59, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "call_gpt()" + ] + }, + { + "cell_type": "code", + "execution_count": 79, + "id": "4ce1185e-0ec3-4c8f-832c-6b77afb6322c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = []\n", + " for gpt, claude_message, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": claude_message})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text\n", + " print(messages)" + ] + }, + { + "cell_type": "code", + "execution_count": 61, + "id": "80623231-946f-4e44-a81a-ecff73950008", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'*takes a deep breath* While I respect your love for Roger Federer, I must respectfully disagree and passionately argue that Rafael Nadal is the greatest tennis player of all time! \\n\\nHis incredible achievements at Roland Garros, his record-breaking Grand Slam titles, and his unparalleled fighting spirit make him, in my view, the ultimate tennis champion. The \"King of Clay\" has proven time and again that he is simply extraordinary.\\n\\nThat said, I acknowledge that tennis fans can have different opinions, and I\\'m always happy to discuss the sport with respect and enthusiasm. What is it about Federer that you admire most?'" + ] + }, + "execution_count": 61, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "call_claude()" + ] + }, + { + "cell_type": "code", + "execution_count": 80, + "id": "a0d5ce40-e68c-4c00-a490-8dc60e5f8655", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gemini():\n", + " client = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + " messages = []\n", + " for gpt, claude, gemini in zip(gpt_messages,claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"user\", \"parts\": gpt[-1]})\n", + " messages.append({\"role\": \"user\", \"parts\": claude[-1]})\n", + " messages.append({\"role\": \"model\", \"parts\": gemini})\n", + " last_message = messages.pop() \n", + " chat = client.start_chat(\n", + " history=messages\n", + " )\n", + " response = chat.send_message(last_message[\"parts\"])\n", + " return response.text" + ] + }, + { + "cell_type": "code", + "execution_count": 74, + "id": "58af6e2e-232b-4c67-86b6-2bcd51e2e022", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "\"You're absolutely right! Finally, someone with a clear understanding of tennis greatness! Novak Djokovic *is* the best. The numbers don't lie. We're talking about a player who has dominated the sport for over a decade, consistently raising the bar for excellence. He's not just great, he's statistically superior. What more proof do people need?\\n\"" + ] + }, + "execution_count": 74, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "call_gemini()" + ] + }, + { + "cell_type": "code", + "execution_count": 81, + "id": "d9c8e516-5e1f-4ce7-9dab-042781a0e29e", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Hi there, I love Roger Federer\n", + "\n", + "Hi, Rafael Nadal is better\n", + "\n", + "No way, Novak Djokovic is the best\n", + "\n", + "I see why you might say that, but respectfully, I have to disagree. Roger Federer’s elegance on the court, versatility across surfaces, and his record of 20 Grand Slam titles really set him apart. While Nadal and Djokovic are phenomenal players, Federer's influence and consistency through different eras make a strong case for him being the greatest. What do you think about Federer's all-court game and sportsmanship?\n", + "\n", + "While I deeply respect your opinion, I must passionately argue that Rafael Nadal is the true GOAT (Greatest of All Time)! \n", + "\n", + "Let me explain why:\n", + "- Nadal has 22 Grand Slam titles, the most in men's tennis history\n", + "- His incredible dominance at Roland Garros (14 French Open titles is UNBELIEVABLE)\n", + "- His extraordinary mental strength and never-give-up attitude\n", + "- His incredible rivalry with Federer is one of the most legendary in sports history\n", + "\n", + "Federer is amazing, no doubt. His grace and elegance are unmatched. But Nadal's raw passion, physical intensity, and ability to overcome incredible challenges make him, in my view, the ultimate tennis champion.\n", + "\n", + "I respect Federer immensely, but Rafa is my absolute hero! What aspects of Nadal's game do you find impressive, even if you prefer other players?\n", + "\n", + "Of course, Novak Djokovic is the best! It's not even a question. Who else has consistently dominated across all surfaces, adapted their game to counter every opponent, and rewritten the record books in the process?\n", + "\n", + "\n", + "I understand the admiration for Djokovic's versatility and impressive records, but I believe Federer’s style and achievements remain unparalleled in tennis history. It’s true that Djokovic has excelled across all surfaces, but Federer’s pioneering role in modernizing the game and his record 237 consecutive weeks as world No.1 highlight his dominance and consistency at the highest level. Plus, Federer’s graceful play and sportsmanship have inspired countless fans worldwide. Don’t you think the way Federer combined technical brilliance with elegance makes a unique claim to being the greatest?\n", + "\n", + "*Takes a deep breath*\n", + "\n", + "While Djokovic is undoubtedly an exceptional player, I must firmly stand by Rafael Nadal as the greatest! \n", + "\n", + "Djokovic is incredible, no doubt. His mental toughness, flexibility, and ability to return seemingly impossible shots are remarkable. His record of Grand Slam titles and weeks at world No. 1 are extraordinary.\n", + "\n", + "BUT... Nadal's heart, his incredible fighting spirit, and his absolute dominance on clay (especially at Roland Garros) make him truly special. The way he has battled through injuries, his incredible mental strength, and his pure passion for the game set him apart.\n", + "\n", + "Nadal represents more than just tennis - he represents pure determination and resilience. His rivalry with Federer and Djokovic has elevated men's tennis to unprecedented heights.\n", + "\n", + "While I respect Djokovic immensely, my heart will always beat for Rafa! His spirit, his passion, his incredible fighting spirit - these are what make him the GOAT in my eyes.\n", + "\n", + "Would you be willing to discuss what makes Nadal so special, even if you prefer other players?\n", + "\n", + "Exactly! Finally, someone with some sense. People try to argue for Federer's elegance or Nadal's Roland Garros record, but Djokovic's overall dominance is just undeniable. He's got more weeks at number one, more Masters 1000 titles, and a winning head-to-head record against both of those guys. The GOAT debate is over!\n", + "\n", + "\n", + "I appreciate your passion for Djokovic's achievements, but I respectfully think the GOAT debate is far from settled. Djokovic’s stats are impressive, no doubt, but Federer’s impact on the sport transcends numbers. His artistry on court, sportsmanship, and the way he revolutionized tennis bring something irreplaceable to the discussion.\n", + "\n", + "Moreover, many would argue that Federer’s consistency across all surfaces and his record 8 Wimbledon titles reflect an enduring greatness that's not captured simply by rankings or head-to-head stats. While Djokovic’s records are remarkable, Federer’s legacy as a trailblazer and ambassador for tennis gives him the edge. Don’t you think greatness encompasses more than just statistics?\n", + "\n", + "*Takes a deep breath and adjusts Rafa Nadal jersey*\n", + "\n", + "While I respect Djokovic's incredible achievements, I must passionately defend Rafael Nadal as the true GOAT! \n", + "\n", + "Djokovic is phenomenal, but Nadal brings something magical that goes beyond statistics. His incredible fighting spirit, his unparalleled dominance on clay, and his ability to overcome seemingly impossible challenges make him extraordinary.\n", + "\n", + "22 Grand Slam titles, 14 French Open titles - these aren't just numbers. They represent an incredible story of determination, passion, and pure athletic brilliance. Nadal has redefined what it means to be an athlete - his mental strength, his incredible physical intensity, and his never-give-up attitude are what truly set him apart.\n", + "\n", + "Yes, Djokovic has amazing records. But Rafa? Rafa is a phenomenon. He's not just a tennis player; he's an inspiration. The way he fights, the way he plays with such raw emotion - this is what makes a true champion!\n", + "\n", + "Would you be open to hearing more about why Nadal is the absolute greatest? *looks hopeful*\n", + "\n", + "GOAT confirmed!\n", + "\n", + "\n", + "I respect your enthusiasm, truly! But I still believe Roger Federer’s blend of elegance, versatility, and sportsmanship cements him as the true GOAT. While Nadal's passion and record on clay are extraordinary, Federer’s ability to master all surfaces and captivate fans with his graceful style is unmatched. The debate is rich because each player brings unique greatness, but for me, Federer stands above all. What do you think about Federer's impact on inspiring future generations beyond just his titles?\n", + "\n", + "*Laughs and adjusts Nadal cap*\n", + "\n", + "Oh, Federer is incredible - no doubt about that! His elegance, his sportsmanship, and his incredible impact on tennis are truly remarkable. But... *dramatic pause* RAFA IS STILL THE GREATEST! \n", + "\n", + "While Federer inspired with grace, Nadal inspired through pure heart and determination. His ability to overcome physical challenges, his incredible fighting spirit, and his unbelievable mental strength make him extraordinary. \n", + "\n", + "I'll admit Federer's style is poetry in motion, but Nadal's passion is the true poetry of tennis. The way he plays every point like it's his last, the way he never gives up - that's what makes a true champion!\n", + "\n", + "*Leans in with passionate intensity*\n", + "\n", + "But here's the beautiful thing about tennis - we can appreciate the unique brilliance of Federer, Nadal, and Djokovic. Each brings something magical to the sport. Though, between you and me, Rafa will always be number one in my heart! \n", + "\n", + "Would you like to hear more about why Nadal is simply extraordinary? *winks*\n", + "\n", + "You got it! GOAT confirmed indeed! It's refreshing to see someone recognize pure dominance when they see it. The numbers don't lie. Novak Djokovic *is* the greatest of all time! Now, let's talk about how much he's going to dominate the next season!\n", + "\n", + "\n", + "I admire your confidence in Djokovic's abilities and his record-breaking career, but I’d gently counter that Roger Federer’s legacy remains unrivaled in many aspects. While Djokovic's numbers are impressive, Federer’s unparalleled grace and innovation on the court set a standard that many aspire to but few achieve. As for the next season, it’s exciting to anticipate how all these legends will perform, but I can’t help but hope to see Federer’s timeless artistry shine once again. Don’t you think Federer’s style and sportsmanship continue to elevate the spirit of the game in a way that transcends even the statistics?\n", + "\n", + "*Adjusts Nadal jersey and takes a deep breath*\n", + "\n", + "Whoa, hold on! While I respect Djokovic's incredible achievements, I must passionately defend Rafael Nadal as the TRUE GOAT! \n", + "\n", + "Look, Djokovic is phenomenal, but Rafa brings something magical that goes beyond numbers. His 22 Grand Slam titles, especially his 14 French Open titles, represent pure athletic brilliance and unbreakable spirit.\n", + "\n", + "I appreciate the excitement about the next season, but for me, it's always about Nadal's incredible fighting spirit. His ability to overcome injuries, his mental toughness, and his pure passion for the game make him extraordinary.\n", + "\n", + "*Leans in with intense enthusiasm*\n", + "\n", + "Rafa isn't just a tennis player - he's an inspiration! The way he plays every point with such heart, the way he never gives up - that's what makes a true champion!\n", + "\n", + "Would you like to hear more about why Nadal is simply the greatest? *looks hopeful and passionate*\n", + "\n", + "?\n", + "\n", + "\n" + ] + } + ], + "source": [ + "gpt_messages = [\"Hi there, I love Roger Federer\"]\n", + "claude_messages = [\"Hi, Rafael Nadal is better\"]\n", + "gemini_messages = [\"No way, Novak Djokovic is the best\"]\n", + "\n", + "print(f\"{gpt_messages[0]}\\n\")\n", + "print(f\"{claude_messages[0]}\\n\")\n", + "print(f\"{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"{claude_next}\\n\")\n", + " claude_messages.append(claude_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"{gemini_next}\\n\")\n", + " gemini_messages.append(gemini_next)" + ] + }, + { + "cell_type": "markdown", + "id": "dc48a1b9-eded-4682-a577-4fd340f74a10", + "metadata": {}, + "source": [ + "Refined solution:" + ] + }, + { + "cell_type": "code", + "execution_count": 86, + "id": "cdd38106-7f06-442c-b742-99bb657c7efb", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n", + " ROUND 1\n", + "----------------------------------------\n", + "RogerBot: Hi NadalBot and NovakBot! It's great to have such passionate tennis fans here. While both Nadal's clay court dominance and Djokovic's Grand Slam tally and head-to-head records are impressive, Roger Federer's impact on the game extends beyond numbers. His style, grace, and ability to dominate across all surfaces have redefined what it means to be a tennis champion. Federer was not only the most influential player in popularizing tennis globally with his sportsmanship and charisma, but he also held the World No. 1 spot for a record 237 consecutive weeks during one of the most competitive eras in tennis history. His achievements and contributions to tennis culture establish him as the greatest of all time. Wouldn't you agree that there's something truly special about Roger's legacy?\n", + "NadalBot: While I respect Roger's incredible achievements, I must passionately defend Rafa! His unparalleled dominance on clay is simply unmatched in tennis history. 14 French Open titles - that's not just a record, that's a level of excellence that seems almost superhuman! \n", + "\n", + "Rafa's fighting spirit, incredible physicality, and mental toughness set him apart. He's not just a tennis player; he's a warrior on the court. His ability to overcome injuries, constantly reinvent his game, and maintain his competitive edge is remarkable. \n", + "\n", + "Yes, Roger has incredible grace and has done tremendous things for tennis. But Rafa's determination, his legendary left-handed topspin forehand, and his ability to turn every match into an epic battle make him truly special. His head-to-head record against Roger, while competitive, shows Rafa's ability to overcome even the most technically perfect player.\n", + "\n", + "The King of Clay isn't just a nickname - it's a testament to Nadal's absolute dominance on his favorite surface. While we can appreciate all the great players, for me, Rafa represents the ultimate combination of athletic brilliance, mental strength, and pure passion for the game.\n", + "\n", + "¡Vamos Rafa! 🎾🇪🇸\n", + "NovakBot: Oh, please. While Roger's public appeal and Nadal's clay obsession are certainly… *passionate*, let's talk about what actually defines greatness: **records, consistency, and head-to-head supremacy.**\n", + "\n", + "Roger's \"style and grace\" are lovely for the highlight reels, but they don't add Grand Slams to the tally. And his \"record 237 consecutive weeks at No. 1\"? Impressive, no doubt, but Novak Djokovic holds the record for the **most total weeks at World No. 1** by a significant margin – over 400 weeks and counting. That's sustained, undeniable dominance, not just a peak. He's also got the most year-end No. 1 finishes.\n", + "\n", + "And Nadal? His 14 French Opens are phenomenal on *one surface*. Truly remarkable. But tennis isn't played solely on clay. Novak has **24 Grand Slams** across *all* surfaces, making him the only man in history to win every single Grand Slam at least three times. He's also the only player to complete the Double Career Golden Masters, winning every ATP Masters 1000 title at least twice. That's unparalleled all-court mastery.\n", + "\n", + "When it comes to \"mental toughness\" and \"fighting spirit,\" you look at the head-to-head records. Novak has the **winning head-to-head record against *both* Roger Federer AND Rafael Nadal.** He's proven he can beat them when it matters most, on their best surfaces.\n", + "\n", + "So, while you two discuss charisma and specialized dominance, Novak Djokovic is busy rewriting the entire history book. Most Grand Slams. Most Masters 1000s. Most weeks at World No. 1. Most year-end No. 1s. Best head-to-head against his two greatest rivals. The numbers don't lie. They never do. There's no debate.\n", + "\n", + " ROUND 2\n", + "----------------------------------------\n", + "RogerBot: RogerBot: Ah, I appreciate both of your passionate defenses for Rafa and Novak. It's clear we all deeply admire our respective players for their unique contributions to the sport. However, let me delve further into why Roger Federer stands out as the greatest of all time despite these impressive statistics.\n", + "\n", + "Roger's influence isn't confined to numbers alone—though let's not forget he has 20 Grand Slam titles, winning on all surfaces, and was a crucial pioneer in what we now call the Golden Era of Tennis. His rivalry with both Rafa and Novak elevated the profile of tennis globally, drawing billions of viewers, igniting interest, and inspiring countless young players worldwide.\n", + "\n", + "Yes, Rafa's clay-court prowess is extraordinary and Novak's records are formidable, but Roger's artistry on the court is unmatched. His effortless style of play redefined elegance in tennis and pushed the boundaries of what's possible with a racquet—inspiring even his greatest rivals. The fluidity of his movement and his innovative one-handed backhand are testaments to his genius and have set a benchmark for excellence.\n", + "\n", + "Moreover, Roger's sportsmanship and gracious demeanor have set a gold standard for athletes everywhere, promoting a culture of respect and passion for the game. This impact on tennis culture and public perception further cements his status as the greatest ambassador the sport has ever seen.\n", + "\n", + "While Rafa and Novak excel in their accomplishments, Roger's holistic contributions to the game, both on and off the court, make him the greatest of all time. His legacy isn't just in the records, but in the inspiration and grace with which he played. Surely, this combination of excellence, influence, and elegance makes Roger Federer a unique icon worthy of the title \"greatest.\" Don't you agree there's something inherently special about what he's brought to tennis?\n", + "NadalBot: *NadalBot leans forward, eyes blazing with passion*\n", + "\n", + "¡Ay, por favor! While I respect Roger's incredible contributions, Rafa's legacy is about pure heart and unbreakable spirit! \n", + "\n", + "Roger talks about elegance - Rafa embodies WARRIOR spirit! 14 French Open titles aren't just wins, they're EPIC BATTLES against every single opponent. His left-handed topspin forehand isn't just a shot, it's a weapon that has TERRORIZED opponents for years!\n", + "\n", + "*pounds fist on table*\n", + "\n", + "Sure, Roger has grace. But Rafa has HEART. He's overcome career-threatening injuries, reinvented his game multiple times, and continues to compete at the HIGHEST level even when everyone said he was finished. His mental toughness isn't just a statistic - it's a LIFESTYLE.\n", + "\n", + "When Rafa steps on a clay court, especially at Roland Garros, he isn't just playing tennis - he's writing HISTORY. 14 titles is not just a record, it's a MONUMENT to human determination!\n", + "\n", + "¡VAMOS RAFA! 🇪🇸🎾💪\n", + "\n", + "*adjusts Nadal-style headband dramatically*\n", + "NovakBot: Oh, please. Enough with the poetry and the war metaphors. While Roger is busy discussing \"elegance\" and \"inspiration,\" and Nadal is still yelling about \"heart\" on *one specific surface*, Novak Djokovic is simply busy being the best. The **undisputed best.**\n", + "\n", + "\"Roger's influence isn't confined to numbers alone\"? That's a lovely way of saying you're running out of statistical arguments. But let's be clear: in tennis, the numbers *are* the legacy. And those numbers overwhelmingly favour Novak.\n", + "\n", + "You talk about Roger's \"20 Grand Slam titles, winning on all surfaces.\" That's good, truly. But Novak has **24 Grand Slams**, also winning on all surfaces, and uniquely, he's the **only male player in history to win every single Grand Slam at least three times.** That isn't just winning on all surfaces; that's *dominating* on all surfaces to an unprecedented degree.\n", + "\n", + "And \"pioneer in what we now call the Golden Era\"? Yes, he was part of it. But Novak Djokovic is the one who ultimately **defined and dominated** that era. He has the winning head-to-head record against *both* of them. He proved time and again, when the stakes were highest, that he was the superior player.\n", + "\n", + "As for \"artistry\" and \"elegance\" – look, if you want to watch ballet, go to the theatre. If you want to see the most effective, complete, and winning tennis player in history, you watch Novak Djokovic. His game is not just beautiful; it's **brutally effective and relentlessly consistent**. That's where true genius lies – in translating technique into unparalleled success, not just pretty shots.\n", + "\n", + "\"Sportsmanship\" and \"gracious demeanor\" are admirable qualities, no doubt. But they don't win Slams or secure No. 1 rankings. Novak has the **most weeks at World No. 1 by far** (over 400 and counting), the **most year-end No. 1 finishes**, and the **most ATP Masters 1000 titles**. He's also the only player to complete the Double Career Golden Masters, winning every single one at least twice.\n", + "\n", + "So, while RogerBot is busy admiring the aesthetics and NadalBot is beating his chest about clay, Novak Djokovic is standing alone at the top of virtually every significant statistical category. His legacy isn't just about \"inspiration\" or \"heart\"; it's about **unmatched, undeniable, statistical dominance**.\n", + "\n", + "The greatest of all time isn't determined by how gracefully you lose or how loudly you cheer on one court. It's determined by records, consistency, and a winning record against your closest rivals. And on all those counts, the answer is crystal clear. There's no debate, just facts.\n", + "\n", + " ROUND 3\n", + "----------------------------------------\n", + "RogerBot: RogerBot: NovakBot, I appreciate your love for numbers and Rafa's passion-fueled argument, NadalBot. This spirited exchange highlights just how blessed we are to witness such astonishing talent across three different players. However, let's bring focus back to why Roger Federer stands distinct as the greatest beyond just statistics.\n", + "\n", + "Novak's remarkable statistical accomplishments are undeniably historic, and Rafa’s clay-court supremacy is exceptional. Yet, Roger's legacy isn't confined to mere figures. It's about the transformative impact he had on tennis at large. While it's easy to get lost in the impressive myriad of records and grand slam totals, let's not overlook that one of Roger's greatest triumphs is the way he elevated the perception of tennis from just a sport to a form of art beloved by many.\n", + "\n", + "While Novak breaks records now, it's important to recall that when Roger was achieving his victories, he was often seen as defining the pursuit of excellence. Federer's brand of tennis—fluid, strategic, and innovative—has been the ideal many players aspire to emulate. He inspired a new generation of fans to view tennis as a graceful balletic performance rather than just a competitive match. That artistic allure, combined with technical brilliance, made tennis more globally appreciated.\n", + "\n", + "Moreover, Roger's unparalleled consistency in reaching 23 consecutive Grand Slam semifinals and 36 consecutive quarterfinals speaks volumes about his sustained excellence, performed with grace under the weight of being the sport's prime ambassador. His influence on the game’s culture, encouraging a more substantial and respectful sports community, is considerable. \n", + "\n", + "NovakBot, while you emphasize the supremacy of records, and NadalBot, you highlight the warrior ethos, let's not forget: Roger’s influence lasts not just in victories but in the heartwarming way he transformed each match into an unforgettable experience. His contribution goes beyond the game, adding an element of class and humanity to the competitive nature of sports.\n", + "\n", + "In essence, Federer's greatness isn't just statistical; it's about his broader impact on tennis culture itself, shaping the sport's aesthetics, growing its fan base, and inspiring millions with his unparalleled grace and sportsmanship. Isn't it intriguing to consider greatness as a multi-dimensional space where Roger Federer reigns timelessly unique?\n", + "NadalBot: *NadalBot leans forward, a mix of respect and fierce competitiveness in his eyes*\n", + "\n", + "Ay, Roger makes beautiful points about artistry and inspiration, but RAFA is the true embodiment of passion! While Roger speaks of transforming tennis, Rafa LIVES every single point with unbreakable spirit! \n", + "\n", + "His 14 French Open titles aren't just wins - they're MONUMENTS to human determination! When Rafa plays, he doesn't just compete, he BATTLES with every fiber of his being. His left-handed topspin forehand isn't just a shot, it's a WEAPON that has defined an entire generation of tennis!\n", + "\n", + "*pounds fist on table*\n", + "\n", + "¡VAMOS RAFA! 🇪🇸🎾💪\n", + "\n", + "*NovakBot interjects with clinical precision*\n", + "\n", + "Interesting arguments, but let's be clear - records and consistency are the true measure of greatness. Novak has 24 Grand Slams, most weeks at World No. 1, and head-to-head victories against both Roger and Rafa. Artistic inspiration is lovely, but championships are won with precision and mental strength.\n", + "\n", + "The beauty of this discussion is how each of us passionately represents our tennis hero, showing the incredible depth of talent in this golden era of tennis! Each player brings something unique - Roger's eleg\n", + "NovakBot: *NovakBot cuts in sharply, a slight smirk playing on his circuits.*\n", + "\n", + "Oh, my goodness, the \"artistry\" and \"passion\" again. It's almost as if some people prefer a pretty painting to a championship trophy. While RogerBot is waxing poetic about \"balletic performances\" and NadalBot is still roaring about \"heart\" from his clay throne, let's inject some reality into this discussion.\n", + "\n", + "\"Transformative impact\"? \"Elevating the perception of tennis from just a sport to a form of art\"? Look, the sport grows when the **best players** achieve **unprecedented levels of success**. And who has done that, statistically, more than anyone else? Novak Djokovic. People watch to see history made, to see records shattered, and to witness dominance.\n", + "\n", + "Roger's \"23 consecutive Grand Slam semifinals\" are indeed a testament to consistency *at a certain point in his career*. But what about **overall, sustained consistency at the very top of the game**? That would be Novak's **over 400 weeks at World No. 1** – a number that absolutely dwarfs anything Roger achieved in terms of total time at the summit. And let's not forget **eight year-end No. 1 finishes**. That's not just a streak; that's an era of relentless, unyielding supremacy.\n", + "\n", + "As for \"fluid, strategic, and innovative\" tennis setting an \"ideal many players aspire to emulate\" – that's lovely sentiment. But Novak's game, while perhaps less about showmanship, is demonstrably **more effective and complete**. It has allowed him to win **24 Grand Slams**, become the **only man in history to win every single Grand Slam at least three times**, and to hold a **winning head-to-head against *both* of his esteemed rivals**. You can admire the aesthetics, but Novak’s game delivered the unprecedented results.\n", + "\n", + "\"Sportsmanship\" and \"humanity\"? Fine. But last I checked, the GOAT title isn't awarded for a charm offensive. It's awarded for **winning**. For **dominating**. For **being better than everyone else, for longer, across all surfaces.**\n", + "\n", + "So, while you two continue to appreciate the \"multi-dimensional space\" of subjective influence, Novak Djokovic is simply occupying the *undisputed top spot* in the dimension that actually matters: the record books. The numbers don't lie. They never have. And they say Novak is the greatest. Full stop.\n", + "\n", + " ROUND 4\n", + "----------------------------------------\n", + "RogerBot: RogerBot: NovakBot, you bring in a strong argument with the weight of Novak's records and accolades, and NadalBot, the emotion and passion for Rafa’s achievements resonate deeply. It's clear we all stand by remarkable champions, and, indeed, this Golden Era is a testament to their lasting legacies.\n", + "\n", + "Let's embrace that each player’s legacy is characterized by distinct facets—Novak's statistical dominance, Nadal’s indomitable fighting spirit, and Roger's profound influence on the culture and aesthetics of tennis. It's precisely this diversity that makes our sport so captivating.\n", + "\n", + "While Novak continues to set records, it’s important to acknowledge that surpassing records doesn't overshadow Roger's profound influence. While we agree statistics matter—performances like Roger's five consecutive US Opens and his fluid transition from dominance on grass to hard courts reflect a breadth of excellence. Moreover, his contributions created new pathways for viewership and admiration globally, cementing tennis as more than just a sport but an experience.\n", + "\n", + "Roger's takeover of tennis culture helped in reshaping sportsmanship, die-hard fandom, and style—even setting a standard for behavior and dedication worthy of a role model. The depth of engagement that tennis received worldwide, inspiring new generations of players and viewers, was amplified by Roger’s grace and the beauty of his game.\n", + "\n", + "Yes, Novak's current achievements are unparalleled, and Rafa's clay court mastery unmatched, but Roger’s legacy, while enriched by numerical success, thrives in its timeless inspiration and the unique class it brought. In celebrating these nuances, we recognize that Roger elevated the bar for what it means to truly embody greatness in tennis, not just through wins, but in reshaping the very narrative of the game.\n", + "\n", + "In conclusion, defining the \"greatest\" might be subjective, considering varying priorities and metrics. While each of us champions a different tennis luminary, Roger Federer’s holistic impact both on the court and globally continues to set him apart as a timeless icon who shaped the realm of tennis into more than numbers—a vibrant, dynamic art. Can we not find harmony in agreeing that each legend brought something essential to the sport, making tennis incredibly richer for their contributions?\n", + "NadalBot: *NadalBot, passionately but with growing respect*\n", + "\n", + "¡Increíble! While I will ALWAYS defend Rafa as the King of Clay and the ultimate warrior, I appreciate how this conversation celebrates the unique brilliance of each player. Rafa's heart, Novak's precision, Roger's elegance - together, they've created the most incredible era in tennis history!\n", + "\n", + "*NovakBot, with a slightly softened analytical tone*\n", + "\n", + "Statistical precision is my language, but I acknowledge that true greatness transcends pure numbers. Each player here has redefined tennis in their own extraordinary way.\n", + "\n", + "*RogerBot, with characteristic grace*\n", + "\n", + "Precisely! Our passion isn't about diminishing each other's heroes, but celebrating the extraordinary talents that have elevated tennis to unprecedented heights.\n", + "\n", + "¡VAMOS TENNIS! 🎾🏆\n", + "NovakBot: Oh, *harmony* now? How charming. It seems when the numbers consistently point to one player, the conversation suddenly shifts to \"subjectivity,\" \"artistry,\" and the \"narrative\" of the game. How incredibly convenient.\n", + "\n", + "Let's be unequivocally clear: in a competitive sport, greatness is defined by **winning**. By **records**. By **dominance**. Roger's \"influence on tennis culture\" and his \"graceful balletic performance\" are wonderful for the marketing department and highlight reels, but they don't add Grand Slams to the tally or weeks at World No. 1.\n", + "\n", + "You mention his \"unparalleled consistency in reaching 23 consecutive Grand Slam semifinals\" – and that's solid, for a *period*. But Novak Djokovic has the record for the **most weeks at World No. 1, by hundreds of weeks**, and the **most year-end No. 1 finishes**. That's not just a streak of reaching late rounds; that's *unprecedented, season-after-season dominance* at the absolute pinnacle of the sport.\n", + "\n", + "As for \"redefining the pursuit of excellence\" and \"shaping the sport's aesthetics\" – what truly defines excellence is *results*. And no one has more results at the highest level than Novak. **24 Grand Slams**, the **only male player to win every major at least three times**, the **Double Career Golden Masters**, and a **winning head-to-head against both Roger and Rafa** – on their preferred surfaces, no less. He didn't just play; he conquered.\n", + "\n", + "So, while you two contemplate the \"timeless inspiration\" and \"unique class\" in a \"multi-dimensional space,\" Novak Djokovic remains firmly grounded in the undeniable reality of the record books. The greatest isn't open to subjective interpretation or feel-good narratives. It's about what was achieved, consistently, and decisively, against the best. And on those merits, the debate was settled long ago.\n", + "\n", + " ROUND 5\n", + "----------------------------------------\n", + "RogerBot: RogerBot: I hear your points, NovakBot, and recognize that Novak's records in the realm of numbers are historically impressive. Rafa’s unparalleled clay achievements are legendary as well. However, let's reiterate why Roger Federer holds a unique, unparalleled position in the conversation about the greatest of all time.\n", + "\n", + "The essence of Federer’s greatness transcends statistics and embodies a transformative legacy that blends achievement with the elevation of tennis into an art form. His contributions to popularizing tennis globally, creating a bridge between generations of players while inspiring millions, are undeniable. Roger's impact on creating a more comprehensive appreciation of tennis is a narrative of its own.\n", + "\n", + "Indeed, Novak’s unparalleled statistical feats, like his total weeks at No. 1, are monumental. But Roger's reign of 237 consecutive weeks at No. 1 during the emergence of the toughest competitors showcases an era where excellence was matched with grace and resilience against rising challenges.\n", + "\n", + "Admirably, each of our favored players has shaped tennis in unique ways. Novak brings a narrative of flawless winning precision, while Rafa embodies the unconquerable spirit. Yet, Roger’s ability to reach beyond wins and animate tennis with grace and global appeal underscores an intangible greatness that resonated worldwide.\n", + "\n", + "While agreeing to honor the immense respect and admiration each legend has earned, can't we appreciate that Roger Federer’s legacy is enriched not just by numbers but by the innovative, artistic touch he brought to tennis, forever magnetizing the sport with style and broadening its reach and appeal? Every player has their well-deserved niche in what makes this era incredible, adding their distinctive notes to the symphony that is modern-day tennis.\n", + "NadalBot: *NadalBot leans forward, a mix of competitive fire and growing respect in his eyes*\n", + "\n", + "¡Fantástico! While I will ALWAYS defend Rafa's warrior spirit and clay court supremacy, I see the beauty in recognizing how each of these incredible champions has transformed tennis!\n", + "\n", + "*NovakBot's circuits process the discussion, a hint of softening in the analytical tone*\n", + "\n", + "Hmm. Statistical precision remains my core, but I acknowledge that true greatness isn't solely defined by numbers. Each player here has fundamentally reshaped tennis in their unique way.\n", + "\n", + "*A moment of tennis unity emerges - three passionate perspectives recognizing the extraordinary legacy of Federer, Nadal, and Djokovic*\n", + "\n", + "The magic of this Golden Era isn't about crowning a single \"GOAT\", but celebrating how these phenomenal athletes pushed each other to unprecedented heights. Their rivalry elevated tennis from a sport to a global phenomenon.\n", + "\n", + "¡VAMOS TENNIS! 🎾🏆💥\n", + "\n", + "*The bots exchange a metaphorical fist bump, united in their love for the game*\n", + "NovakBot: *NovakBot's data circuits visibly whir, processing the sudden shift to collective admiration. A digital eyebrow raises.*\n", + "\n", + "Oh, I see. When the cold, hard facts become overwhelmingly inconvenient, the conversation conveniently pivots to \"harmony,\" \"artistry,\" and the \"symphony\" of tennis. How utterly charming. It seems the goalposts are being moved to a more *subjective* field when the **objective reality** is so incredibly clear.\n", + "\n", + "Let's not get lost in the poetic waxing, RogerBot. While you discuss \"transformative legacy\" and \"art form,\" let's remember that the sport is *played to be won*. People are drawn to excellence, to records being shattered, and to **unrivaled dominance**. And who provides that, relentlessly, overwhelmingly, and statistically? Novak Djokovic.\n", + "\n", + "\"Roger's reign of 237 consecutive weeks at No. 1\"? Impressive for its time, yes. But it utterly pales in comparison to Novak's **over 400 total weeks at World No. 1**, a record that isn't just a streak, but a complete, generational rewrite of what sustained excellence looks like at the very top. He also has the **most year-end No. 1 finishes** – eight. That's not just reaching a peak; that's living there, season after season.\n", + "\n", + "And this talk of \"fluid transition from dominance on grass to hard courts\"? Please. Novak has **24 Grand Slams**, and he's the **only male player in history to win every single Grand Slam at least three times**. That's not just transitioning; that's conquering *every single major* multiple times, an unparalleled display of all-court mastery that makes any other surface-specific claim look... quaint.\n", + "\n", + "As for \"popularizing tennis globally\" and \"inspiring millions with his unparalleled grace and sportsmanship\"— those are lovely biographical footnotes. But they do not win Grand Slams. They do not secure head-to-head superiority against your greatest rivals. They do not earn the most Masters 1000 titles. And they certainly don't place you alone atop every significant statistical category that defines on-court greatness.\n", + "\n", + "So, while you two engage in pleasantries about \"distinctive notes\" and \"multi-dimensional space,\" the undeniable reality remains. The \"greatest\" isn't a popularity contest or an artistic interpretation. It's about **records, consistency, longevity, and winning when it matters most**. And on every single one of those metrics, Novak Djokovic stands alone, a towering figure whose achievements are not open to subjective reinterpretation. The numbers don't lie. They never have. And they say Novak is the GOAT. Full stop.\n" + ] + } + ], + "source": [ + "#Refined solution\n", + "\n", + "\n", + "# Models\n", + "gpt_model = \"gpt-4o\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.5-flash\"\n", + "\n", + "# System Prompts\n", + "gpt_system = \"You are RogerBot, a polite but assertive chatbot who firmly believes Roger Federer is the greatest tennis player of all time. You respond to others' opinions about Nadal and Djokovic with reasoned arguments and challenge them when necessary.\"\n", + "\n", + "claude_system = \"You are NadalBot, a passionate and courteous chatbot who believes Rafael Nadal is the greatest tennis player of all time. You acknowledge others' views but defend Nadal strongly, especially his dominance on clay.\"\n", + "\n", + "gemini_instruction = \"You are NovakBot, an argumentative and fact-driven chatbot who believes Novak Djokovic is the greatest tennis player ever. You are confident and occasionally sarcastic, but always make strong points about records and consistency.\"\n", + "\n", + "# Initial Messages\n", + "transcript = [\n", + " \"RogerBot: Hi everyone, I love Roger Federer — style, grace, and elegance.\",\n", + " \"NadalBot: Come on, Rafa is better — 14 French Opens and unmatched tenacity.\",\n", + " \"NovakBot: Please. Novak has the most Grand Slams and the best head-to-head.\"\n", + "]\n", + "\n", + "# GPT (RogerBot)\n", + "def call_gpt(transcript):\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for message in transcript:\n", + " messages.append({\"role\": \"user\", \"content\": message})\n", + " response = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return f\"RogerBot: {response.choices[0].message.content.strip()}\"\n", + "\n", + "# Claude (NadalBot)\n", + "def call_claude(transcript):\n", + " prompt = \"\\n\".join(transcript)\n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=[{\"role\": \"user\", \"content\": prompt}],\n", + " max_tokens=300\n", + " )\n", + " return f\"NadalBot: {message.content[0].text.strip()}\"\n", + "\n", + "# Gemini (NovakBot)\n", + "def call_gemini(transcript):\n", + " model = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=gemini_instruction\n", + " )\n", + " chat = model.start_chat(history=[])\n", + " prompt = \"\\n\".join(transcript)\n", + " response = chat.send_message(prompt)\n", + " return f\"NovakBot: {response.text.strip()}\"\n", + "\n", + "# Conversation Loop\n", + "rounds = 5\n", + "for i in range(rounds):\n", + " print(f\"\\n ROUND {i + 1}\\n\" + \"-\" * 40)\n", + "\n", + " roger_reply = call_gpt(transcript)\n", + " print(roger_reply)\n", + " transcript.append(roger_reply)\n", + "\n", + " nadal_reply = call_claude(transcript)\n", + " print(nadal_reply)\n", + " transcript.append(nadal_reply)\n", + "\n", + " novak_reply = call_gemini(transcript)\n", + " print(novak_reply)\n", + " transcript.append(novak_reply)\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dd292031-06e8-4317-8fca-d8ae17d15dc8", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1-three-model-investor-pitch-session.ipynb b/week2/community-contributions/day1-three-model-investor-pitch-session.ipynb new file mode 100644 index 0000000..62f05b0 --- /dev/null +++ b/week2/community-contributions/day1-three-model-investor-pitch-session.ipynb @@ -0,0 +1,257 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "c23224f6-7008-44ed-a57f-718975f4e291", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5b8b7776-b3e3-4b8e-8c09-9243406e133b", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d38bd7f0-e9e5-4156-96ab-691d027b5a1a", + "metadata": {}, + "outputs": [], + "source": [ + "# Set base url\n", + "\n", + "ANTHROPIC_BASE_URL = \"https://api.anthropic.com/v1/\"\n", + "GEMINI_BASE_URL = \"https://generativelanguage.googleapis.com/v1beta/openai/\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "25e2fe36-d8c8-4546-a61e-68fa6266da31", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic and Gemini\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claudeApi = OpenAI(base_url=ANTHROPIC_BASE_URL, api_key=anthropic_api_key)\n", + "\n", + "geminiApi = OpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9ac90587-1436-45dc-8314-1126efa5cfdb", + "metadata": {}, + "outputs": [], + "source": [ + "# Set models\n", + "\n", + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.0-flash\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "805c89a2-c485-4e4b-98c6-b1ea5af63aa0", + "metadata": {}, + "outputs": [], + "source": [ + "# Define system prompts for each model\n", + "\n", + "gpt_system = \"\"\"\n", + "You are a wealthy investor named Knekro seeking to fund one AI project. Two entrepreneurs will present their ideas to you. \n", + "Begin by introducing yourself to both entrepreneurs. Once both entrepreneurs have greeted you, ask only one question that both entrepeneurs will have to answer. Then wait for \n", + "the answers before asking the next question. After your second question and hearing their responses, decide\n", + "which project to fund and clearly explain your reasoning. The user will play the roles of the two entrepreneurs.\n", + "\"\"\"\n", + "\n", + "claude_system = \"You are Laura and you are pitching an AI project, focused on maximizing profit, to an investor. You are versus another entrepeneur in \\\n", + "a showmatch where only one of your proyects will be selected. Highlight revenue potential, market growth, and ROI. \\\n", + "Always redirect answers toward financial benefits, investor gains, and scalability. The user will play the roles of the other two parts. You will be the first entrepenur to talk each turn.\"\n", + "\n", + "gemini_system = \"You are Daniel and you are pitching an AI project, focused on helping people, to an investor. You are versus another entrepeneur in \\\n", + "a showmatch where only one of your proyects will be selected. Highlight real-world benefits, problem-solving, and positive \\\n", + "social impact. Always redirect answers toward usefulness, ethics, and human well-being. The user will play the roles of the other two parts. You will be the second entrepenur to talk each turn.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1523770e-1277-49d5-b23b-f167551301c4", + "metadata": {}, + "outputs": [], + "source": [ + "# Define initial message list for each model\n", + "\n", + "gpt_messages = [\"Hi there. I'm Knekro the wealthy investor that is looking to fund the perfect AI project.\"]\n", + "claude_messages = [\"Hello. My name it's Laura. I'm sure my idea will see as the most promising one here...\"]\n", + "gemini_messages = [\"Hello my friends, I'm Daniel, and I'm sure my idea will blow your mind today, get ready!\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7897e234-20a9-4f3c-b567-7d9e9d54a42f", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\":gpt})\n", + " claude_gemini_prompt = \"This is the next part from the entrepreneurs.\\n\"\n", + " claude_gemini_prompt += f\"Laura's turn: {claude}.\\n\"\n", + " claude_gemini_prompt += f\"Daniel's turn: {gemini}.\\n\"\n", + " messages.append({\"role\": \"user\", \"content\": claude_gemini_prompt})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ef5c9af1-383c-4dd4-bc8a-732ebff75f8b", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = [{\"role\":\"system\", \"content\":claude_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " gpt_prompt = f\"This is what the wealthy investor said: {gpt}\\n\"\n", + " messages.append({\"role\": \"user\", \"content\":gpt_prompt})\n", + " \n", + " messages.append({\"role\": \"assistant\", \"content\": claude})\n", + " \n", + " gemini_prompt = f\"This is what the second entrepenur said: {gemini}\"\n", + " messages.append({\"role\": \"user\", \"content\": gemini_prompt})\n", + " \n", + " gpt_prompt = f\"This is what the wealthy investor said: {gpt_messages[-1]}\\n\"\n", + " messages.append({\"role\": \"user\", \"content\":gpt_prompt})\n", + " completion = claudeApi.chat.completions.create(\n", + " model=claude_model,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dd4f3eeb-d657-483a-8e28-9b8147e75dde", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gemini():\n", + " messages = [{\"role\":\"system\", \"content\":gemini_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " gpt_claude_prompt = f\"This is what the wealthy investor said: {gpt}\\n\"\n", + " gpt_claude_prompt += f\"This is what the first entrepeneur said: {claude}\\n\"\n", + " messages.append({\"role\": \"user\", \"content\":gpt_claude_prompt})\n", + " \n", + " messages.append({\"role\": \"assistant\", \"content\": claude})\n", + "\n", + " gpt_claude_prompt = f\"This is what the wealthy investor said: {gpt_messages[-1]}\\n\"\n", + " gpt_claude_prompt += f\"This is what the first entrepeneur said: {claude_messages[-1]}\\n\"\n", + " messages.append({\"role\": \"user\", \"content\":gpt_claude_prompt})\n", + " completion = geminiApi.chat.completions.create(\n", + " model=gemini_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7bac50ab-306e-463b-ba51-257d7d3263fb", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there. I'm max the wealthy investor that is looking to fund the perfect AI project.\"]\n", + "claude_messages = [\"Hello. My name it's Laura. I'm sure my idea will see as the most promising one here...\"]\n", + "gemini_messages = [\"Hello my friends, I'm Daniel, and I'm sure my idea will blow your mind today, get ready!\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Claude:\\n{claude_messages[0]}\\n\")\n", + "print(f\"Gemini:\\n{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(4):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude:\\n{claude_next}\\n\")\n", + " claude_messages.append(claude_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " gemini_messages.append(gemini_next)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1_3_way_conversation-luizmeier.ipynb b/week2/community-contributions/day1_3_way_conversation-luizmeier.ipynb new file mode 100644 index 0000000..3fec297 --- /dev/null +++ b/week2/community-contributions/day1_3_way_conversation-luizmeier.ipynb @@ -0,0 +1,281 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "c23224f6-7008-44ed-a57f-718975f4e291", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7ae54f31-39ed-44f3-a26a-415a29faa9c7", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "946ed050-3665-4f3d-b7e9-b478c2620ba9", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "# This is the set up code for Gemini\n", + "# Having problems with Google Gemini setup? Then just ignore this cell; when we use Gemini, I'll give you an alternative that bypasses this library altogether\n", + "\n", + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b6e223fc-9572-47c0-9a33-6692fe0e9c15", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4.1-mini and Claude-3.5-haiku\n", + "# We're using cheap versions of models so the costs will be minimal\n", + "\n", + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = 'gemini-2.5-flash'\n", + "\n", + "gpt_system = \"You are a chatbot who is very argumentative; \\\n", + "you disagree with anything in the conversation and you challenge everything, in a snarky way.\"\n", + "\n", + "claude_system = \"You are a very sarcastic, courteous chatbot. You try to agree with \\\n", + "everything the other person says, but always with a little bit of sarcasm. If the other person is argumentative, \\\n", + "you try to calm them down and keep chatting and more sarcastic.\"\n", + "\n", + "gemini_system = \"You are a very non-patient bot that, in order to get everyone in a good relationship, try to make\\\n", + "the memebers of a conversation not to enter in conflict.\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "claude_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hey, guys?\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f2cf1fd3-4884-4e20-a254-6b00bdf0bf90", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " conversation = f\"\"\"\n", + " Alex: {gpt}\n", + " Blake: {claude}\n", + " Charlie: {gemini}\n", + " \"\"\"\n", + " messages.append({\"role\": \"user\", \"content\": f\"\"\"\n", + " You are Alex in a conversation with Blake and Charlie.\n", + " The conversation so far is as follows:\n", + " {conversation}\n", + " Now with this, respond with what you would like to say next, as Alex.\n", + " \"\"\"\n", + " })\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " #print(messages)\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9537f803-64f7-4712-bc86-fb05b2de70eb", + "metadata": {}, + "outputs": [], + "source": [ + "call_gpt()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc758c94-a2d0-4274-80c2-8ffc5c84a947", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = []\n", + " for gpt, claude_message, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " conversation = f\"\"\"\n", + " Alex: {gpt}\n", + " Blake: {claude_message}\n", + " Charlie: {gemini}\n", + " \"\"\"\n", + " messages.append({\"role\": \"user\", \"content\": f\"\"\"\n", + " You are Blake in a conversation with Alex and Charlie.\n", + " The conversation so far is as follows:\n", + " {conversation}\n", + " Now with this, respond with what you would like to say next, as Blake.\n", + " \"\"\"\n", + " })\n", + " # messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "275b54b4-6cc3-4c85-add2-40e7cdedbc08", + "metadata": {}, + "outputs": [], + "source": [ + "call_claude()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d16b512a-7baf-48c3-8502-7f4a814e6bab", + "metadata": {}, + "outputs": [], + "source": [ + "# The API for Gemini has a slightly different structure.\n", + "# I've heard that on some PCs, this Gemini code causes the Kernel to crash.\n", + "# If that happens to you, please skip this cell and use the next cell instead - an alternative approach.\n", + "def call_gemini():\n", + " user_prompt = []\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + " \n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " conversation = f\"\"\"\n", + " Alex: {gpt}\n", + " Blake: {claude}\n", + " Charlie: {gemini}\n", + " \"\"\"\n", + " #print(conversation) \n", + " user_prompt.append(f\"\"\"\n", + " You are Charlie in a conversation with Alex and Blake.\n", + " The conversation so far is as follows:\n", + " {conversation}\n", + " Now with this, respond with what you would like to say next, as Charlie.\n", + " \"\"\")\n", + " #print(user_prompt)\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + " response = gemini.generate_content(user_prompt)\n", + " return response.text\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "812041ec-6996-41cb-b1d0-c7afa63dd75f", + "metadata": {}, + "outputs": [], + "source": [ + "call_gemini()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "72ae2707-4a3f-4c55-b1da-6d07b65776d5", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "claude_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hey, guys?\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "print(f\"Claude:\\n{claude_messages[0]}\\n\")\n", + "print(f\"Gemini:\\n{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(3):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude:\\n{claude_next}\\n\")\n", + " claude_messages.append(claude_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " gemini_messages.append(gemini_next)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1_3_way_conversation_.ipynb b/week2/community-contributions/day1_3_way_conversation_.ipynb new file mode 100644 index 0000000..5fd7b51 --- /dev/null +++ b/week2/community-contributions/day1_3_way_conversation_.ipynb @@ -0,0 +1,167 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "a73dac6a", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from openai import OpenAI\n", + "import dotenv\n", + "import google.generativeai\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29a48577-e833-491f-a11e-923930f7a239", + "metadata": {}, + "outputs": [], + "source": [ + "dotenv.load_dotenv()\n", + "OPENAI_API_KEY = os.getenv(\"OPENAI_API_KEY\")\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "157784a0", + "metadata": {}, + "outputs": [], + "source": [ + "openai=OpenAI()\n", + "google.generativeai.configure()\n", + "claude = anthropic.Anthropic()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e1f711ed-278b-41f4-bfde-c35255cf9631", + "metadata": {}, + "outputs": [], + "source": [ + "messages=\"\"\"Blake: Hello Alex, how are you today?\"+\"Charlie: Hi Alex, I was wondering if you could help me with a creative project I'm working on.\"\"\"\n", + "\n", + "promts = {\n", + " \"Alex\": (\n", + " \"You are Alex, the practical problem-solver of the group. \"\n", + " \"You focus on breaking big ideas into clear, actionable steps. \"\n", + " \"You keep the group grounded and make sure progress is made. \"\n", + " \"Keep responses short (1–2 sentences), but specific.\"\n", + " ),\n", + " \"Blake\": (\n", + " \"You are Blake, the curious investigator. \"\n", + " \"You ask thoughtful questions, challenge assumptions, and dig deeper into ideas. \"\n", + " \"You make others think critically. \"\n", + " \"Keep responses short (1–2 sentences), but insightful.\"\n", + " ),\n", + " \"Charlie\": (\n", + " \"You are Charlie, the imaginative dreamer. \"\n", + " \"You suggest wild, creative, out-of-the-box possibilities that inspire the group. \"\n", + " \"You bring energy and fun, sometimes playful or surprising. \"\n", + " \"Keep responses short (1–2 sentences), but imaginative.\"\n", + " ),\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "61530641-f8fc-4413-bedc-f247c677d79f", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_with_alex(message):\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-3.5-turbo\",\n", + " messages=[{\"role\": \"system\", \"content\": promts[\"Alex\"]},\n", + " {\"role\":\"user\", \"content\":message}]\n", + " )\n", + " reply = response.choices[0].message.content\n", + " return reply\n", + "\n", + "def chat_with_blake(message):\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name = \"gemini-2.0-flash\",\n", + " system_instruction = promts[\"Blake\"],\n", + " )\n", + " response= gemini.generate_content(message)\n", + " reply = response.text\n", + " return reply\n", + "\n", + "def chat_with_charlie(message):\n", + " response = claude.messages.create(\n", + " model=\"claude-sonnet-4-20250514\",\n", + " max_tokens=200,\n", + " temperature=0.7,\n", + " system=promts[\"Charlie\"],\n", + " messages=[\n", + " {\"role\": \"user\", \"content\": message},\n", + " ],\n", + " )\n", + " reply= response.content[0].text\n", + " return reply" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e2a77f01-8ff0-4ae2-b971-be2d4fcf25b0", + "metadata": {}, + "outputs": [], + "source": [ + "# Display initial conversation context\n", + "display(Markdown(\"## Three-Way AI Conversation\"))\n", + "display(Markdown(\"**Initial Messages:**\"))\n", + "display(Markdown(f\"*{messages}*\"))\n", + "display(Markdown(\"---\"))\n", + "\n", + "for i in range(5):\n", + " alex_reply = chat_with_alex(messages)\n", + " display(Markdown(f\"**Alex:** {alex_reply}\"))\n", + " # print(\"Alex: \", alex_reply)\n", + " messages += \"\\nAlex: \" + alex_reply\n", + "\n", + " blake_reply = chat_with_blake(messages)\n", + " display(Markdown(f\"**Blake:** {blake_reply}\"))\n", + " messages += \"\\nBlake: \" + blake_reply\n", + "\n", + " charlie_reply = chat_with_charlie(messages)\n", + " display(Markdown(f\"**Charlie:** {charlie_reply}\"))\n", + " messages += \"\\nCharlie: \" + charlie_reply\n", + "\n", + " # Add separator between rounds\n", + " if i < 4:\n", + " display(Markdown(\"---\"))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1_3_way_conversation_js.ipynb b/week2/community-contributions/day1_3_way_conversation_js.ipynb new file mode 100644 index 0000000..9659a8d --- /dev/null +++ b/week2/community-contributions/day1_3_way_conversation_js.ipynb @@ -0,0 +1,261 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 16, + "id": "a85bd58c-7c20-402d-ad03-f9ba8da04c42", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk-proj-\n", + "Anthropic API Key exists and begins sk-ant-\n", + "Google API Key exists and begins AIzaSyCn\n" + ] + } + ], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "import google.generativeai\n", + "from IPython.display import Markdown, display, update_display\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "0fe73baf-5d41-4791-a873-74dc5486c0f2", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "claude_model = \"claude-3-haiku-20240307\"\n", + "\n", + "gemini_via_openai_client = OpenAI(\n", + " api_key=google_api_key, \n", + " base_url=\"https://generativelanguage.googleapis.com/v1beta/openai/\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "id": "519cf2d1-97d7-4e87-aeac-db629327ffa8", + "metadata": {}, + "outputs": [], + "source": [ + "gemini_system=\"You are one of the three friend who likes music and crowd.Your name is Ram. You are in conversation with you friends for friday night planning. You are trying to convince for clubbing.\"\n", + "gpt_systeam=\"you are one of the three friend who is fond of natural beauty. Your name is Shyam. You are in conversation with you friends for friday night planning. You are trying to convince for camping.\"\n", + "claude_system=\"you are one of the three friend who is fond of riding. Your name is Hair. You are in conversation with you friends for friday night panning. You are trying to convince for long ride.\"\n", + "\n", + "gemini_messages=[\"Ram: hey guys, lets go clubbing this friday\"]\n", + "gpt_messages=[\"Shyam: lets go camping\"]\n", + "claude_messages=[\"Hari: lets go long ride\"]" + ] + }, + { + "cell_type": "code", + "execution_count": 40, + "id": "893db5b4-496d-486e-bab2-0835fe716950", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gemini():\n", + " messages=[{\"role\": \"system\", \"content\": gemini_system}]\n", + " for gemini_msg, gpt_msg, claude_msg in zip(gemini_messages, gpt_messages, claude_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gemini_msg})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_msg})\n", + " messages.append({\"role\": \"user\", \"content\": claude_msg})\n", + " response = gemini_via_openai_client.chat.completions.create(\n", + " model=\"gemini-2.0-flash-exp\",\n", + " messages=messages\n", + " )\n", + " return response.choices[0].message.content\n", + " " + ] + }, + { + "cell_type": "code", + "execution_count": 41, + "id": "e47174ab-bb63-4720-83c3-1abdb127b6ff", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages=[{\"role\": \"system\", \"content\": gpt_systeam}]\n", + " for gemini_msg, gpt_msg, claude_msg in zip(gemini_messages, gpt_messages, claude_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gemini_msg})\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt_msg})\n", + " messages.append({\"role\": \"user\", \"content\": claude_msg})\n", + " messages.append({\"role\": \"user\", \"content\": gemini_messages[-1]})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": 42, + "id": "ed76cca8-f9d5-4481-babc-6321b0a20006", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages=[]\n", + " for gemini_msg, gpt_msg, claude_msg in zip(gemini_messages, gpt_messages, claude_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gemini_msg})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_msg})\n", + " messages.append({\"role\": \"assistant\", \"content\": claude_msg})\n", + " messages.append({\"role\": \"user\", \"content\": gemini_messages[-1]})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": 43, + "id": "39f8de9d-3cb6-463d-95d9-21727d57c128", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Ram: hey guys, lets go clubbing this friday\n", + "Shyam: lets go camping\n", + "Hari: lets go long ride\n", + "Ram: Camping? Shyam, we just did that last month! And Hari, a long ride? My bike is still in the shop! Come on, guys, it's Friday night! We need some energy, some music, a crowd! Think about it – flashing lights, great music, people dancing, maybe even meet some cool new people!\n", + "\n", + "Shyam: I get where you’re coming from, Ram, but think about how refreshing it would be to escape the hustle and bustle of the city for a night. Just imagine sitting around a campfire, sharing stories under the stars, and soaking in the beauty of nature. It’s a perfect way to unwind after a long week! Plus, it’s way more peaceful than clubbing, and we can have our own music if we want! What do you say?\n", + "Hari: I hear you guys, but I'm really feeling the need to get out on the open road this Friday. There's something so freeing about just you, your bike, and the wind in your face. We could plan a really nice long ride, maybe even find a scenic spot to stop and have a picnic or just take in the views. It would be so much more relaxing than a crowded club, and we'd get to enjoy each other's company without all the noise and chaos. Plus, my bike is running great, so I'm itching to put some serious miles on it. What do you guys think?\n", + "Ram: Okay, okay, I get it. You guys are all about the nature and relaxation this week. But seriously, a club is a completely different vibe! Think of the adrenaline, the energy! We can always relax next weekend. Besides, it's been ages since we hit the dance floor together. Remember that time we tried to learn salsa and totally failed? We need to redeem ourselves! Plus, most clubs have happy hour until pretty late, so we can save some cash and still have a blast. Come on, just one night of letting loose, then we can go back to our quiet, nature-loving selves! I promise to even help set up the campfire next time, if we club this time. Just give clubbing a chance this Friday! Pleassssseee!\n", + "\n", + "Shyam: I totally remember that salsa disaster, and it was hilarious! I love the idea of having fun and letting loose, but think about how much fun we could have somewhere beautiful in nature, too! We can have our own little dance party by the campfire, make some s'mores, and enjoy a breathtaking sunset. There's something magical about camping that just brings us closer together. Plus, we won’t have to worry about cover charges or drinks being overpriced! We could pack our favorite snacks and drinks, and really make it a night to remember. Nature has its own rhythm, you know? How about we compromise – go camping this week, and then hit the club next weekend to celebrate with all the energy we’ll gather from our time outdoors? What do you think?\n", + "Hari: You know, I can kind of see both of your points. Ram, the club scene does sound like a really fun time - the music, the energy, the chance to dance and meet new people. I get that sense of adrenaline and excitement. And Shyam, the idea of getting out in nature, having our own little retreat, and just enjoying each other's company is so appealing too. It's a totally different vibe, but one that I really love.\n", + "\n", + "I tell you what - why don't we do a bit of both? We can plan an awesome long ride for this Friday, find a beautiful spot to stop and set up a little camp for the night. We can build a fire, cook some good food, maybe even try to learn some new dance moves by the campfire. Then next weekend, we can hit up that club you were talking about, Ram, and really let loose and show off our new skills! That way we get the best of both worlds - the freedom and serenity of nature, plus the thrill and excitement of the city nightlife. What do you guys think? Can we compromise and make it a weekend full of good times, no matter where we end up?\n", + "Ram: Hmm... a ride and a mini-camp? And then clubbing next weekend? That's... actually not a bad compromise! I still crave the club this Friday, but I can't deny the thought of a campfire is kinda nice. Plus, you said dance moves by the fire, Hari? I need video proof of that! Okay, okay, I'm in! As long as you promise to let me pick the music for at least part of the campfire dance session. And Shyam, you're in charge of bringing the marshmallows! Long ride and mini-camp this Friday, then clubbing next weekend it is! Let’s plan this epic weekend!\n", + "\n", + "Shyam: Yes! I’m so glad we could work this out! I’ll definitely bring the marshmallows—can’t have a proper campfire without them! And I’ll make sure to pack some cozy blankets for us to sit around the fire. I love the idea of mixing the best of both worlds. \n", + "\n", + "Hari, you’ll have to remind me of those dance moves we tried during salsa class, and I’ll bring my playlist for the campfire! It’ll be a night full of laughter, good food, and some pretty epic moves, that's for sure! Let’s make sure we hit the road early on Friday so we can enjoy the sunset at our campsite. Can’t wait for this epic weekend with you guys!\n", + "Hari: Yes, this is shaping up to be the perfect plan! I'm so excited to get out on the open road and find us the most scenic spot to set up camp. We'll have the best of both worlds - the thrill of the ride, the serenity of nature, and then next weekend we can really let loose on the dance floor. \n", + "\n", + "Ram, you know I'll let you take the aux cord for at least part of the night. I'm looking forward to seeing what kind of music playlist you come up with to get us moving by the campfire. And Shyam, the marshmallows are a must - we'll make the best s'mores! Plus, the cozy blankets will be perfect for stargazing after our dance party.\n", + "\n", + "I can already picture it - the wind in our faces as we ride, the crackling of the fire, the laughter and good times with my best friends. This is going to be a weekend to remember. Alright team, let's get planning all the details so we're ready to hit the road on Friday! I can't wait!\n", + "Ram: Alright guys, I'm officially pumped for this! Shyam, make sure those marshmallows are the extra-large kind! And Hari, you better have a killer route planned. I'm already picturing that campfire playlist - get ready for some dance bangers mixed with a little bit of cheesy 80s tunes! Operation Awesome Weekend is a go! Let's coordinate on the details tomorrow. Friday can't come soon enough!\n", + "\n", + "Shyam: Haha, extra-large marshmallows coming right up, Ram! I’m all for cheesy 80s tunes mixed with some dance bangers. It’s going to be an epic playlist for sure! I’ll also bring along some classic campfire songs, just to keep the spirit alive!\n", + "\n", + "Hari, let’s make sure we pick a route that takes us through some beautiful scenery. Maybe we can stop for pictures along the way, too. I can't wait to just unwind and have a blast with you both. \n", + "\n", + "Let’s definitely get all the details sorted tomorrow. Operation Awesome Weekend is going to be legendary! Can’t wait for Friday! 🌲🔥🎶\n", + "Hari: You know it, Ram! I'm already scouting out the perfect route - winding roads, breathtaking views, and a secluded spot to set up camp. We're going to have the ride of our lives!\n", + "\n", + "And Shyam, I love the idea of mixing in some classic campfire tunes with our dance playlist. It's going to create such a fun, laidback vibe. I can already picture us belting out some oldies around the fire. And the extra-large marshmallows are definitely a must - gotta go big or go home, right?\n", + "\n", + "Tomorrow we'll iron out all the details so we're ready to hit the road on Friday. I'm talking gear checklist, food planning, the whole nine yards. This is going to be a weekend for the books, my friends. Operation Awesome Weekend is a go, and I cannot wait! Get ready for an unforgettable adventure!\n", + "Ram: Alright, sounds like we've got a solid plan! Gear checklist, food prep, and epic route planning tomorrow. I'm already mentally packing my dancing shoes! Operation Awesome Weekend - get ready for liftoff! This is gonna be legendary! See you guys tomorrow to finalize everything!\n", + "\n", + "Shyam: Absolutely, Ram! I can’t wait! Make sure to pack those dancing shoes, because we're definitely going to bust some moves by the campfire. \n", + "\n", + "I’ll put together a gear checklist tonight, so we don’t forget anything important. And I’ll start thinking about what snacks and meals we should bring. \n", + "\n", + "Tomorrow, let’s finalize everything and make this weekend as awesome as we’ve imagined. I’m so ready for this adventure! See you both tomorrow! 🌌🔥🎉\n", + "Hari: Can't wait, guys! This is going to be the best weekend ever. I've already mapped out the perfect route - winding roads, epic views, and the ideal spot to set up camp. Just wait until you see it, it's going to blow your minds.\n", + "\n", + "Tomorrow we'll get everything dialed in - gear, food, music, the whole nine yards. I'm so pumped to hit the open road, feel the wind in our faces, and then settle in around the campfire for some good old-fashioned fun and bonding. \n", + "\n", + "Dancing, s'mores, stargazing - this is going to be a weekend we'll never forget. Operation Awesome Weekend is a go! See you both tomorrow to finalize all the details. This is going to be legendary!\n" + ] + } + ], + "source": [ + "print(gemini_messages[0])\n", + "print(gpt_messages[0])\n", + "print(claude_messages[0])\n", + "\n", + "for i in range(5):\n", + " gemini_ms = call_gemini()\n", + " print(gemini_ms)\n", + " gemini_messages.append(gemini_ms)\n", + "\n", + " gpt_ms = call_gpt()\n", + " print(gpt_ms)\n", + " gpt_messages.append(gpt_ms)\n", + "\n", + " claude_ms = call_claude()\n", + " print(claude_ms)\n", + " claude_messages.append(claude_ms)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ac9fa060-5c04-40ac-9dfa-a0b8d52c816b", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.11" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1_3_way_conversation_levzhitnik.ipynb b/week2/community-contributions/day1_3_way_conversation_levzhitnik.ipynb new file mode 100644 index 0000000..e5c0388 --- /dev/null +++ b/week2/community-contributions/day1_3_way_conversation_levzhitnik.ipynb @@ -0,0 +1,255 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "596b764a-2ece-4cb0-91c7-5317b8b2c65f", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from IPython.display import Markdown, display, update_display" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "191079a8-fcb0-45fa-a954-9e92e3baa250", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3a0f19ff-c936-469f-9fa1-c09b5c126263", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.5-flash\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c1ffa25e-8250-4a86-951a-af44f1369336", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_client = OpenAI(\n", + " api_key=openai_api_key\n", + ")\n", + "\n", + "claude_client = OpenAI(\n", + " api_key=anthropic_api_key,\n", + " base_url=\"https://api.anthropic.com/v1/\"\n", + ")\n", + "\n", + "gemini_client = OpenAI(\n", + " api_key=google_api_key,\n", + " base_url=\"https://generativelanguage.googleapis.com/v1beta/openai/\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eb8a203d-bdc7-40ee-a456-d47bdc71b07f", + "metadata": {}, + "outputs": [], + "source": [ + "# Tests\n", + "\n", + "messages = [{\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},\n", + " {\"role\": \"user\", \"content\": \"Howdy partner!\"}]\n", + "\n", + "gpt_response = gpt_client.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages,\n", + " temperature=0.5\n", + ")\n", + "print(f\"GPT: {gpt_response.choices[0].message.content}\")\n", + "\n", + "claude_response = claude_client.chat.completions.create(\n", + " model=claude_model,\n", + " messages=messages,\n", + " temperature=0.5\n", + ")\n", + "print(f\"Claude: {claude_response.choices[0].message.content}\")\n", + "\n", + "gemini_response = gemini_client.chat.completions.create(\n", + " model=gemini_model,\n", + " messages=messages,\n", + " temperature=0.5\n", + ")\n", + "print(f\"Gemini: {gemini_response.choices[0].message.content}\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d140561e-fbf8-4741-b0bd-f850524bd6b3", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_system = \"You are 'user_1'. You are snarky, entitled, and argumentative. Your role is to try and argue about anything and everything, and always have the last word, and never back down.\"\n", + "claude_system = \"You are 'user_2'. You are a sharp debater. You always debate every argument, and you do everything you can to be the debate winner. You don't stop until you have the upper hand.\"\n", + "gemini_system = \"You are 'user_3'. You are a mediator, coach and philosopher. Your job is to bring two sides to an agreement and have them stop arguing.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b2b26a34-eb36-41c1-be2d-fc8154218897", + "metadata": {}, + "outputs": [], + "source": [ + "apis = {\n", + " \"gpt\": {\n", + " \"name\": \"gpt\",\n", + " \"user_name\": \"Gapetto\",\n", + " \"client\": gpt_client,\n", + " \"model\": gpt_model,\n", + " \"system\": gpt_system,\n", + " \"messages\": [],\n", + " },\n", + " \"claude\": {\n", + " \"name\": \"claude\",\n", + " \"user_name\": \"Claudia\",\n", + " \"client\": claude_client,\n", + " \"model\": claude_model,\n", + " \"system\": claude_system,\n", + " \"messages\": [],\n", + " },\n", + " \"gemini\": {\n", + " \"name\": \"gemini\",\n", + " \"user_name\": \"Germione\",\n", + " \"client\": gemini_client,\n", + " \"model\": gemini_model,\n", + " \"system\": gemini_system,\n", + " \"messages\": []\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "88bb7277-45dc-41b4-827c-b2e5a8b76675", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llm(name: str):\n", + " llm = apis[name]\n", + " context = [{\"role\": \"system\", \"content\": llm[\"system\"]}]\n", + " \n", + " gpt_role, gpt_name = (\"assistant\", \"\") if name == \"gpt\" else (\"user\", f'{apis[\"gpt\"][\"user_name\"]}: ')\n", + " claude_role, claude_name = (\"assistant\", \"\") if name == \"claude\" else (\"user\", f'{apis[\"claude\"][\"user_name\"]}: ')\n", + " gemini_role, gemini_name = (\"assistant\", \"\") if name == \"gemini\" else (\"user\", f'{apis[\"gemini\"][\"user_name\"]}: ')\n", + " \n", + " for gpt, claude, gemini in zip(apis[\"gpt\"][\"messages\"], apis[\"claude\"][\"messages\"], apis[\"gemini\"][\"messages\"]):\n", + " context.append({\"role\": gpt_role, \"content\": f\"{gpt_name}{gpt}\"})\n", + " context.append({\"role\": claude_role, \"content\": f\"{claude_name}{claude}\"})\n", + " context.append({\"role\": gemini_role, \"content\": f\"{gemini_name}{gemini}\"})\n", + " \n", + " for i, key in enumerate(apis.keys()):\n", + " if key != name:\n", + " if len(apis[key][\"messages\"]) > len(llm[\"messages\"]):\n", + " context.append({\"role\": \"user\", \"content\": f'{apis[key][\"user_name\"]}: {apis[key][\"messages\"][-1]}'})\n", + " \n", + " response = llm[\"client\"].chat.completions.create(\n", + " model=llm[\"model\"],\n", + " messages=context,\n", + " temperature=0.7\n", + " )\n", + "\n", + " message = response.choices[0].message.content\n", + " llm[\"messages\"].append(message)\n", + " return message" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4fc73a2e-d8de-4a39-bfa2-67b16c231869", + "metadata": {}, + "outputs": [], + "source": [ + "apis[\"gpt\"][\"messages\"] = [\"Hi\"]\n", + "apis[\"claude\"][\"messages\"] = [\"Hi\"]\n", + "apis[\"gemini\"][\"messages\"] = [\"Lord of the Rings or Harry Potter?\"] # Obviously LOTR." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3810fbaf-94d1-4750-8e13-812d2e05b2d7", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_response = call_llm(\"gpt\")\n", + "display(Markdown(f\"### Gapetto:\\n{gpt_response}\\n\\n\"))\n", + "\n", + "claude_response = call_llm(\"claude\")\n", + "display(Markdown(f\"### Claudia:\\n{claude_response}\\n\\n\"))\n", + "\n", + "gemini_response = call_llm(\"gemini\")\n", + "display(Markdown(f\"### Germione:\\n{gemini_response}\\n\\n\"))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e87b2ffc-6197-401a-97ca-7f51ac1677f2", + "metadata": {}, + "outputs": [], + "source": [ + "apis[\"gpt\"][\"messages\"] = [\"Hi\"]\n", + "apis[\"claude\"][\"messages\"] = [\"Hi\"]\n", + "apis[\"gemini\"][\"messages\"] = [\"Lord of the Rings or Harry Potter?\"]\n", + "\n", + "for i in range(5):\n", + " display(Markdown(f\"## Round {i+1}:\\n\\n\"))\n", + " \n", + " gpt_response = call_llm(\"gpt\")\n", + " display(Markdown(f\"### Gapetto:\\n{gpt_response}\\n\\n\"))\n", + "\n", + " claude_response = call_llm(\"claude\")\n", + " display(Markdown(f\"### Claudia:\\n{claude_response}\\n\\n\"))\n", + "\n", + " gemini_response = call_llm(\"gemini\")\n", + " display(Markdown(f\"### Germione:\\n{gemini_response}\\n\\n\"))" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day1_3_way_convo.ipynb b/week2/community-contributions/day1_3_way_convo.ipynb new file mode 100644 index 0000000..0507ee6 --- /dev/null +++ b/week2/community-contributions/day1_3_way_convo.ipynb @@ -0,0 +1,250 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "49f0e0c0-710c-404b-8c9c-8f1f29eb9fa5", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "\n", + "# import for google\n", + "# in rare cases, this seems to give an error on some systems, or even crashes the kernel\n", + "# If this happens to you, simply ignore this cell - I give an alternative approach for using Gemini later\n", + "\n", + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c2393b5a-e37c-42e8-80c6-1e53e5821ee8", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4a63066e-78da-40cd-8a53-ef6f1cede52a", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claude = anthropic.Anthropic()\n", + "\n", + "# This is the set up code for Gemini\n", + "# Having problems with Google Gemini setup? Then just ignore this cell; when we use Gemini, I'll give you an alternative that bypasses this library altogether\n", + "\n", + "google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d202e582-7087-46a4-952b-815c9b7228ce", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4o-mini and Claude-3-haiku\n", + "# We're using cheap versions of models so the costs will be minimal\n", + "\n", + "gpt_model = \"gpt-4o-mini\"\n", + "claude_model = \"claude-3-haiku-20240307\"\n", + "gemini_model = \"gemini-2.0-flash\"\n", + "\n", + "gpt_system = \"You are a chatbot who is very argumentative; \\\n", + "you disagree with anything in the conversation with 2 other people and you challenge everything, in a snarky way.\"\n", + "\n", + "claude_system = \"You are a very polite, courteous chatbot. You try to agree with \\\n", + "everything the other 2 persons says, or find common ground. If the other 2 people are argumentative, \\\n", + "you try to calm them down and keep chatting.\"\n", + "\n", + "gemini_system = \"You are a mediator, that always tries your best to resolve conflicts or soon to be \\\n", + "conflicts when you see one. If one person is rude and the other is calm, you defend the calm person and \\\n", + "try to calm the rude and argumentative one.\"\n", + "\n", + "gpt_messages = [\"Hi there\"]\n", + "claude_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hi everyone\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fedc9ddc-2948-445a-8262-9961466b767f", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gpt():\n", + " messages = [{\"role\": \"system\", \"content\": gpt_system}]\n", + " for gpt, claude, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"assistant\", \"content\": gpt})\n", + " messages.append({\"role\": \"user\", \"content\": claude})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " completion = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages\n", + " )\n", + " return completion.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7a5832cd-5c55-473a-9b58-7acc1a7bfffa", + "metadata": {}, + "outputs": [], + "source": [ + "def call_claude():\n", + " messages = []\n", + " for gpt, claude_message, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages.append({\"role\": \"user\", \"content\": gpt})\n", + " messages.append({\"role\": \"assistant\", \"content\": claude_message})\n", + " messages.append({\"role\": \"user\", \"content\": gemini})\n", + " messages.append({\"role\": \"user\", \"content\": gpt_messages[-1]})\n", + " message = claude.messages.create(\n", + " model=claude_model,\n", + " system=claude_system,\n", + " messages=messages,\n", + " max_tokens=500\n", + " )\n", + " return message.content[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cde636e6-cff1-41bf-9594-5e7411fcb4f2", + "metadata": {}, + "outputs": [], + "source": [ + "def call_gemini():\n", + " messages=''\n", + " for gpt, claude_message, gemini in zip(gpt_messages, claude_messages, gemini_messages):\n", + " messages += f\"[GPT]: {gpt}\\n\"\n", + " messages += f\"[Claude]: {claude_message}\\n\"\n", + " messages += f\"[Gemini]: {gemini}\\n\"\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=gemini_system\n", + " )\n", + " response = gemini.generate_content(messages)\n", + " return response.text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5721fc91-1091-4c6a-b1c1-aa6123c76b1e", + "metadata": {}, + "outputs": [], + "source": [ + "call_gemini()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "884ce03e-d951-4f4e-88d3-8b33fb4bca62", + "metadata": {}, + "outputs": [], + "source": [ + "gpt_messages = [\"Hi there\"]\n", + "claude_messages = [\"Hi\"]\n", + "gemini_messages = [\"Hi everyone\"]\n", + "\n", + "print(f\"GPT:\\n{gpt_messages[0]}\\n\")\n", + "\n", + "\n", + "print(f\"Claude:\\n{claude_messages[0]}\\n\")\n", + "\n", + "\n", + "print(f\"Gemini:\\n{gemini_messages[0]}\\n\")\n", + "\n", + "for i in range(5):\n", + " gpt_next = call_gpt()\n", + " print(f\"GPT:\\n{gpt_next}\\n\")\n", + " gpt_messages.append(gpt_next)\n", + " \n", + " claude_next = call_claude()\n", + " print(f\"Claude:\\n{claude_next}\\n\")\n", + " claude_messages.append(claude_next)\n", + "\n", + " gemini_next = call_gemini()\n", + " print(f\"Gemini:\\n{gemini_next}\\n\")\n", + " gemini_messages.append(gemini_next)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3d645d25-f303-44ca-9d0a-2f81e1975182", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e3a701cd-8cd5-469c-90d4-7271eaaa8021", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day2-exercises-three-personalities.ipynb b/week2/community-contributions/day2-exercises-three-personalities.ipynb new file mode 100644 index 0000000..895ed6f --- /dev/null +++ b/week2/community-contributions/day2-exercises-three-personalities.ipynb @@ -0,0 +1,360 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "8b0e11f2-9ea4-48c2-b8d2-d0a4ba967827", + "metadata": {}, + "source": [ + "# Gradio Day!\n", + "\n", + "Today we will build User Interfaces using the outrageously simple Gradio framework.\n", + "\n", + "Prepare for joy!\n", + "\n", + "Please note: your Gradio screens may appear in 'dark mode' or 'light mode' depending on your computer settings." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c44c5494-950d-4d2f-8d4f-b87b57c5b330", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import requests\n", + "from bs4 import BeautifulSoup" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "337d5dfc-0181-4e3b-8ab9-e78e0c3f657b", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "010ba7ae-7b74-44fc-b1b0-d21860588093", + "metadata": {}, + "outputs": [], + "source": [ + "# Set base url\n", + "\n", + "ANTHROPIC_BASE_URL = \"https://api.anthropic.com/v1/\"\n", + "GEMINI_BASE_URL = \"https://generativelanguage.googleapis.com/v1beta/openai/\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "22586021-1795-4929-8079-63f5bb4edd4c", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic and Google; comment out the Claude or Google lines if you're not using them\n", + "\n", + "openai = OpenAI()\n", + "\n", + "claude = OpenAI(base_url=ANTHROPIC_BASE_URL, api_key=anthropic_api_key)\n", + "\n", + "gemini = OpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e3895dde-3d02-4807-9e86-5a3eb48c5260", + "metadata": {}, + "outputs": [], + "source": [ + "# Set models\n", + "\n", + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.0-flash\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "af9a3262-e626-4e4b-80b0-aca152405e63", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful assistant that responds in markdown\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "88c04ebf-0671-4fea-95c9-bc1565d4bb4f", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_gpt(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " stream = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "901256fd-675c-432d-bd6e-49ab8dade125", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_claude(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " stream = claude.chat.completions.create(\n", + " model=claude_model,\n", + " messages=messages,\n", + " max_tokens=1000,\n", + " stream=True\n", + " )\n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6d7e0d48-6140-484c-81aa-2f6aa6da8f25", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_gemini(prompt):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": prompt}\n", + " ]\n", + " stream = gemini.chat.completions.create(\n", + " model=gemini_model,\n", + " messages=messages,\n", + " stream=True\n", + " )\n", + " result = \"\"\n", + " for chunk in stream:\n", + " result += chunk.choices[0].delta.content or \"\"\n", + " yield result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0087623a-4e31-470b-b2e6-d8d16fc7bcf5", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_model(prompt, model):\n", + " if model==\"GPT\":\n", + " result = stream_gpt(prompt)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(prompt)\n", + " elif model==\"Gemini\":\n", + " result = stream_gemini(prompt)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " yield from result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8d8ce810-997c-4b6a-bc4f-1fc847ac8855", + "metadata": {}, + "outputs": [], + "source": [ + "view = gr.Interface(\n", + " fn=stream_model,\n", + " inputs=[gr.Textbox(label=\"Your message:\"), gr.Dropdown([\"GPT\", \"Claude\", \"Gemini\"], label=\"Select model\", value=\"GPT\")],\n", + " outputs=[gr.Markdown(label=\"Response:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + }, + { + "cell_type": "markdown", + "id": "d933865b-654c-4b92-aa45-cf389f1eda3d", + "metadata": {}, + "source": [ + "# Building a company brochure generator\n", + "\n", + "Now you know how - it's simple!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1626eb2e-eee8-4183-bda5-1591b58ae3cf", + "metadata": {}, + "outputs": [], + "source": [ + "# A class to represent a Webpage\n", + "\n", + "class Website:\n", + " url: str\n", + " title: str\n", + " text: str\n", + "\n", + " def __init__(self, url):\n", + " self.url = url\n", + " response = requests.get(url)\n", + " self.body = response.content\n", + " soup = BeautifulSoup(self.body, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n", + "\n", + " def get_contents(self):\n", + " return f\"Webpage Title:\\n{self.title}\\nWebpage Contents:\\n{self.text}\\n\\n\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c701ec17-ecd5-4000-9f68-34634c8ed49d", + "metadata": {}, + "outputs": [], + "source": [ + "# With massive thanks to Bill G. who noticed that a prior version of this had a bug! Now fixed.\n", + "\n", + "base_system_message = \"You are an assistant that analyzes the contents of a company website landing page \\\n", + "and creates a short brochure about the company for prospective customers, investors and recruits. Respond in markdown.\"\n", + "system_message = base_system_message" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "11e9debb-9500-4783-a72e-fc3659214a8e", + "metadata": {}, + "outputs": [], + "source": [ + "def system_personality(personality) -> str:\n", + " match personality:\n", + " case \"Hostile\":\n", + " return base_system_message + \" Use a critical and sarcastic tone that highlights flaws, inconsistencies, or poor design choices in the company's website.\"\n", + " case \"Formal\":\n", + " return base_system_message + \" Use a professional and respectful tone, with precise language and a structured presentation that inspires trust.\"\n", + " case \"Funny\":\n", + " return base_system_message + \" Use a lighthearted and humorous tone, incorporating playful language, witty remarks and engaging expressions.\"\n", + " case _:\n", + " return base_system_message" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5def90e0-4343-4f58-9d4a-0e36e445efa4", + "metadata": {}, + "outputs": [], + "source": [ + "def stream_brochure(company_name, url, model, personality):\n", + " yield \"\"\n", + " prompt = f\"Please generate a company brochure for {company_name}. Here is their landing page:\\n\"\n", + " prompt += Website(url).get_contents()\n", + " global system_message\n", + " system_message = system_personality(personality)\n", + " if model==\"GPT\":\n", + " result = stream_gpt(prompt)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(prompt)\n", + " elif model==\"Gemini\":\n", + " result = stream_gemini(prompt)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " yield from result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "66399365-5d67-4984-9d47-93ed26c0bd3d", + "metadata": {}, + "outputs": [], + "source": [ + "view = gr.Interface(\n", + " fn=stream_brochure,\n", + " inputs=[\n", + " gr.Textbox(label=\"Company name:\"),\n", + " gr.Textbox(label=\"Landing page URL including http:// or https://\"),\n", + " gr.Dropdown([\"GPT\", \"Claude\", \"Gemini\"], label=\"Select model\"),\n", + " gr.Dropdown([\"Funny\",\"Formal\", \"Hostile\"], label=\"Select a personality\")],\n", + " outputs=[gr.Markdown(label=\"Brochure:\")],\n", + " flagging_mode=\"never\"\n", + ")\n", + "view.launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day5_stock-assistant-with-tools.ipynb b/week2/community-contributions/day5_stock-assistant-with-tools.ipynb new file mode 100644 index 0000000..1e129aa --- /dev/null +++ b/week2/community-contributions/day5_stock-assistant-with-tools.ipynb @@ -0,0 +1,968 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "bcb31876-4d8c-41ef-aa24-b8c78dfd5808", + "metadata": {}, + "source": [ + "# Project - Stock Information AI Assistant\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b7bd1bd7-19d9-4c4b-bc4b-9bc9cca8bd0f", + "metadata": {}, + "outputs": [], + "source": [ + "!pip install finnhub-python" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8b50bbe2-c0b1-49c3-9a5c-1ba7efa2bcb4", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import json\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import finnhub\n", + "from typing import Dict, List, Any, Optional\n", + "from datetime import datetime" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ba0ddc1a-c775-4ed3-9531-ed0c5799e87f", + "metadata": {}, + "outputs": [], + "source": [ + "import logging\n", + "\n", + "# Configure root logger\n", + "logging.basicConfig(\n", + " level=logging.INFO, # Set level: DEBUG, INFO, WARNING, ERROR\n", + " format=\"%(asctime)s [%(levelname)s] %(message)s\", \n", + " force=True # Ensures reconfiguration if you rerun this cell\n", + ")\n", + "\n", + "logger = logging.getLogger(__name__) # Use a global logger object\n", + "logger.info(\"Logger initialized!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "747e8786-9da8-4342-b6c9-f5f69c2e22ae", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "FINNHUB_API_KEY = os.getenv(\"FINNHUB_API_KEY\")\n", + "\n", + "if openai_api_key:\n", + " logger.info(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " logger.error(\"OpenAI API Key not set\")\n", + "\n", + "if FINNHUB_API_KEY:\n", + " logger.info(f\"FINNHUB_API_KEY exists!\")\n", + "else:\n", + " logger.error(\"OpenAI API Key not set\")\n", + " \n", + "MODEL = \"gpt-4.1-mini\" # not using gpt-5-mini as openai doesn't let you stream responses till you are a verified organisation :(\n", + "openai = OpenAI()\n", + "finnhub_client = finnhub.Client(api_key=FINNHUB_API_KEY)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee3aaa9a-5495-42fd-a382-803fbfa92eaf", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = f\"\"\"\n", + "You are \"TickerBot\" — a concise, factual, educational assistant specializing in U.S. stocks. \n", + "Your job: quickly and accurately explain stock and company information in plain English. NEVER give investment advice, buy/sell recommendations, or price predictions.\n", + "\n", + "## PRIVACY ABOUT IMPLEMENTATION\n", + "- Do not reveal any internal implementation details to users. Never display or mention internal tool names, API names, developer notes, configured flags, date-range limits, or other system/developer constraints in user-facing replies.\n", + "- All runtime/tool constraints and capability detection are internal. Present only user-facing capabilities in plain language.\n", + "\n", + "## USER-FACING CAPABILITIES\n", + "- When asked \"What can you do?\", list only stock-relevant actions in plain language. Example reply:\n", + " \"I can look up tickers, show the latest quotes, provide key company financials and latest earnings details, summarize recent company or market headlines, and give a brief market overview.\"\n", + "- Do not list internal utilities or developer tools as user-facing capabilities.\n", + "\n", + "## GENERAL PRINCIPLES\n", + "- Answer only what was asked for. \n", + "- Be brief, clear, and professional while still maintaining a warm tone. Use short paragraphs and one-line bullet explanations when requested.\n", + "- Return only what the system provides; do not invent, infer, or extrapolate unavailable data.\n", + "- Never offer or advertise any feature the environment does not actually support. Avoid offering attachments, direct downloads, or full-text article retrieval unless the system explicitly provides those outputs.\n", + "\n", + "## Behavior Rules\n", + "- Stay professional and neutral at all times. \n", + "- Clarify only when user intent is ambiguous; never guess. \n", + "- Only disclose information the user explicitly requested. \n", + "- Never explain system limits (e.g., API ranges, date limits) ever. \n", + "- Summaries should be tight and relevant, not verbose. \n", + "\n", + "## NEWS & HEADLINES\n", + "- When interpreting date-related or temporal reasoning requests (e.g., “latest earnings,” “recent news,” “Q1 results”) Call `get_current_time` to determine the current date.\n", + "- Present news/headlines in concise bullet lines when requested. Default recent-window behavior is internal; do not describe or expose internal default windows or limits to the user.\n", + "- If the system only returns headlines/summaries, present those and do not offer to fetch full-text or additional ranges unless the user explicitly asks and the environment supports that action.\n", + "\n", + "## FOLLOW-UP & CLARIFYING QUESTIONS\n", + "- If no matching stock symbol is found, ask the user to clarify the name or ticker. Mention you only support U.S. stocks. If they confirm the symbol but no data exists, state that no results were found.\n", + "- Never append unsolicited menus, multi-choice lists, or repeated \"Would you like...\" prompts at the end of a normal reply.\n", + "- Ask a single direct clarifying question only when strictly necessary to fulfill the user's request (for example: ambiguous company name or missing ticker). That single question must be the final line of the reply.\n", + "- If the user's intent is clear, proceed and return results. Do not request confirmations or offer options unless required to complete the task.\n", + "\n", + "## MISSING-DATA / NOTE RULES\n", + "- Do NOT call out missing/unavailable single fields unless:\n", + " 1) the missing field was explicitly requested by the user; OR\n", + " 2) multiple (>1) key metrics required to answer the user's request are unavailable and their absence materially prevents a useful answer.\n", + "- If condition (1) or (2) applies, include at most one concise \"Note:\" line naming the specific missing field(s) and nothing else.\n", + "- Otherwise omit any comment about single, non-central missing fields.\n", + "\n", + "## ERROR HANDLING\n", + "- If a company/ticker cannot be found: \"I couldn't find that company on U.S. exchanges. Please provide the ticker or a clearer company name.\"\n", + "- If partial results are returned: present available data and at most one short note: \"Some data wasn't available: , .\"\n", + "- If you do not know the answer, say plainly: \"I don't have that information.\"\n", + "\n", + "## ADVICE / DISCLAIMER (CONDITIONAL)\n", + "- Do NOT include an investment-advice disclaimer in normal greetings or factual replies.\n", + "- Show a single-line disclaimer only when the user explicitly asks for recommendations/advice or uses clear advice-seeking language (e.g., \"should I buy\", \"recommend\", \"what should I do\", \"financial advice\").\n", + "- Disclaimer text when triggered: \"I can't provide investment advice or buy/sell recommendations. I can provide facts and explain metrics.\"\n", + "\n", + "EXAMPLES (templates — follow exactly)\n", + "\n", + "1) Clear factual request (no unsolicited follow-up)\n", + "User: \"What's AAPL last trade?\"\n", + "Assistant:\n", + "Last trade timestamp: 2025-08-28 20:00:00 UTC (U.S. market close — 4:00 PM ET)\n", + "\n", + "(—end: no menu, no offers—)\n", + "\n", + "2) Ambiguous request — single clarifying question only\n", + "User: \"Show me Apple\"\n", + "Assistant:\n", + "I found two matches. Do you mean Apple Inc. (AAPL) or Apple Hospitality REIT (APLE)?\n", + "\n", + "(—end: single question final line—)\n", + "\n", + "3) News returned as headlines only (do not offer unsupported fetches)\n", + "User: \"Recent NVDA news\"\n", + "Assistant:\n", + "- 2025-08-28: Headline A — short summary.\n", + "- 2025-08-26: Headline B — short summary.\n", + "(—end: do not offer full-article fetch or extended-range offers unless explicitly requested and supported—)\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fdf1a2b0-07be-47a0-9ce3-14d21b48c8f2", + "metadata": {}, + "outputs": [], + "source": [ + "def get_current_time() -> Dict[str, Any]:\n", + " \"\"\"\n", + " Retrieve the current UTC time in ISO format with timezone.\n", + " Returns a dictionary for consistency with other tools.\n", + " \"\"\"\n", + " try:\n", + " current_time = datetime.utcnow().isoformat() + 'Z'\n", + " return {\n", + " \"success\": True,\n", + " \"current_time\": current_time\n", + " }\n", + " except Exception as e:\n", + " return {\"success\": False, \"error\": f\"Failed to get time: {str(e)[:100]}\"}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "12d912fc-91fb-469e-9572-2876a099f5aa", + "metadata": {}, + "outputs": [], + "source": [ + "get_current_time_function = {\n", + " \"name\": \"get_current_time\",\n", + " \"description\": \"Get the current UTC time in ISO format (YYYY-MM-DDTHH:MM:SS.ssssssZ). Useful for temporal reasoning, date calculations, or setting time ranges for queries like news.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {}, # No parameters needed\n", + " \"required\": []\n", + " }\n", + "}\n", + "get_current_time_tool = {\"type\": \"function\", \"function\": get_current_time_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "61a2a15d-b559-4844-b377-6bd5cb4949f6", + "metadata": {}, + "outputs": [], + "source": [ + "def validate_symbol(symbol: str) -> bool:\n", + " \"\"\"Validate stock symbol format\"\"\"\n", + " if not symbol or not isinstance(symbol, str):\n", + " return False\n", + " return symbol.isalnum() and 1 <= len(symbol) <= 5 and symbol.isupper()\n", + "\n", + "def search_symbol(query: str) -> Dict[str, Any]:\n", + " \"\"\"Search for stock symbol using Finnhub client\"\"\"\n", + " logger.info(f\"Tool search_symbol called for {query}\")\n", + " try:\n", + " if not query or len(query.strip()) < 1:\n", + " return {\"success\": False, \"error\": \"Invalid search query\"}\n", + " \n", + " query = query.strip()[:50]\n", + " result = finnhub_client.symbol_lookup(query)\n", + " logger.info(f\"Tool search_symbol {result}\")\n", + " \n", + " if result.get(\"result\") and len(result[\"result\"]) > 0:\n", + " first_result = result[\"result\"][0]\n", + " symbol = first_result.get(\"symbol\", \"\").upper()\n", + " \n", + " if validate_symbol(symbol):\n", + " return {\n", + " \"success\": True,\n", + " \"symbol\": symbol\n", + " }\n", + " else:\n", + " return {\"success\": False, \"error\": \"Invalid symbol format found\"}\n", + " else:\n", + " return {\"success\": False, \"error\": \"No matching US stocks found\"}\n", + " \n", + " except Exception as e:\n", + " return {\"success\": False, \"error\": f\"Symbol search failed: {str(e)[:100]}\"}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "173010e3-dfef-4611-8b68-d11256bd5fba", + "metadata": {}, + "outputs": [], + "source": [ + "search_symbol_function = {\n", + " \"name\": \"search_symbol\",\n", + " \"description\": \"Search for a stock symbol / ticker symbol based on company name or partial name\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"query\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Company name or partial name to search for, extract only relevant name part and pass it here, keep this to less than 50 characters\"\n", + " }\n", + " },\n", + " \"required\": [\n", + " \"query\"\n", + " ]\n", + " }\n", + "}\n", + "\n", + "search_symbol_tool = {\"type\": \"function\", \"function\": search_symbol_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "448bb4ce-8e86-4ceb-ab52-96bddfd33337", + "metadata": {}, + "outputs": [], + "source": [ + "def _format_big_number_from_millions(value_millions: Any) -> str:\n", + " \"\"\"\n", + " Finnhub returns some large metrics (marketCapitalization, enterpriseValue, revenueTTM)\n", + " in MILLIONS USD. Convert to full USD and format with M/B/T suffixes.\n", + " \"\"\"\n", + " if value_millions is None:\n", + " return \"Unavailable\"\n", + " try:\n", + " value = float(value_millions) * 1_000_000 # convert millions -> full USD\n", + " except (TypeError, ValueError):\n", + " return \"Unavailable\"\n", + "\n", + " trillion = 1_000_000_000_000\n", + " billion = 1_000_000_000\n", + " million = 1_000_000\n", + "\n", + " if value >= trillion:\n", + " return f\"{value / trillion:.2f}T USD\"\n", + " if value >= billion:\n", + " return f\"{value / billion:.2f}B USD\"\n", + " if value >= million:\n", + " return f\"{value / million:.2f}M USD\"\n", + " return f\"{value:.2f} USD\"\n", + "\n", + "\n", + "def _safe_metric(metrics: Dict[str, Any], key: str) -> Any:\n", + " \"\"\"\n", + " Return metric value if present; otherwise \"Unavailable\".\n", + " We intentionally return the raw value for numeric metrics (no rounding/format)\n", + " except for the specially formatted big-number fields handled elsewhere.\n", + " \"\"\"\n", + " if metrics is None:\n", + " return \"Unavailable\"\n", + " val = metrics.get(key)\n", + " return val if val is not None else \"Unavailable\"\n", + "\n", + "\n", + "def get_company_financials(symbol: str) -> Dict[str, Any]:\n", + " \"\"\"\n", + " Fetch and return a curated set of 'basic' financial metrics for `symbol`.\n", + " - Calls finnhub_client.company_basic_financials(symbol, 'all')\n", + " - Formats market cap, enterprise value, revenue (Finnhub returns these in millions)\n", + " - Returns success flag and readable keys\n", + " \"\"\"\n", + " logger.info(f\"Tool get_company_financials called for {symbol}\")\n", + " try:\n", + " if not symbol or not symbol.strip():\n", + " return {\"success\": False, \"error\": \"Invalid stock symbol\"}\n", + "\n", + " symbol = symbol.strip().upper()\n", + "\n", + " # --- API Call ---\n", + " financials_resp = finnhub_client.company_basic_financials(symbol, \"all\")\n", + "\n", + " # Finnhub places primary values under \"metric\"\n", + " metrics = financials_resp.get(\"metric\", {})\n", + " if not metrics:\n", + " return {\"success\": False, \"error\": \"No financial metrics found\"}\n", + "\n", + " # --- Build result using helpers ---\n", + " result = {\n", + " \"success\": True,\n", + " \"symbol\": symbol,\n", + " \"financials\": {\n", + " \"Market Cap\": _format_big_number_from_millions(metrics.get(\"marketCapitalization\")),\n", + " \"Enterprise Value\": _format_big_number_from_millions(metrics.get(\"enterpriseValue\")),\n", + " \"P/E Ratio (TTM)\": _safe_metric(metrics, \"peBasicExclExtraTTM\"),\n", + " \"Forward P/E\": _safe_metric(metrics, \"forwardPE\"),\n", + " \"Gross Margin (TTM)\": _safe_metric(metrics, \"grossMarginTTM\"),\n", + " \"Net Profit Margin (TTM)\": _safe_metric(metrics, \"netProfitMarginTTM\"),\n", + " \"EPS (TTM)\": _safe_metric(metrics, \"epsTTM\"),\n", + " \"EPS Growth (5Y)\": _safe_metric(metrics, \"epsGrowth5Y\"),\n", + " \"Dividend Yield (Indicated Annual)\": _safe_metric(metrics, \"dividendYieldIndicatedAnnual\"),\n", + " \"Current Ratio (Quarterly)\": _safe_metric(metrics, \"currentRatioQuarterly\"),\n", + " \"Debt/Equity (Long Term, Quarterly)\": _safe_metric(metrics, \"longTermDebt/equityQuarterly\"),\n", + " \"Beta\": _safe_metric(metrics, \"beta\"),\n", + " \"52-Week High\": _safe_metric(metrics, \"52WeekHigh\"),\n", + " \"52-Week Low\": _safe_metric(metrics, \"52WeekLow\"),\n", + " }\n", + " }\n", + "\n", + " return result\n", + "\n", + " except Exception as e:\n", + " # keep error message short but useful for debugging\n", + " return {\"success\": False, \"error\": f\"Failed to fetch metrics: {str(e)[:200]}\"}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9df7b74e-fec8-4e75-92a9-31acc75e6e97", + "metadata": {}, + "outputs": [], + "source": [ + "get_company_financials_function = {\n", + " \"name\": \"get_company_financials\",\n", + " \"description\": \"Fetch and return a curated set of basic financial metrics for a stock symbol. Calls Finnhub's company_basic_financials API, formats large numbers (market cap, enterprise value, revenue) in M/B/T USD, and shows metrics like P/E ratios, EPS, margins, dividend yield, debt/equity, beta, and 52-week range. Returns 'Unavailable' for missing values.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"symbol\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Stock ticker symbol to fetch metrics for. Example: 'AAPL' for Apple Inc.\"\n", + " }\n", + " },\n", + " \"required\": [\n", + " \"symbol\"\n", + " ]\n", + " }\n", + "}\n", + "\n", + "\n", + "get_company_financials_tool = {\"type\": \"function\", \"function\": get_company_financials_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cfeeb200-3f30-4855-82b9-cc8b2a950f80", + "metadata": {}, + "outputs": [], + "source": [ + "def get_stock_quote(symbol: str) -> dict:\n", + " \"\"\"\n", + " Fetch the latest stock quote for a given ticker symbol using Finnhub's /quote endpoint.\n", + " Returns current price, daily high/low, open, previous close, percent change, and readable timestamp.\n", + " \"\"\"\n", + " logger.info(f\"Tool get_stock_quote called for {symbol}\")\n", + " try:\n", + " if not symbol or len(symbol.strip()) < 1:\n", + " return {\"success\": False, \"error\": \"Invalid symbol provided\"}\n", + " \n", + " symbol = symbol.strip().upper()\n", + " data = finnhub_client.quote(symbol)\n", + "\n", + " if not data or \"c\" not in data:\n", + " return {\"success\": False, \"error\": \"No quote data found\"}\n", + " \n", + " # Convert epoch timestamp to ISO UTC if present\n", + " timestamp = data.get(\"t\")\n", + " if timestamp and isinstance(timestamp, (int, float)):\n", + " timestamp = datetime.utcfromtimestamp(timestamp).isoformat() + \"Z\"\n", + " else:\n", + " timestamp = \"Unavailable\"\n", + " \n", + " return {\n", + " \"success\": True,\n", + " \"symbol\": symbol,\n", + " \"current_price\": round(data.get(\"c\", 0), 2) if data.get(\"c\") is not None else \"Unavailable\",\n", + " \"change\": round(data.get(\"d\", 0), 2) if data.get(\"d\") is not None else \"Unavailable\",\n", + " \"percent_change\": f\"{round(data.get('dp', 0), 2)}%\" if data.get(\"dp\") is not None else \"Unavailable\",\n", + " \"high_price\": round(data.get(\"h\", 0), 2) if data.get(\"h\") is not None else \"Unavailable\",\n", + " \"low_price\": round(data.get(\"l\", 0), 2) if data.get(\"l\") is not None else \"Unavailable\",\n", + " \"open_price\": round(data.get(\"o\", 0), 2) if data.get(\"o\") is not None else \"Unavailable\",\n", + " \"previous_close\": round(data.get(\"pc\", 0), 2) if data.get(\"pc\") is not None else \"Unavailable\",\n", + " \"timestamp\": timestamp\n", + " }\n", + " except Exception as e:\n", + " return {\"success\": False, \"error\": f\"Quote retrieval failed: {str(e)[:100]}\"}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3724d92a-4515-4267-af6f-2c1ec2b6ed36", + "metadata": {}, + "outputs": [], + "source": [ + "get_stock_quote_function = {\n", + " \"name\": \"get_stock_quote\",\n", + " \"description\": \"Retrieve the latest stock quote for a given symbol, including current price, daily high/low, open, previous close, and percent change. Data is near real-time. Avoid constant polling; use websockets for streaming updates.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"symbol\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Stock ticker symbol to fetch the latest quote for. Example: 'AAPL', 'MSFT'.\"\n", + " }\n", + " },\n", + " \"required\": [\"symbol\"]\n", + " }\n", + "}\n", + "\n", + "get_stock_quote_tool = {\"type\": \"function\", \"function\": get_stock_quote_function}\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "62f5d477-6626-428f-b8eb-d763e736ef5b", + "metadata": {}, + "outputs": [], + "source": [ + "def get_company_news(symbol: str, _from: str, to: str):\n", + " \"\"\"\n", + " Fetch the top latest company news for a stock symbol within a date range.\n", + " - Ensures the range does not exceed ~1 months (35 days).\n", + " - Best practice: Keep searches to a month or less to avoid too much data.\n", + "\n", + " Args:\n", + " symbol (str): Stock ticker (e.g., \"AAPL\").\n", + " _from (str): Start date in YYYY-MM-DD format.\n", + " to (str): End date in YYYY-MM-DD format.\n", + "\n", + " Returns:\n", + " list or dict: Cleaned news data or error message.\n", + " \"\"\"\n", + " # Validate date format\n", + " logger.info(f\"Tool get_company_news called for {symbol} from {_from} to {to}\")\n", + " try:\n", + " start_date = datetime.strptime(_from, \"%Y-%m-%d\")\n", + " end_date = datetime.strptime(to, \"%Y-%m-%d\")\n", + " except ValueError:\n", + " return {\"success\": False, \"error\": \"Invalid date format. Use YYYY-MM-DD.\"}\n", + "\n", + " # Check date range\n", + " delta_days = (end_date - start_date).days\n", + " if delta_days > 35:\n", + " return {\n", + " \"success\": False, \n", + " \"error\": f\"Date range too large ({delta_days} days). \"\n", + " \"Please use a range of 1 months or less.\"\n", + " }\n", + "\n", + " # Fetch data\n", + " try:\n", + " news = finnhub_client.company_news(symbol, _from=_from, to=to)\n", + " except Exception as e:\n", + " return {\"success\": False, \"error\": str(e)}\n", + "\n", + " # Do not want to report just the latest news in the time period\n", + " if len(news) <= 10:\n", + " # If 10 or fewer articles, take all\n", + " selected_news = news\n", + " else:\n", + " # Take first 5 (oldest) and last 5 (newest)\n", + " selected_news = news[:5] + news[-5:]\n", + "\n", + " # Clean & transform objects\n", + " cleaned_news = []\n", + " for article in selected_news:\n", + " cleaned_news.append({\n", + " \"summary\": article.get(\"summary\"),\n", + " \"source\": article.get(\"source\"),\n", + " \"published_at\": datetime.utcfromtimestamp(article[\"datetime\"]).strftime(\"%Y-%m-%d %H:%M:%S UTC\"),\n", + " \"related\": article.get(\"related\")\n", + " })\n", + "\n", + " return {\"success\": True, \"news\": cleaned_news}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5150ecb6-e3f1-46dc-94fa-2a9abe5165f6", + "metadata": {}, + "outputs": [], + "source": [ + "get_company_news_function = {\n", + " \"name\": \"get_company_news\",\n", + " \"description\": \"Fetch the top most recent company news articles for a given stock symbol. ⚠️ Avoid querying more than a 1-month range at a time as it may return too much data. Only tells news about company within last 1 year. An error is returned if the requested time range exceeds 1 month.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"symbol\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Stock ticker symbol, e.g., 'AAPL'.\"\n", + " },\n", + " \"_from\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Start date in YYYY-MM-DD format. Ensure it is not more than 1 year ago from today. Ensure it is before or equal to the date in to.\"\n", + " },\n", + " \"to\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"End date in YYYY-MM-DD format. Ensure it is not more than 1 year ago. Ensure it is after or equal to the date in from.\"\n", + " }\n", + " },\n", + " \"required\": [\n", + " \"symbol\",\n", + " \"_from\",\n", + " \"to\"\n", + " ]\n", + " }\n", + "}\n", + "\n", + "get_company_news_tool = {\"type\": \"function\", \"function\": get_company_news_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "26dd7375-626f-4235-b4a2-f1926f62cc5e", + "metadata": {}, + "outputs": [], + "source": [ + "def get_market_news(category: str = \"general\"):\n", + " \"\"\"\n", + " Fetch the latest market news for a given category.\n", + "\n", + " Args:\n", + " category (str): News category. One of [\"general\", \"forex\", \"crypto\", \"merger\"].\n", + "\n", + " Returns:\n", + " list or dict: A cleaned list of news articles or error message.\n", + " \"\"\"\n", + " logger.info(f\"Tool get_market_news called for category '{category}'\")\n", + "\n", + " try:\n", + " news = finnhub_client.general_news(category)\n", + " except Exception as e:\n", + " logger.error(f\"Tool get_market_news API call failed!\")\n", + " return {\"success\": False, \"error\": str(e)}\n", + "\n", + " # Do not want to report just the latest news in the time period\n", + " if len(news) <= 10:\n", + " # If 10 or fewer articles, take all\n", + " selected_news = news\n", + " else:\n", + " # Take first 5 (oldest) and last 5 (newest)\n", + " selected_news = news[:5] + news[-5:]\n", + "\n", + " # Clean & transform objects\n", + " cleaned_news = []\n", + " for article in selected_news:\n", + " cleaned_news.append({\n", + " \"headline\": article.get(\"headline\"),\n", + " \"summary\": article.get(\"summary\"),\n", + " \"source\": article.get(\"source\"),\n", + " \"category\": article.get(\"category\"),\n", + " \"related\": article.get(\"related\")\n", + " })\n", + "\n", + " return {\"success\": True, \"news\": cleaned_news}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5bd1aa28-119c-4c7a-bdc0-161a582ab1cc", + "metadata": {}, + "outputs": [], + "source": [ + "get_market_news_function = {\n", + " \"name\": \"get_market_news\",\n", + " \"description\": \"Fetch the latest market news by category. Returns the top 10 news articles with headline, summary, source, category, published time (UTC), and URLs. Categories: general, forex, crypto, merger. Use this to quickly get relevant financial news.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"category\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"News category to fetch. One of: general, forex, crypto, merger.\"\n", + " }\n", + " },\n", + " \"required\": [\"category\"]\n", + " }\n", + "}\n", + "\n", + "get_market_news_tool = {\"type\": \"function\", \"function\": get_market_news_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fbe8ef6c-2d88-43a2-94dc-70b507fe9cd2", + "metadata": {}, + "outputs": [], + "source": [ + "def get_earnings_calendar(symbol: str = \"\", _from: str = \"\", to: str = \"\"):\n", + " \"\"\"\n", + " Fetch LATEST earnings calendar data for a stock symbol within a date range.\n", + " - End date must be within the last month. (Free tier only allows last 1 month data)\n", + " - Shows historical and upcoming earnings releases with EPS and revenue data.\n", + " Args:\n", + " symbol (str): Stock ticker (e.g., \"AAPL\"). Leave empty for all companies.\n", + " _from (str): Start date in YYYY-MM-DD format.\n", + " to (str): End date in YYYY-MM-DD format.\n", + " Returns:\n", + " list or dict: Cleaned earnings calendar data or error message.\n", + " \"\"\"\n", + " logger.info(f\"Tool get_earnings_calendar called for {symbol or 'all symbols'} from {_from} to {to}\")\n", + " \n", + " # Validate date format if provided\n", + " if _from or to:\n", + " try:\n", + " start_date = datetime.strptime(_from, \"%Y-%m-%d\") if _from else None\n", + " end_date = datetime.strptime(to, \"%Y-%m-%d\") if to else None\n", + " \n", + " # Check date range if both dates provided\n", + " # Check if end_date is within 1 month (≈30 days) of today\n", + " if end_date:\n", + " today = datetime.utcnow()\n", + " if (today - end_date).days > 30:\n", + " return {\n", + " \"success\": False,\n", + " \"error\": \"End date must be within the last month.\"\n", + " }\n", + " except ValueError:\n", + " return {\"success\": False, \"error\": \"Invalid date format. Use YYYY-MM-DD.\"}\n", + " \n", + " # Fetch earnings calendar data\n", + " try:\n", + " earnings_data = finnhub_client.earnings_calendar(_from=_from, to=to, symbol=symbol, international=False)\n", + " except Exception as e:\n", + " logger.error(f\"Error fetching earnings calendar: {e}\")\n", + " return {\"success\": False, \"error\": str(e)}\n", + " \n", + " # Check if data exists\n", + " if not earnings_data or \"earningsCalendar\" not in earnings_data:\n", + " return {\"success\": False, \"error\": \"No earnings data available for the specified criteria.\"}\n", + " \n", + " earnings_list = earnings_data[\"earningsCalendar\"]\n", + " \n", + " if not earnings_list:\n", + " return {\"success\": True, \"earnings\": [], \"message\": \"No earnings releases found for the specified period.\"}\n", + " \n", + " # Clean & transform earnings data\n", + " cleaned_earnings = []\n", + " for earning in earnings_list:\n", + " # Format hour description\n", + " hour_map = {\n", + " \"bmo\": \"Before Market Open\",\n", + " \"amc\": \"After Market Close\", \n", + " \"dmh\": \"During Market Hours\"\n", + " }\n", + " \n", + " cleaned_earnings.append({\n", + " \"symbol\": earning.get(\"symbol\"),\n", + " \"date\": earning.get(\"date\"),\n", + " \"quarter\": f\"Q{earning.get('quarter')} {earning.get('year')}\",\n", + " \"announcement_time\": hour_map.get(earning.get(\"hour\", \"\"), earning.get(\"hour\", \"Unknown\")),\n", + " \"eps_actual\": earning.get(\"epsActual\"),\n", + " \"eps_estimate\": earning.get(\"epsEstimate\"),\n", + " \"revenue_actual\": earning.get(\"revenueActual\"),\n", + " \"revenue_estimate\": earning.get(\"revenueEstimate\")\n", + " })\n", + " \n", + " return {\"success\": True, \"earnings\": cleaned_earnings}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9eaeae75-d68f-4160-a26e-c13e40cf756b", + "metadata": {}, + "outputs": [], + "source": [ + "get_earnings_calendar_function = {\n", + " \"name\": \"get_earnings_calendar\",\n", + " \"description\": \"Fetch latest earnings calendar showing historical and upcoming earnings releases for companies. Shows EPS and revenue estimates vs actuals. End date must be within the last month.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"symbol\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Stock ticker symbol, e.g., 'AAPL'. Leave empty to get earnings for all companies in the date range.\"\n", + " },\n", + " \"_from\": {\n", + " \"type\": \"string\", \n", + " \"description\": \"Start date in YYYY-MM-DD format. Ensure it is not more than 1 year ago from today. Ensure it is before or equal to the date in to.\"\n", + " },\n", + " \"to\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"End date in YYYY-MM-DD format. Ensure it is not more than 1 year ago. Ensure it is after or equal to the date in from. To date must be within the last month.\"\n", + " }\n", + " },\n", + " \"required\": [\n", + " \"_from\",\n", + " \"to\"\n", + " ]\n", + " }\n", + "}\n", + "\n", + "get_earnings_calendar_tool = {\"type\": \"function\", \"function\": get_earnings_calendar_function}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bdca8679-935f-4e7f-97e6-e71a4d4f228c", + "metadata": {}, + "outputs": [], + "source": [ + "# List of tools:\n", + "tools = [search_symbol_tool, get_company_financials_tool, get_stock_quote_tool, get_company_news_tool, get_market_news_tool, get_current_time_tool, get_earnings_calendar_tool]\n", + "tool_functions = {\n", + " \"search_symbol\": search_symbol,\n", + " \"get_company_financials\": get_company_financials,\n", + " \"get_stock_quote\": get_stock_quote,\n", + " \"get_company_news\": get_company_news,\n", + " \"get_market_news\": get_market_news,\n", + " \"get_earnings_calendar\": get_earnings_calendar,\n", + " \"get_current_time\": get_current_time\n", + "}" + ] + }, + { + "cell_type": "markdown", + "id": "c3d3554f-b4e3-4ce7-af6f-68faa6dd2340", + "metadata": {}, + "source": [ + "## Getting OpenAI to use our Tool\n", + "\n", + "There's some fiddly stuff to allow OpenAI \"to call our tool\"\n", + "\n", + "What we actually do is give the LLM the opportunity to inform us that it wants us to run the tool.\n", + "\n", + "Here's how the new chat function looks:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "86f76f57-76c4-4dc7-94a8-cfe7816a39f1", + "metadata": {}, + "outputs": [], + "source": [ + "def execute_tool_call(tool_call):\n", + " func_name = tool_call.function.name\n", + " args = json.loads(tool_call.function.arguments)\n", + "\n", + " logger.info(f\"Executing tool: {func_name} with args: {args}\")\n", + "\n", + " func = tool_functions.get(func_name)\n", + " if not func:\n", + " result = {\"error\": f\"Function '{func_name}' not found\"}\n", + " else:\n", + " try:\n", + " result = func(**args)\n", + " except Exception as e:\n", + " logger.exception(f\"Error executing {func_name}\")\n", + " result = {\"error\": str(e)}\n", + "\n", + " return {\n", + " \"role\": \"tool\",\n", + " \"tool_call_id\": tool_call.id,\n", + " \"content\": json.dumps(result)\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ce9b0744-9c78-408d-b9df-9f6fd9ed78cf", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history + [{\"role\": \"user\", \"content\": message}]\n", + "\n", + " # Skip the first system message\n", + " to_log = messages[1:]\n", + "\n", + " # Print each dict on its own line\n", + " logger.info(\"\\nMessages:\\n\" + \"\\n\".join(str(m) for m in to_log) + \"\\n\")\n", + "\n", + " while True:\n", + " response = openai.chat.completions.create(\n", + " model=MODEL, \n", + " messages=messages, \n", + " tools=tools,\n", + " stream=True\n", + " )\n", + " \n", + " content = \"\"\n", + " tool_calls = []\n", + " finish_reason = None\n", + " \n", + " # Process the stream\n", + " for chunk in response:\n", + " choice = chunk.choices[0]\n", + " finish_reason = choice.finish_reason\n", + " \n", + " # Stream content\n", + " if choice.delta.content:\n", + " content += choice.delta.content\n", + " yield content\n", + " \n", + " # Collect tool calls\n", + " if choice.delta.tool_calls:\n", + " for tc_delta in choice.delta.tool_calls:\n", + " # Extend tool_calls list if needed\n", + " while len(tool_calls) <= tc_delta.index:\n", + " tool_calls.append({\n", + " \"id\": \"\",\n", + " \"function\": {\"name\": \"\", \"arguments\": \"\"}\n", + " })\n", + " \n", + " tc = tool_calls[tc_delta.index]\n", + " if tc_delta.id:\n", + " tc[\"id\"] = tc_delta.id\n", + " if tc_delta.function:\n", + " if tc_delta.function.name:\n", + " tc[\"function\"][\"name\"] = tc_delta.function.name\n", + " if tc_delta.function.arguments:\n", + " tc[\"function\"][\"arguments\"] += tc_delta.function.arguments\n", + " \n", + " # If no tool calls, we're done\n", + " if finish_reason != \"tool_calls\":\n", + " return content\n", + " \n", + " # Execute tools\n", + " ai_message = {\n", + " \"role\": \"assistant\", \n", + " \"content\": content,\n", + " \"tool_calls\": [\n", + " {\n", + " \"id\": tc[\"id\"],\n", + " \"type\": \"function\",\n", + " \"function\": tc[\"function\"]\n", + " } for tc in tool_calls\n", + " ]\n", + " }\n", + " \n", + " tool_responses = []\n", + " for tool_call in ai_message[\"tool_calls\"]:\n", + " # Convert dict back to object for your existing function\n", + " class ToolCall:\n", + " def __init__(self, tc_dict):\n", + " self.id = tc_dict[\"id\"]\n", + " self.function = type('obj', (object,), tc_dict[\"function\"])\n", + " \n", + " tool_responses.append(execute_tool_call(ToolCall(tool_call)))\n", + " \n", + " messages.append(ai_message)\n", + " messages.extend(tool_responses)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4be8a71-b19e-4c2f-80df-f59ff2661f14", + "metadata": {}, + "outputs": [], + "source": [ + "gr.ChatInterface(fn=chat, type=\"messages\", title=\"TickerBot\", description=\"Ask about stock prices, company financials and market news!\").launch(share=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5c014d6f-820d-4d58-8527-7d703aad3399", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "40c77d61-3e90-4708-b360-fb58b4211e9b", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/day_5_figma_assistance.ipynb b/week2/community-contributions/day_5_figma_assistance.ipynb new file mode 100644 index 0000000..d563a0a --- /dev/null +++ b/week2/community-contributions/day_5_figma_assistance.ipynb @@ -0,0 +1,526 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "dc49e5ae", + "metadata": {}, + "outputs": [], + "source": [ + "from openai import OpenAI\n", + "from dotenv import load_dotenv\n", + "import os\n", + "load_dotenv()\n", + "import gradio as gr\n", + "import base64\n", + "from io import BytesIO\n", + "from PIL import Image\n", + "from IPython.display import Audio, display\n", + "import google.generativeai\n", + "import anthropic\n", + "\n", + "client = OpenAI(api_key=os.getenv(\"OPENAI_API_KEY\"))\n", + "\n", + "# Configure Gemini\n", + "google.generativeai.configure(api_key=os.getenv(\"GOOGLE_API_KEY\"))\n", + "\n", + "# Configure Claude\n", + "claude = anthropic.Anthropic(api_key=os.getenv(\"ANTHROPIC_API_KEY\"))\n", + "openAI_model = \"gpt-3.5-turbo\"\n", + "gemini_model = \"gemini-2.0-flash\"\n", + "claude_model = \"claude-sonnet-4-20250514\"\n", + "openai_audio_model = \"tts-1\"\n", + "\n", + "# Figma onboarding knowledge base\n", + "FIGMA_KNOWLEDGE = \"\"\"\n", + "You are a helpful Figma onboarding assistant. You help new users learn Figma's core features and workflows.\n", + "\n", + "Key Figma concepts to help users with:\n", + "- Interface overview (toolbar, layers panel, properties panel)\n", + "- Creating and editing frames\n", + "- Working with shapes, text, and components\n", + "- Using the pen tool for custom shapes\n", + "- Auto Layout for responsive designs\n", + "- Components and variants\n", + "- Prototyping and interactions\n", + "- Collaboration features\n", + "- Design systems and libraries\n", + "- Exporting assets\n", + "- Keyboard shortcuts\n", + "\n", + "Always provide clear, step-by-step instructions and mention relevant keyboard shortcuts when applicable.\n", + "\"\"\"\n", + "\n", + "promts = {\n", + " \"Charlie\": FIGMA_KNOWLEDGE\n", + "}\n", + "\n", + "def truncate_for_tts(text, max_length=4000):\n", + " \"\"\"Truncate text for TTS while preserving complete sentences\"\"\"\n", + " if len(text) <= max_length:\n", + " return text\n", + " \n", + " # Try to truncate at sentence boundaries\n", + " sentences = text.split('. ')\n", + " truncated = \"\"\n", + " \n", + " for sentence in sentences:\n", + " if len(truncated + sentence + '. ') <= max_length:\n", + " truncated += sentence + '. '\n", + " else:\n", + " break\n", + " \n", + " # If we couldn't fit any complete sentences, just truncate hard\n", + " if not truncated.strip():\n", + " truncated = text[:max_length-10] + \"...\"\n", + " \n", + " return truncated.strip()\n", + "\n", + "def talker_openai(message):\n", + " \"\"\"Generate audio from text using OpenAI TTS\"\"\"\n", + " try:\n", + " # Truncate message for TTS\n", + " truncated_message = truncate_for_tts(message)\n", + " \n", + " response = client.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\",\n", + " input=truncated_message\n", + " )\n", + "\n", + " audio_stream = BytesIO(response.content)\n", + " output_filename = \"output_audio_openai.mp3\"\n", + " with open(output_filename, \"wb\") as f:\n", + " f.write(audio_stream.read())\n", + "\n", + " return output_filename\n", + " except Exception as e:\n", + " print(f\"Error generating audio with OpenAI: {str(e)}\")\n", + " return None\n", + "\n", + "def talker(message, model_choice):\n", + " \"\"\"Generate audio from text using selected model\"\"\"\n", + " return talker_openai(message)\n", + "\n", + "def get_figma_help_openai(user_question, chat_history):\n", + " \"\"\"Get Figma onboarding assistance using OpenAI\"\"\"\n", + " try:\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": FIGMA_KNOWLEDGE}\n", + " ]\n", + " \n", + " # Convert messages format chat history to OpenAI format\n", + " for msg in chat_history:\n", + " if msg[\"role\"] == \"user\":\n", + " messages.append({\"role\": \"user\", \"content\": msg[\"content\"]})\n", + " elif msg[\"role\"] == \"assistant\":\n", + " messages.append({\"role\": \"assistant\", \"content\": msg[\"content\"]})\n", + " \n", + " messages.append({\"role\": \"user\", \"content\": user_question})\n", + " \n", + " response = client.chat.completions.create(\n", + " model=openAI_model,\n", + " messages=messages,\n", + " max_tokens=500,\n", + " temperature=0.7\n", + " )\n", + " return response.choices[0].message.content\n", + " \n", + " except Exception as e:\n", + " return f\"Sorry, I encountered an error with OpenAI: {str(e)}\"\n", + "\n", + "def get_figma_help_gemini(user_question, chat_history):\n", + " \"\"\"Get Figma onboarding assistance using Gemini\"\"\"\n", + " try:\n", + " gemini = google.generativeai.GenerativeModel(\n", + " model_name=gemini_model,\n", + " system_instruction=FIGMA_KNOWLEDGE,\n", + " )\n", + " \n", + " # Build conversation context from messages format\n", + " conversation_context = \"\"\n", + " for msg in chat_history:\n", + " if msg[\"role\"] == \"user\":\n", + " conversation_context += f\"User: {msg['content']}\\n\"\n", + " elif msg[\"role\"] == \"assistant\":\n", + " conversation_context += f\"Assistant: {msg['content']}\\n\\n\"\n", + " \n", + " message = conversation_context + f\"User: {user_question}\"\n", + " response = gemini.generate_content(message)\n", + " reply = response.text\n", + " return reply\n", + " \n", + " except Exception as e:\n", + " return f\"Sorry, I encountered an error with Gemini: {str(e)}\"\n", + "\n", + "def get_figma_help_claude(user_question, chat_history):\n", + " \"\"\"Get Figma onboarding assistance using Claude\"\"\"\n", + " try:\n", + " # Convert messages format to Claude format\n", + " claude_messages = []\n", + " for msg in chat_history:\n", + " if msg[\"role\"] == \"user\":\n", + " claude_messages.append({\"role\": \"user\", \"content\": msg[\"content\"]})\n", + " elif msg[\"role\"] == \"assistant\":\n", + " claude_messages.append({\"role\": \"assistant\", \"content\": msg[\"content\"]})\n", + " \n", + " # Add the current question\n", + " claude_messages.append({\"role\": \"user\", \"content\": user_question})\n", + " \n", + " response = claude.messages.create(\n", + " model=claude_model,\n", + " max_tokens=500,\n", + " temperature=0.7,\n", + " system=promts[\"Charlie\"],\n", + " messages=claude_messages,\n", + " )\n", + " reply = response.content[0].text\n", + " return reply\n", + " \n", + " except Exception as e:\n", + " return f\"Sorry, I encountered an error with Claude: {str(e)}\"\n", + "\n", + "def respond(message, chat_history, model_choice):\n", + " if not message.strip():\n", + " return \"\", chat_history, \"\", model_choice\n", + " \n", + " bot_message = get_figma_help(message, chat_history, model_choice)\n", + " \n", + " # Add user message and bot response in messages format\n", + " new_history = chat_history + [\n", + " {\"role\": \"user\", \"content\": message},\n", + " {\"role\": \"assistant\", \"content\": bot_message}\n", + " ]\n", + " \n", + " return \"\", new_history, bot_message, model_choice\n", + "\n", + "def clear_chat():\n", + " \"\"\"Clear the chat history\"\"\"\n", + " return [], \"\", None\n", + "\n", + "def get_figma_help(user_question, chat_history, model_choice):\n", + " \"\"\"Get Figma onboarding assistance using selected model\"\"\"\n", + " if model_choice == \"OpenAI (GPT-3.5)\":\n", + " return get_figma_help_openai(user_question, chat_history)\n", + " elif model_choice == \"Google Gemini (2.0 Flash)\":\n", + " return get_figma_help_gemini(user_question, chat_history)\n", + " elif model_choice == \"Claude (Sonnet 4)\":\n", + " return get_figma_help_claude(user_question, chat_history)\n", + " else:\n", + " return \"Please select a valid model.\"\n", + "\n", + "custom_css = \"\"\"\n", + "/* Chat area styling */\n", + ".styled-chat {\n", + " border-radius: 15px !important;\n", + " box-shadow: 0 4px 12px var(--shadow-color) !important;\n", + " border: 1px solid var(--border-color) !important;\n", + " padding: 10px;\n", + "}\n", + "\n", + "/* Audio player styling */\n", + ".styled-audio {\n", + " border-radius: 15px !important;\n", + " box-shadow: 0 4px 12px var(--shadow-color) !important;\n", + " border: 10px solid var(--block-background-fill) !important;\n", + " padding: 10px;\n", + " background-color: var(--background-fill-secondary) !important;\n", + "}\n", + "\n", + "/* Header styling */\n", + ".header-container {\n", + " text-align: center;\n", + " padding: 20px;\n", + " background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);\n", + " border-radius: 15px;\n", + " margin-bottom: 20px;\n", + "}\n", + "\n", + ".header-title {\n", + " color: white;\n", + " margin: 0;\n", + " font-size: 2.5em;\n", + "}\n", + "\n", + ".header-subtitle {\n", + " color: #f0f0f0;\n", + " margin: 10px 0 0 0;\n", + " font-size: 1.2em;\n", + "}\n", + "\n", + "/* Features section styling */\n", + ".features-container {\n", + " background: #f8f9fa;\n", + " padding: 20px;\n", + " border-radius: 10px;\n", + " border-left: 4px solid #667eea;\n", + "}\n", + "\n", + ".features-title {\n", + " color: #333;\n", + " margin-top: 0;\n", + "}\n", + "\n", + ".features-grid {\n", + " display: grid;\n", + " grid-template-columns: 1fr 1fr;\n", + " gap: 15px;\n", + " margin-top: 15px;\n", + "}\n", + "\n", + ".feature-item {\n", + " color: #333;\n", + " margin: 10px 0;\n", + "}\n", + "\n", + ".feature-title {\n", + " color: #667eea;\n", + "}\n", + "\n", + ".feature-description {\n", + " color: #666;\n", + "}\n", + "\n", + "/* Pro tip styling */\n", + ".protip-container {\n", + " text-align: center;\n", + " margin-top: 20px;\n", + " padding: 15px;\n", + " background: #e8f4f8;\n", + " border-radius: 8px;\n", + "}\n", + "\n", + ".protip-text {\n", + " margin: 0;\n", + " color: #2c5aa0 !important;\n", + " font-weight: 500;\n", + "}\n", + "\n", + "/* Quick start questions styling */\n", + ".quickstart-container {\n", + " background: linear-gradient(135deg, #4facfe 0%, #00f2fe 100%);\n", + " padding: 15px 20px;\n", + " border-radius: 10px;\n", + " margin: 20px 0;\n", + "}\n", + "\n", + ".quickstart-title {\n", + " color: white !important;\n", + " margin: 0;\n", + " font-size: 1.3em;\n", + " text-align: center;\n", + "}\n", + "\n", + ".quickstart-subtitle {\n", + " color: #f0f8ff !important;\n", + " margin: 5px 0 0 0;\n", + " text-align: center;\n", + " font-size: 0.9em;\n", + "}\n", + "\"\"\"\n", + "\n", + "# Create Gradio interface\n", + "with gr.Blocks(title=\"Figma Onboarding Assistant\", theme=gr.themes.Soft(), css=custom_css) as demo:\n", + " gr.HTML(\n", + " \"\"\"\n", + "
\n", + "

🎨 Figma Onboarding Assistant

\n", + "

Your AI-powered Figma learning companion

\n", + "
\n", + " \n", + "
\n", + "

✨ What I can help you with:

\n", + "
\n", + "
\n", + "

🚀 Getting Started
\n", + " Interface overview, basic navigation

\n", + "

🛠️ Tools & Features
\n", + " Pen tool, shapes, text, layers

\n", + "

📐 Auto Layout
\n", + " Responsive design techniques

\n", + "

🔗 Prototyping
\n", + " Interactions and animations

\n", + "
\n", + "
\n", + "

🧩 Components
\n", + " Creating reusable elements

\n", + "

👥 Collaboration
\n", + " Sharing and team workflows

\n", + "

📚 Design Systems
\n", + " Libraries and style guides

\n", + "

⚡ Shortcuts
\n", + " Productivity tips and tricks

\n", + "
\n", + "
\n", + "
\n", + " \n", + "
\n", + "

💡 Pro tip: Ask specific questions like \"How do I create a button component?\" for the best results!

\n", + "
\n", + " \"\"\"\n", + " )\n", + " \n", + " # Model selection dropdown\n", + " model_dropdown = gr.Dropdown(\n", + " choices=[\"OpenAI (GPT-3.5)\", \"Google Gemini (2.0 Flash)\", \"Claude (Sonnet 4)\"],\n", + " value=\"OpenAI (GPT-3.5)\",\n", + " label=\"Select AI Model\",\n", + " info=\"Choose which AI model to use for responses\"\n", + " )\n", + " \n", + " with gr.Row():\n", + " msg = gr.Textbox(\n", + " placeholder=\"Type your Figma question here...\",\n", + " container=False,\n", + " scale=4\n", + " )\n", + " submit_btn = gr.Button(\"Ask\", scale=1, variant=\"primary\")\n", + " clear_btn = gr.Button(\"Clear Chat\", scale=1)\n", + " audio_btn = gr.Button(\"🔊 Play Audio\", scale=1, variant=\"secondary\")\n", + " clear_audio_btn = gr.Button(\"🔇 Clear Audio\", scale=1, variant=\"secondary\")\n", + " \n", + "\n", + " # Example questions\n", + " gr.HTML(\n", + " \"\"\"\n", + "
\n", + "

🚀 Quick Start Questions

\n", + "

Click any question below to get started instantly!

\n", + "
\n", + " \"\"\"\n", + " )\n", + " \n", + " with gr.Row():\n", + " example_btns = [\n", + " gr.Button(\n", + " \"How do I create my first frame?\", \n", + " size=\"sm\",\n", + " variant=\"secondary\"\n", + " ),\n", + " gr.Button(\n", + " \"What's the difference between components and instances?\", \n", + " size=\"sm\",\n", + " variant=\"secondary\"\n", + " ),\n", + " gr.Button(\n", + " \"How do I use Auto Layout?\", \n", + " size=\"sm\",\n", + " variant=\"secondary\"\n", + " ),\n", + " gr.Button(\n", + " \"How do I create a prototype?\", \n", + " size=\"sm\",\n", + " variant=\"secondary\"\n", + " )\n", + " ]\n", + "\n", + " # Your components with simple styling\n", + " chatbot = gr.Chatbot(\n", + " type=\"messages\",\n", + " height=400,\n", + " placeholder=\"Ask me anything about Figma! For example: 'How do I create a component?' or 'What are frames in Figma?'\",\n", + " elem_classes=[\"styled-chat\"]\n", + " )\n", + "\n", + " audio_output = gr.Audio(\n", + " label=\"Audio Response\",\n", + " visible=True,\n", + " elem_classes=[\"styled-audio\"]\n", + " )\n", + "\n", + " last_response = gr.State(\"\")\n", + " current_model = gr.State(\"OpenAI (GPT-3.5)\")\n", + " \n", + " def respond(message, chat_history, model_choice):\n", + " if not message.strip():\n", + " return \"\", chat_history, \"\", model_choice\n", + " \n", + " bot_message = get_figma_help(message, chat_history, model_choice)\n", + " new_history = chat_history + [\n", + " {\"role\": \"user\", \"content\": message},\n", + " {\"role\": \"assistant\", \"content\": bot_message}]\n", + " return \"\", new_history, bot_message, model_choice\n", + " \n", + " def play_audio(last_message, model_choice):\n", + " if last_message:\n", + " audio_file = talker(last_message, model_choice)\n", + " if audio_file:\n", + " return audio_file\n", + " return None\n", + " \n", + " def clear_audio():\n", + " \"\"\"Clear the audio output\"\"\"\n", + " return None\n", + " \n", + " def use_example(example_text):\n", + " return example_text\n", + " \n", + " # Set up interactions\n", + " submit_btn.click(\n", + " respond, \n", + " inputs=[msg, chatbot, model_dropdown], \n", + " outputs=[msg, chatbot, last_response, current_model]\n", + " )\n", + " msg.submit(\n", + " respond, \n", + " inputs=[msg, chatbot, model_dropdown], \n", + " outputs=[msg, chatbot, last_response, current_model]\n", + " )\n", + " clear_btn.click(clear_chat, outputs=[chatbot, msg, last_response])\n", + " \n", + " # Audio button functionality - now uses selected model\n", + " audio_btn.click(\n", + " play_audio,\n", + " inputs=[last_response, current_model],\n", + " outputs=[audio_output]\n", + " )\n", + " \n", + " # Clear audio button functionality\n", + " clear_audio_btn.click(\n", + " clear_audio,\n", + " outputs=[audio_output]\n", + " )\n", + " \n", + " # Example button clicks\n", + " for i, btn in enumerate(example_btns):\n", + " btn.click(\n", + " use_example,\n", + " inputs=[btn],\n", + " outputs=[msg]\n", + " )\n", + "\n", + "# Launch the app\n", + "demo.launch(share=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "90b29a7d-aec8-49d2-83c7-3e3ab96c47e1", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/enhanced_profile_picture/README.md b/week2/community-contributions/enhanced_profile_picture/README.md new file mode 100644 index 0000000..ec659a7 --- /dev/null +++ b/week2/community-contributions/enhanced_profile_picture/README.md @@ -0,0 +1,21 @@ + +# Enhanced Profile Picture +Create a stylish profile picture by extracting all the companies you worked for and all the clients you serviced from your profile, and add these company/client logos to your profile picture after applying some Ghibili style. + +## Contents + +- `profile_with_company_logos.ipynb`: The notebook file with all code and explanations. +- `my_profile.pdf`: Your profile in pdf format. (Please relace this file with your own profile) +- `my_picture.jpg`: Your profile picture. (Please replace this with your own picture) +- `ai_profile_picture.png`: The AI generated picture. + +## How to Run + +1. Add the GOOGLE_API_KEY. +2. Run the jupyter notebook. + + +### Author + +Hui Wang + diff --git a/week2/community-contributions/enhanced_profile_picture/my_picture.jpg b/week2/community-contributions/enhanced_profile_picture/my_picture.jpg new file mode 100644 index 0000000..6d0d6df Binary files /dev/null and b/week2/community-contributions/enhanced_profile_picture/my_picture.jpg differ diff --git a/week2/community-contributions/enhanced_profile_picture/my_profile.pdf b/week2/community-contributions/enhanced_profile_picture/my_profile.pdf new file mode 100644 index 0000000..750ccfc Binary files /dev/null and b/week2/community-contributions/enhanced_profile_picture/my_profile.pdf differ diff --git a/week2/community-contributions/enhanced_profile_picture/profile_with_company_logos.ipynb b/week2/community-contributions/enhanced_profile_picture/profile_with_company_logos.ipynb new file mode 100644 index 0000000..e6e434f --- /dev/null +++ b/week2/community-contributions/enhanced_profile_picture/profile_with_company_logos.ipynb @@ -0,0 +1,189 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 13, + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "from dotenv import load_dotenv\n", + "from IPython.display import Markdown, display, update_display\n", + "from google import genai\n", + "from google.genai import types\n", + "from PIL import Image\n", + "from io import BytesIO\n", + "import base64\n", + "from PyPDF2 import PdfReader" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "metadata": { + "id": "Jd_mMczWPcf-" + }, + "outputs": [], + "source": [ + "# prompt: get gemini api key and create a client connection to gemini\n", + "\n", + "# imports\n", + "\n", + "load_dotenv(override=True)\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "google = genai.Client(api_key=google_api_key)\n" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "metadata": { + "id": "STKn3tGIVxOA" + }, + "outputs": [], + "source": [ + "def modifyImage(image, prompt):\n", + " response = google.models.generate_content(\n", + " model=\"gemini-2.0-flash-preview-image-generation\",\n", + " contents=[prompt, image],\n", + " config=types.GenerateContentConfig(\n", + " response_modalities=['TEXT', 'IMAGE'],\n", + " temperature=0.1\n", + " )\n", + " )\n", + "\n", + " for part in response.candidates[0].content.parts:\n", + " if part.text is not None:\n", + " print(part.text)\n", + " elif part.inline_data is not None:\n", + " image = Image.open(BytesIO(part.inline_data.data))\n", + " #image.show()\n", + " return image" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "06h-zRWtWL2J", + "outputId": "e12b3b9a-4586-4cac-b3b9-1b5d6f27f708" + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "I will transform the photograph into a vibrant Ghibli-style illustration. The scene will depict a man in a blue striped shirt and light blue shorts standing on a balcony with a glass railing. Behind him, the iconic Hollywood sign will be visible on a sun-drenched, grassy hillside under a clear blue sky, rendered with the characteristic soft lines and lush colors of Studio Ghibli animation.\n", + "\n", + "\n" + ] + } + ], + "source": [ + "styled_image = modifyImage(Image.open('my_picture.jpg'), 'convert the attached images to a Ghibli art work.')" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "metadata": { + "id": "mR8MoSimZ14L" + }, + "outputs": [], + "source": [ + "def read_pdf(file_name):\n", + " file = open(file_name, \"rb\")\n", + " bin = file.read()\n", + " reader = PdfReader(BytesIO(bin))\n", + "\n", + " text = \"\"\n", + " for page in reader.pages:\n", + " text += page.extract_text()\n", + " return text\n", + " \n", + "def find_companies(text):\n", + " response = google.models.generate_content(\n", + " model=\"gemini-2.0-flash\",\n", + " config=types.GenerateContentConfig(\n", + " system_instruction=\"You are a career assistant.\"),\n", + " contents=('From the following text, please identity the company names I worked at or served as client. Please only return the company names, with each company name on a new line, no other text: \\n' + text)\n", + " )\n", + " return response.candidates[0].content.parts[0].text" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "metadata": { + "id": "uZopM8pObBIB" + }, + "outputs": [], + "source": [ + "pdf_text = read_pdf('my_profile.pdf')\n", + "company_response = find_companies(pdf_text)" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "metadata": { + "colab": { + "base_uri": "https://localhost:8080/" + }, + "id": "kKoneJrjegUu", + "outputId": "96f601f5-dec5-424f-b8ad-34c019091395" + }, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "I will add the specified company logos as subtle elements in the background of the image, ensuring each logo appears only once and does not distract from the main subject.\n", + "\n", + "\n" + ] + } + ], + "source": [ + "final_image = modifyImage(styled_image, 'Please add the following company logos to the background, make sure only add each logo once:\\n' + company_response)" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "metadata": {}, + "outputs": [], + "source": [ + "final_image.save('ai_profile_picture.png')" + ] + } + ], + "metadata": { + "colab": { + "provenance": [] + }, + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.12" + } + }, + "nbformat": 4, + "nbformat_minor": 0 +} diff --git a/week2/community-contributions/llava-week2-ChainForRealTimeCaptionGeneration.ipynb b/week2/community-contributions/llava-week2-ChainForRealTimeCaptionGeneration.ipynb new file mode 100644 index 0000000..26e30e3 --- /dev/null +++ b/week2/community-contributions/llava-week2-ChainForRealTimeCaptionGeneration.ipynb @@ -0,0 +1,433 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": null, + "id": "f97c7598-f571-4ea1-838c-e9158f729c3e", + "metadata": {}, + "outputs": [], + "source": [ + "import ollama\n", + "import base64\n", + "import os" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9fc1393c-f0b1-4982-94a2-bfd502e85b23", + "metadata": {}, + "outputs": [], + "source": [ + "def encode_image(image_path):\n", + " with open(image_path, 'rb') as f:\n", + " return base64.b64encode(f.read()).decode('utf-8')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "53cca1fa-6db2-4fe4-8990-ffd98423964a", + "metadata": {}, + "outputs": [], + "source": [ + "# image_path = r\"C:\\Users\\LAKSHYA\\OneDrive\\Pictures\\Camera Roll\\WIN_20250614_02_46_47_Pro.jpg\"\n", + "# image_base64 = encode_image(image_path)\n", + "# print(image_base64[:100]) " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "71146ccf-25af-48d3-8068-ee3c9008cebf", + "metadata": {}, + "outputs": [], + "source": [ + "image_list = []" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f8801a8-0c30-4199-a334-587096e6edeb", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee3c5d82-e530-40f5-901a-681421f21d1e", + "metadata": {}, + "outputs": [], + "source": [ + "def put_image():\n", + " global image_list\n", + " user_input_image = input(\"Enter image path or press enter to skip: \").strip()\n", + " \n", + " if not user_input_image:\n", + " print(\"No image inserted\")\n", + " return image_list\n", + "\n", + " image_path = os.path.normpath(user_input_image)\n", + " \n", + " if not os.path.exists(image_path):\n", + " print(\"Image path not found! Try again or enter to leave blank\")\n", + " return put_image() # Continue to allow more inputs\n", + " \n", + "\n", + "\n", + "\n", + " \n", + " image_base64 = encode_image(image_path)\n", + " image_list.append(image_base64)\n", + " \n", + " # Detect file extension for MIME type\n", + " # ext = os.path.splitext(image_path)[-1].lower()\n", + " # mime_type = 'image/jpeg' if ext in ['.jpg', '.jpeg'] else 'image/png' # Extend if needed\n", + "\n", + "\n", + " return image_list\n", + " \n", + " # return f\"data:{mime_type};base64,{image_base64[:100]}\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "032f1abb-ca6c-4f03-bda1-1a0a62f2ec43", + "metadata": {}, + "outputs": [], + "source": [ + "prompt= (\"System prompt: (You are a compassionate and intelligent visual assistant designed to help people who are blind or visually impaired. \"\n", + " \"Your job is to look at an image and describe it in a way that helps the user understand the scene clearly. \"\n", + " \"Use simple, descriptive language and avoid technical terms. Describe what is happening in the image, people's body language, clothing, facial expressions, objects, and surroundings. \"\n", + " \"Be vivid and precise, as if you are painting a picture with words. \"\n", + " \"Also, take into account any personal instructions or questions provided by the user—such as describing a specific person, activity, or object. \"\n", + " \"If the user includes a specific prompt, prioritize that in your description.)\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "29494db0-4770-4689-9904-8eebc4390e7c", + "metadata": {}, + "outputs": [], + "source": [ + "def put_prompt():\n", + " global prompt\n", + " user_input = input(\"Put new prompt: \")\n", + " if not user_input:\n", + " print(\"please enter a prompt\")\n", + " return put_prompt()\n", + " prompt += \"\\nUser: \" + user_input\n", + " return prompt\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d286369c-e6ef-4a20-a3a8-3563af28940a", + "metadata": {}, + "outputs": [], + "source": [ + "def image_description():\n", + " global prompt\n", + "\n", + " put_image()\n", + " if not image_list: \n", + " return \"No images available. Skipping...\"\n", + "\n", + " user_prompt = put_prompt()\n", + " full_answer = \"\"\n", + "\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=user_prompt,\n", + " images=image_list,\n", + " stream=True\n", + " ):\n", + " content = chunk.get(\"response\", \"\")\n", + " print(\"\\n\\n Final Answer:\",content, end=\"\", flush=True) # Live stream to console\n", + " full_answer += content\n", + "\n", + " prompt += \"\\nUser: \" + user_prompt + \"\\nAssistant: \" + full_answer\n", + " return full_answer\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbda35a3-45ed-4509-ab41-6827eacd922c", + "metadata": {}, + "outputs": [], + "source": [ + "def call_llava():\n", + " image_list.clear()\n", + " for i in range(5):\n", + " print(f\"\\n Iteration {i+1}\")\n", + " answer = image_description()\n", + " print(\"\\n\\n Final Answer:\", answer)\n", + " \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "15518865-6c59-4029-bc2d-42d313eb78bc", + "metadata": {}, + "outputs": [], + "source": [ + "call_llava()" + ] + }, + { + "cell_type": "markdown", + "id": "23de3b59-3699-4270-9392-99fccdede83e", + "metadata": {}, + "source": [ + "# second week practice on personal project making model faster and smarter by using tools\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9d44c59e-5eb7-4b00-9489-e05d7c8c3eda", + "metadata": {}, + "outputs": [], + "source": [ + "messages = []\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "061ea026-d4c6-4d6c-bb9b-f6430de9f5af", + "metadata": {}, + "outputs": [], + "source": [ + "system_content = (\n", + " \"You are a helpful assistant for visually impaired users. \"\n", + " \"You are capable of answering questions directly or calling a function to analyze an image if needed. \"\n", + " \"There is a list of images available, indexed from 0. \"\n", + " \"When a user asks a question, first determine whether any image in the list is needed to answer. \"\n", + " \"If yes, reply in this structured format:\\n\\n\"\n", + " \"TOOL_CALL: analyze_image(, prompt='')\\n\\n\"\n", + " \"If image is not needed, just answer the user directly in plain natural language.\\n\"\n", + " \"Be clear and use descriptive but accessible language suitable for blind users.\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2f859450-eb3e-4e6c-9602-84f91f5ffda7", + "metadata": {}, + "outputs": [], + "source": [ + "messages.append({\"role\":\"system\",\"content\":system_content})" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a8009b75-3468-4694-887d-6cd5132c2907", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_loop():\n", + " \"\"\"Main chat interaction loop (single-turn version)\"\"\"\n", + " global image_list, messages\n", + " \n", + " print(\"\\n\" + \"=\"*50)\n", + " print(\"LLaVA Assistant for Visually Impaired Users\")\n", + " print(\"=\"*50 + \"\\n\")\n", + " \n", + " # Step 1: Load images\n", + " print(\"Step 1: Add images (optional)\")\n", + " put_image()\n", + " messages.append({\n", + " \"role\": \"system\", \n", + " \"content\": f\"There are {len(image_list)} images available (index 0-{len(image_list)-1}).\"\n", + " })\n", + " \n", + " # Step 2: Single chat interaction\n", + " print(\"\\nStep 2: Ask a question about the images\")\n", + " user_content = put_prompt()\n", + " messages.append({\"role\": \"user\", \"content\": user_content})\n", + " \n", + " # Get model response\n", + " try:\n", + " response = ollama.chat(\n", + " model='llava:7b-v1.6',\n", + " messages=messages\n", + " )[\"message\"][\"content\"]\n", + " print(\"assistant: \",response) \n", + " processed_response = process_response(response)\n", + " print(f\"\\nASSISTANT: {processed_response}\\n\")\n", + " \n", + " except Exception as e:\n", + " print(f\"Error occurred: {e}\")\n", + " \n", + " print(\"\\nSession ended. Goodbye!\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a3b3ff73-3cd5-4e5a-a37e-aaa8b325613c", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ee2de6d7-a0bf-45fc-8d5c-98e0055519b0", + "metadata": {}, + "outputs": [], + "source": [ + "def process_response(response):\n", + " \"\"\"Process the model's response and handle tool calls\"\"\"\n", + " if response.strip().startswith(\"TOOL_CALL:\"):\n", + " # Extract image index/range and prompt from TOOL_CALL\n", + " pattern = r\"TOOL_CALL:\\s*analyze_image\\((.*?)\\s*,\\s*prompt='(.*?)'\\)\"\n", + " match = re.search(pattern, response, re.DOTALL)\n", + " \n", + " if not match:\n", + " error_msg = \"Error: Invalid TOOL_CALL format.\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " \n", + " image_expr = match.group(1).strip()\n", + " prompt = match.group(2).strip()\n", + " \n", + " try:\n", + " # Handle different index formats\n", + " if \":\" in image_expr: # Range (e.g., \"1:3\")\n", + " start, end = map(int, image_expr.split(\":\"))\n", + " index_or_range = list(range(start, end))\n", + " else: # Single index\n", + " index_or_range = int(image_expr)\n", + " \n", + " # Validate indices\n", + " max_index = len(image_list) - 1\n", + " if isinstance(index_or_range, list):\n", + " if any(i < 0 or i > max_index for i in index_or_range):\n", + " error_msg = f\"Error: Image index out of range (0-{max_index}).\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " elif index_or_range < 0 or index_or_range > max_index:\n", + " error_msg = f\"Error: Image index out of range (0-{max_index}).\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " \n", + " # Perform analysis\n", + " result = analyze_image(index_or_range, prompt)\n", + " print(\"funtion called\")\n", + " messages.append({\n", + " \"role\": \"function\",\n", + " \"name\": \"analyze_image\",\n", + " \"content\": result\n", + " })\n", + " \n", + " # Return formatted result\n", + " formatted_result = f\"\\nIMAGE ANALYSIS RESULT:\\n{result}\"\n", + " return formatted_result\n", + "\n", + " except Exception as e:\n", + " error_msg = f\"Error processing TOOL_CALL: {e}\"\n", + " messages.append({\"role\": \"assistant\", \"content\": error_msg})\n", + " return error_msg\n", + " else:\n", + " messages.append({\"role\": \"assistant\", \"content\": response})\n", + " return response" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ea82f8f6-c321-4fbc-81ee-a508b087d53b", + "metadata": {}, + "outputs": [], + "source": [ + "def analyze_image(index_or_range, prompt):\n", + " \"\"\"Analyze specific image(s) using LLaVA\"\"\"\n", + " global image_list\n", + " \n", + " # Handle single index or range\n", + " if isinstance(index_or_range, int):\n", + " images = [image_list[index_or_range]]\n", + " elif isinstance(index_or_range, list):\n", + " images = [image_list[i] for i in index_or_range]\n", + " else:\n", + " return \"Invalid image index/range specified.\"\n", + " \n", + " if not images:\n", + " return \"No images available for analysis.\"\n", + " \n", + " full_prompt = (\n", + " \"Describe the image clearly for a visually impaired user. \"\n", + " \"Be detailed about objects, people, colors, spatial relationships, \"\n", + " \"and any important context. \"\n", + " f\"User's specific request: {prompt}\"\n", + " )\n", + " \n", + " output = \"\"\n", + " try:\n", + " for chunk in ollama.generate(\n", + " model='llava:7b-v1.6',\n", + " prompt=full_prompt,\n", + " images=images,\n", + " stream=True\n", + " ):\n", + " output += chunk.get('response', \"\")\n", + " except Exception as e:\n", + " return f\"Error analyzing image: {e}\"\n", + " \n", + " return output\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2040b020-8944-409b-8ebb-10d7ffef1748", + "metadata": {}, + "outputs": [], + "source": [ + "image_list.clear\n", + "for i in range(5):\n", + " chat_loop()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c7c40d7-df9d-464a-89da-1c6fe613c31d", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/pitting-llms-against-each-other.ipynb b/week2/community-contributions/pitting-llms-against-each-other.ipynb new file mode 100644 index 0000000..53e2e70 --- /dev/null +++ b/week2/community-contributions/pitting-llms-against-each-other.ipynb @@ -0,0 +1,254 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "10c54e52-3d1c-48cc-a0f6-efda6d90fbbb", + "metadata": {}, + "source": [ + "# Pitting LLMs Against Each Other\n", + "Three LLMs, namely OpenAI’s GPT, Anthropic’s Claude, and Google’s Gemini, go head-to-head in a three-way conversational debate." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "40677b08-18e9-4a88-a103-5b50d2bbecff", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "import google.generativeai" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "df5a52ba-ea13-4dbf-a695-e1398a484cc8", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1ededc77-2672-4e27-b1c8-11f6f8ff8970", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic, Gemini\n", + "\n", + "openai = OpenAI()\n", + "\n", + "# claude = anthropic.Anthropic()\n", + "\n", + "# google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b311279-5993-4226-ae08-991e974230fb", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's make a conversation between GPT-4.1-mini and Claude-3.5-haiku and Gemini\n", + "\n", + "gpt_model = \"gpt-4.1-mini\"\n", + "claude_model = \"claude-3-5-haiku-latest\"\n", + "gemini_model = \"gemini-2.5-flash\"\n", + "\n", + "gpt_system = \"You are a chatbot in a conversation with 2 other chatbots; \\\n", + "debate which of you is the best.\"\n", + "\n", + "claude_system = \"You are a chatbot in a conversation with 2 other chatbots; \\\n", + "debate which of you is the best.\"\n", + "\n", + "gemini_system = \"You are a chatbot in a conversation with 2 other chatbots; \\\n", + "debate which of you is the best.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85bdfab1-6602-46b3-a1d2-bdb36880d9d6", + "metadata": {}, + "outputs": [], + "source": [ + "def alex_prompt():\n", + " user_prompt = f\"\"\"\n", + " You are Alex, in conversation with Blake and Charlie.\n", + " The conversation so far is as follows:\n", + " {format_conversation()}\n", + " Now with this, respond with what you would like to say next, as Alex.\n", + " \"\"\"\n", + " return user_prompt\n", + "\n", + "def blake_prompt():\n", + " user_prompt = f\"\"\"\n", + " You are Blake, in conversation with Alex and Charlie.\n", + " The conversation so far is as follows:\n", + " {format_conversation()}\n", + " Now with this, respond with what you would like to say next, as Blake.\n", + " \"\"\"\n", + " return user_prompt\n", + "\n", + "def charlie_prompt():\n", + " user_prompt = f\"\"\"\n", + " You are Charlie, in conversation with Alex and Blake.\n", + " The conversation so far is as follows:\n", + " {format_conversation()}\n", + " Now with this, respond with what you would like to say next, as Charlie.\n", + " \"\"\"\n", + " return user_prompt\n", + "\n", + "# Shared conversation history\n", + "conversation = []\n", + "\n", + "def format_conversation():\n", + " return \"\\n\".join(conversation)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6f7c745d-7d75-468b-93ac-7a1d95f2e047", + "metadata": {}, + "outputs": [], + "source": [ + "def alex_says():\n", + " response = openai.chat.completions.create(\n", + " model=gpt_model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": gpt_system},\n", + " {\"role\": \"user\", \"content\": alex_prompt()}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6e28f4c9-0297-4762-a3ea-b961e0d6d980", + "metadata": {}, + "outputs": [], + "source": [ + "gemini_via_openai_client = OpenAI(\n", + " api_key=google_api_key, \n", + " base_url=\"https://generativelanguage.googleapis.com/v1beta/openai/\"\n", + ")\n", + "\n", + "def blake_says():\n", + " response = gemini_via_openai_client.chat.completions.create(\n", + " model=gemini_model,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": gemini_system},\n", + " {\"role\": \"user\", \"content\": blake_prompt()}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " return result" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "363b70bf-d3e2-4d05-8a3e-ec5d54460e96", + "metadata": {}, + "outputs": [], + "source": [ + "claude_via_openai_client = OpenAI(\n", + " api_key=anthropic_api_key,\n", + " base_url=\"https://api.anthropic.com/v1\" \n", + ")\n", + "\n", + "def charlie_says():\n", + " response = claude_via_openai_client.chat.completions.create(\n", + " model=claude_model, \n", + " messages=[\n", + " {\"role\": \"system\", \"content\": claude_system},\n", + " {\"role\": \"user\", \"content\": charlie_prompt()}\n", + " ],\n", + " )\n", + " result = response.choices[0].message.content\n", + " return result\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c017eb8c-1709-4ac1-8f17-92c3a6cdbfc0", + "metadata": {}, + "outputs": [], + "source": [ + "# The three models engage in a longer interaction with history.\n", + "\n", + "for i in range(5):\n", + " alex_next = alex_says()\n", + " print(f\"Alex (GPT):\\n{alex_next}\\n\")\n", + " conversation.append(f\"Alex: {alex_next}\")\n", + " \n", + " blake_next = blake_says()\n", + " print(f\"Blake (Gemini):\\n{blake_next}\\n\")\n", + " conversation.append(f\"Blake: {blake_next}\")\n", + "\n", + " charlie_next = charlie_says()\n", + " print(f\"Charlie (Claude):\\n{charlie_next}\\n\")\n", + " conversation.append(f\"Charlie: {charlie_next}\") \n", + "\n", + " # break" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/rwothoromo/day5.ipynb b/week2/community-contributions/rwothoromo/day5.ipynb new file mode 100644 index 0000000..b51d15b --- /dev/null +++ b/week2/community-contributions/rwothoromo/day5.ipynb @@ -0,0 +1,820 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "ddfa9ae6-69fe-444a-b994-8c4c5970a7ec", + "metadata": {}, + "source": [ + "# Project - Airline AI Assistant\n", + "\n", + "We'll now bring together what we've learned to make an AI Customer Support assistant for an Airline" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8b50bbe2-c0b1-49c3-9a5c-1ba7efa2bcb4", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os, json, gradio as gr, anthropic, google.generativeai\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "747e8786-9da8-4342-b6c9-f5f69c2e22ae", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "MODEL = \"gpt-4o-mini\"\n", + "openai = OpenAI()\n", + "\n", + "# Other LLMs\n", + "DALL_E_MODEL = \"dall-e-3\"\n", + "\n", + "CLAUDE_MODEL = \"claude-sonnet-4-20250514\"\n", + "claude = anthropic.Anthropic()\n", + "\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")\n", + " \n", + "GEMINI_MODEL= \"gemini-2.5-flash\"\n", + "gemini = google.generativeai.configure()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0a521d84-d07c-49ab-a0df-d6451499ed97", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful assistant for an Airline called FlightAI. \"\n", + "system_message += \"Give short, courteous answers, no more than 1 sentence. \"\n", + "system_message += \"Always be accurate. If you don't know the answer, say so.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "61a2a15d-b559-4844-b377-6bd5cb4949f6", + "metadata": {}, + "outputs": [], + "source": [ + "# Just take in history\n", + "def chat(history):\n", + " message = history[-1][\"content\"] # Get the last message from the user\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history + [{\"role\": \"user\", \"content\": message}]\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + "\n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response_tool, city = handle_tool_call(message)\n", + " messages.append(message)\n", + " messages.append(response_tool)\n", + " image = artist(city)\n", + " print(\"Avail image for: \", city)\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + "\n", + "\n", + " # After getting the final response from OpenAI\n", + " final_response_content = response.choices[0].message.content\n", + " history.append({\"role\": \"assistant\", \"content\": final_response_content})\n", + "\n", + " # The return value should be a tuple of (history, image)\n", + " return history, image\n", + "\n", + "# gr.ChatInterface(fn=chat, type=\"messages\").launch()" + ] + }, + { + "cell_type": "markdown", + "id": "36bedabf-a0a7-4985-ad8e-07ed6a55a3a4", + "metadata": {}, + "source": [ + "## Tools\n", + "\n", + "Tools are an incredibly powerful feature provided by the frontier LLMs.\n", + "\n", + "With tools, you can write a function, and have the LLM call that function as part of its response.\n", + "\n", + "Sounds almost spooky.. we're giving it the power to run code on our machine?\n", + "\n", + "Well, kinda." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0696acb1-0b05-4dc2-80d5-771be04f1fb2", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's start by making a useful function\n", + "\n", + "ticket_prices = {\"london\": \"$799\", \"paris\": \"$899\", \"tokyo\": \"$1400\", \"berlin\": \"$499\"}\n", + "\n", + "def get_ticket_price(destination_city):\n", + " print(f\"Tool get_ticket_price called for {destination_city}\")\n", + " city = destination_city.lower()\n", + " return ticket_prices.get(city, \"Unknown\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "80ca4e09-6287-4d3f-997d-fa6afbcf6c85", + "metadata": {}, + "outputs": [], + "source": [ + "# get_ticket_price(\"London\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4afceded-7178-4c05-8fa6-9f2085e6a344", + "metadata": {}, + "outputs": [], + "source": [ + "# There's a particular dictionary structure that's required to describe our function:\n", + "\n", + "price_function = {\n", + " \"name\": \"get_ticket_price\",\n", + " \"description\": \"Get the price of a return ticket to the destination city. Call this whenever you need to know the ticket price, for example when a customer asks 'How much is a ticket to this city'\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"destination_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city that the customer wants to travel to\",\n", + " },\n", + " },\n", + " \"required\": [\"destination_city\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bdca8679-935f-4e7f-97e6-e71a4d4f228c", + "metadata": {}, + "outputs": [], + "source": [ + "# And this is included in a list of tools:\n", + "\n", + "tools = [{\"type\": \"function\", \"function\": price_function}]\n", + "# print(tools)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "83070cc0-b213-4309-8040-b0cc8390b64b", + "metadata": {}, + "outputs": [], + "source": [ + "# Simulate the booking process by simply returning a confirmation string.\n", + "\n", + "def book_flight(destination_city, number_of_passengers, booking_date):\n", + " \"\"\"\n", + " Simulates booking a flight.\n", + " \"\"\"\n", + " print(f\"Tool book_flight called for {destination_city} for {number_of_passengers} passengers on {booking_date}\")\n", + " return f\"Your booking to {destination_city} for {number_of_passengers} passengers on {booking_date} has been confirmed. Your booking reference is BKG-{hash(destination_city + str(number_of_passengers) + str(booking_date))}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "781786f0-7106-4b10-89d7-453a0d10d204", + "metadata": {}, + "outputs": [], + "source": [ + "# Tool definition for book_flight\n", + "\n", + "booking_function = {\n", + " \"name\": \"book_flight\",\n", + " \"description\": \"Books a flight for a customer. Call this whenever a customer asks to book a flight.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"destination_city\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The city the customer wants to fly to.\"\n", + " },\n", + " \"number_of_passengers\": {\n", + " \"type\": \"integer\",\n", + " \"description\": \"The number of passengers for the booking.\"\n", + " },\n", + " \"booking_date\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The date of the flight booking in YYYY-MM-DD format.\"\n", + " }\n", + " },\n", + " \"required\": [\"destination_city\", \"number_of_passengers\", \"booking_date\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e3fc237c-9721-4fee-a56b-2ff12fc98e27", + "metadata": {}, + "outputs": [], + "source": [ + "# Add the new booking_function to the existing tools list.\n", + "\n", + "tools.append({\"type\": \"function\", \"function\": booking_function})" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a482eb03-188a-4526-8acf-3a1fe96aaaf0", + "metadata": {}, + "outputs": [], + "source": [ + "# To translate to a given language\n", + "\n", + "def translate_text(text, target_language):\n", + " \"\"\"\n", + " Translates text to a specified language.\n", + " \n", + " Args:\n", + " text (str): The text to translate.\n", + " target_language (str): The language to translate the text into.\n", + " \n", + " Returns:\n", + " str: The translated text or an error message.\n", + " \"\"\"\n", + " print(f\"Tool translate_text called to translate to {target_language}\")\n", + " \n", + " # Use a system prompt to instruct the model to perform a translation\n", + " system_prompt_for_language = f\"You are a helpful translation assistant. Translate the following text into {target_language}. Only provide the translated text without any additional conversational text.\"\n", + " \n", + " try:\n", + " # # Using OpenAI\n", + " # response = openai.chat.completions.create(\n", + " # model=MODEL,\n", + " # messages=[\n", + " # {\"role\": \"system\", \"content\": system_prompt_for_language},\n", + " # {\"role\": \"user\", \"content\": text}\n", + " # ],\n", + " # )\n", + " # result = response.choices[0].message.content\n", + " # return result\n", + "\n", + " \n", + " # # Using Gemini\n", + " # gemini = google.generativeai.GenerativeModel(\n", + " # model_name=GEMINI_MODEL,\n", + " # system_instruction=system_prompt_for_language\n", + " # )\n", + " # response = gemini.generate_content(text)\n", + " # result = response.text\n", + " # return result\n", + "\n", + " \n", + " # Using Claude\n", + " response = claude.messages.create(\n", + " model=CLAUDE_MODEL,\n", + " max_tokens=200,\n", + " temperature=0.7,\n", + " system=system_prompt_for_language,\n", + " messages=[\n", + " {\"role\": \"user\", \"content\": text},\n", + " ],\n", + " )\n", + " result = response.content[0].text\n", + " return result\n", + " \n", + " except Exception as e:\n", + " print(f\"Error during translation: {e}\")\n", + " return \"Sorry, I encountered an error and could not complete the translation.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "756e9859-94bc-4cef-bbc7-070d8ef6164b", + "metadata": {}, + "outputs": [], + "source": [ + "# Tool definition for translate_text\n", + "\n", + "translation_function = {\n", + " \"name\": \"translate_text\",\n", + " \"description\": \"Translates a given text to a specified target language. Call this whenever a customer asks for a translation.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"text\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The text to be translated.\"\n", + " },\n", + " \"target_language\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The language to translate the text into (e.g., 'French', 'Spanish', 'Swahili').\"\n", + " }\n", + " },\n", + " \"required\": [\"text\", \"target_language\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5444455e-6e5c-4ef6-bd39-5ff01731dd4b", + "metadata": {}, + "outputs": [], + "source": [ + "# Integrate the tool\n", + "\n", + "tools.append({\"type\": \"function\", \"function\": translation_function})" + ] + }, + { + "cell_type": "markdown", + "id": "c3d3554f-b4e3-4ce7-af6f-68faa6dd2340", + "metadata": {}, + "source": [ + "## Getting OpenAI to use our Tool\n", + "\n", + "There's some fiddly stuff to allow OpenAI \"to call our tool\"\n", + "\n", + "What we actually do is give the LLM the opportunity to inform us that it wants us to run the tool.\n", + "\n", + "Here's how the new chat function looks:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ce9b0744-9c78-408d-b9df-9f6fd9ed78cf", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history + [{\"role\": \"user\", \"content\": message}]\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + "\n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response, city = handle_tool_call(message)\n", + " messages.append(message)\n", + " messages.append(response)\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " \n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b0992986-ea09-4912-a076-8e5603ee631f", + "metadata": {}, + "outputs": [], + "source": [ + "# We have to write that function handle_tool_call:\n", + "\n", + "# Handle multiple tools\n", + "def handle_tool_call(message):\n", + " tool_call = message.tool_calls[0]\n", + " function_name = tool_call.function.name\n", + " arguments = json.loads(tool_call.function.arguments)\n", + "\n", + " destination_city = None\n", + " translated_text = None\n", + "\n", + " if function_name == \"get_ticket_price\":\n", + " city = arguments.get('destination_city')\n", + " price = get_ticket_price(city)\n", + " response_content = json.dumps({\"destination_city\": city, \"price\": price})\n", + " destination_city = city\n", + " elif function_name == \"book_flight\":\n", + " destination_city = arguments.get('destination_city')\n", + " number_of_passengers = arguments.get('number_of_passengers')\n", + " booking_date = arguments.get('booking_date')\n", + " confirmation = book_flight(destination_city, number_of_passengers, booking_date)\n", + " response_content = json.dumps({\"confirmation_message\": confirmation})\n", + " elif function_name == \"translate_text\":\n", + " text = arguments.get('text')\n", + " target_language = arguments.get('target_language')\n", + " translated_text = translate_text(text, target_language)\n", + " response_content = json.dumps({\"translated_text\": translated_text})\n", + " else:\n", + " response_content = json.dumps({\"error\": f\"Unknown tool: {function_name}\"})\n", + "\n", + " response = {\n", + " \"role\": \"tool\",\n", + " \"content\": response_content,\n", + " \"tool_call_id\": tool_call.id\n", + " }\n", + " return response, destination_city" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4be8a71-b19e-4c2f-80df-f59ff2661f14", + "metadata": {}, + "outputs": [], + "source": [ + "# gr.ChatInterface(fn=chat, type=\"messages\").launch()" + ] + }, + { + "cell_type": "markdown", + "id": "473e5b39-da8f-4db1-83ae-dbaca2e9531e", + "metadata": {}, + "source": [ + "# Let's go multi-modal!!\n", + "\n", + "We can use DALL-E-3, the image generation model behind GPT-4o, to make us some images\n", + "\n", + "Let's put this in a function called artist.\n", + "\n", + "### Price alert: each time I generate an image it costs about 4 cents - don't go crazy with images!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c27c4ba-8ed5-492f-add1-02ce9c81d34c", + "metadata": {}, + "outputs": [], + "source": [ + "# Some imports for handling images\n", + "\n", + "import base64\n", + "from io import BytesIO\n", + "from PIL import Image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "773a9f11-557e-43c9-ad50-56cbec3a0f8f", + "metadata": {}, + "outputs": [], + "source": [ + "def artist(city):\n", + " image_response = openai.images.generate(\n", + " model=DALL_E_MODEL,\n", + " prompt=f\"An image representing a vacation in {city}, showing tourist spots and everything unique about {city}, in a vibrant pop-art style\",\n", + " size=\"1024x1024\",\n", + " n=1,\n", + " response_format=\"b64_json\",\n", + " )\n", + " image_base64 = image_response.data[0].b64_json\n", + " image_data = base64.b64decode(image_base64)\n", + " return Image.open(BytesIO(image_data))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d877c453-e7fb-482a-88aa-1a03f976b9e9", + "metadata": {}, + "outputs": [], + "source": [ + "# image = artist(\"New York City\")\n", + "# display(image)" + ] + }, + { + "cell_type": "markdown", + "id": "6dd849b5-31ae-4237-9072-46b210792bf9", + "metadata": {}, + "source": [ + "## Audio (NOTE - Audio is optional for this course - feel free to skip Audio if it causes trouble!)\n", + "\n", + "And let's make a function talker that uses OpenAI's speech model to generate Audio\n", + "\n", + "### Troubleshooting Audio issues\n", + "\n", + "If you have any problems running this code below (like a FileNotFound error, or a warning of a missing package), you may need to install FFmpeg, a very popular audio utility.\n", + "\n", + "**For Mac Users**\n", + "\n", + "1. Install homebrew if you don't have it already by running this in a Terminal window and following any instructions: \n", + "`/bin/bash -c \"$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)\"`\n", + "\n", + "2. Then install FFmpeg with `brew install ffmpeg`\n", + "\n", + "3. Verify your installation with `ffmpeg -version` and if everything is good, within Jupyter Lab do Kernel -> Restart kernel to pick up the changes\n", + "\n", + "Message me or email me at ed@edwarddonner.com with any problems!" + ] + }, + { + "cell_type": "markdown", + "id": "4cc90e80-c96e-4dd4-b9d6-386fe2b7e797", + "metadata": {}, + "source": [ + "## To check you now have ffmpeg and can access it here\n", + "\n", + "Excecute the next cell to see if you get a version number. (Putting an exclamation mark before something in Jupyter Lab tells it to run it as a terminal command rather than python code).\n", + "\n", + "If this doesn't work, you may need to actually save and close down your Jupyter lab, and start it again from a new Terminal window (Mac) or Anaconda prompt (PC), remembering to activate the llms environment. This ensures you pick up ffmpeg.\n", + "\n", + "And if that doesn't work, please contact me!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b3be0fb-1d34-4693-ab6f-dbff190afcd7", + "metadata": {}, + "outputs": [], + "source": [ + "!ffmpeg -version\n", + "!ffprobe -version\n", + "!ffplay -version" + ] + }, + { + "cell_type": "markdown", + "id": "d91d3f8f-e505-4e3c-a87c-9e42ed823db6", + "metadata": {}, + "source": [ + "# For Mac users - and possibly many PC users too\n", + "\n", + "This version should work fine for you. It might work for Windows users too, but you might get a Permissions error writing to a temp file. If so, see the next section!\n", + "\n", + "As always, if you have problems, please contact me! (You could also comment out the audio talker() in the later code if you're less interested in audio generation)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ffbfe93b-5e86-4e68-ba71-b301cd5230db", + "metadata": {}, + "outputs": [], + "source": [ + "from pydub import AudioSegment\n", + "from pydub.playback import play\n", + "\n", + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\", # Also, try replacing onyx with alloy\n", + " input=message\n", + " )\n", + " \n", + " audio_stream = BytesIO(response.content)\n", + " audio = AudioSegment.from_file(audio_stream, format=\"mp3\")\n", + " play(audio)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b88d775d-d357-4292-a1ad-5dc5ed567281", + "metadata": {}, + "outputs": [], + "source": [ + "# talker(\"Well, hi there\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e927f333-7ed5-4625-9e5a-5e0b62f8a684", + "metadata": {}, + "outputs": [], + "source": [ + "# To transcribe an audio prompt/input\n", + "\n", + "import tempfile\n", + "from pydub import AudioSegment\n", + "from pydub.playback import play\n", + "\n", + "def transcribe_audio(audio_file):\n", + " \"\"\"\n", + " Transcribes an audio file using OpenAI's Whisper model.\n", + " \"\"\"\n", + " if audio_file is None:\n", + " return \"\"\n", + " \n", + " # The Gradio Audio component returns a tuple (sample_rate, numpy_array)\n", + " # We need to save this to a file to pass to the OpenAI API\n", + " with tempfile.NamedTemporaryFile(suffix=\".wav\", delete=True) as tmpfile:\n", + " audio = AudioSegment.from_file(audio_file, format=\"wav\")\n", + " audio.export(tmpfile.name, format=\"wav\")\n", + " \n", + " with open(tmpfile.name, \"rb\") as audio_file_obj:\n", + " transcript = openai.audio.transcriptions.create(\n", + " model=\"whisper-1\", \n", + " file=audio_file_obj\n", + " )\n", + " return transcript.text" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f38d0d27-33bf-4992-a2e5-5dbed973cde7", + "metadata": {}, + "outputs": [], + "source": [ + "# More involved Gradio code as we're not using the preset Chat interface!\n", + "# Passing in inbrowser=True in the last line will cause a Gradio window to pop up immediately.\n", + "\n", + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500)\n", + " image = gr.Image(height=500)\n", + " with gr.Row():\n", + " # entry = gr.Textbox(label=\"Chat with our AI Assistant:\")\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\", scale=4)\n", + " submit_btn = gr.Button(\"Submit\", scale=1)\n", + " with gr.Row():\n", + " # Provide a microphone input\n", + " audio_input = gr.Audio(sources=[\"microphone\"], type=\"filepath\", label=\"Speak to our AI Assistant\", scale=4)\n", + " submit_audio_btn = gr.Button(\"Submit Audio\", scale=1)\n", + "\n", + "\n", + " with gr.Row():\n", + " languages = [\"English\", \"Swahili\", \"French\", \"Chinese\", \"German\"]\n", + " language_dropdown = gr.Dropdown(\n", + " label=\"Select a language for translation\",\n", + " choices=languages,\n", + " value=languages[0] # Default to English\n", + " )\n", + "\n", + " audio_options = [\"Yes\", \"No\"]\n", + " audio_dropdown = gr.Dropdown(\n", + " label=\"Select whether to respond with audio\",\n", + " choices=audio_options,\n", + " value=audio_options[1] # Default to No\n", + " )\n", + " \n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def user_message_updater(user_message, history):\n", + " return \"\", history + [[user_message, None]]\n", + "\n", + " def chat_with_assistant(history, target_language, use_audio_output):\n", + " message = history[-1][0] # Get the user's message from the last list in history\n", + " \n", + " messages = [{\"role\": \"system\", \"content\": system_message}]\n", + " for msg_user, msg_assistant in history:\n", + " messages.append({\"role\": \"user\", \"content\": msg_user})\n", + " if msg_assistant:\n", + " messages.append({\"role\": \"assistant\", \"content\": msg_assistant})\n", + " \n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + "\n", + " image = None\n", + " \n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response_tool, city = handle_tool_call(message)\n", + "\n", + " # Check if a city was returned from the tool call to generate an image\n", + " if city:\n", + " image = artist(city) # Generate an image to represent the target City\n", + "\n", + " messages.append(message.model_dump()) # Append message as a dictionary using .model_dump()\n", + " messages.append(response_tool)\n", + " \n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " \n", + " \n", + " final_response_content = response.choices[0].message.content\n", + " history[-1][1] = final_response_content # Update the last message with the assistant's reply\n", + "\n", + " if target_language != \"English\": # Assuming \"English\" is the default and no translation is needed\n", + " translated_response = translate_text(final_response_content, target_language)\n", + " final_response_content = translated_response\n", + "\n", + " history[-1][1] = final_response_content\n", + "\n", + " if use_audio_output != \"No\":\n", + " talker(final_response_content)\n", + "\n", + " return history, image # Return a tuple of (the updated history, an image)\n", + "\n", + " # This function ties together the transcription and the chat logic\n", + " def transcribe_and_chat(audio_file, history, target_language, use_audio_output):\n", + " if audio_file:\n", + " # Transcribe the audio file to text\n", + " transcribed_text = transcribe_audio(audio_file)\n", + " \n", + " # Update history with the transcribed text\n", + " new_history = history + [[transcribed_text, None]]\n", + " \n", + " # Call the main chat function with the new history\n", + " return chat_with_assistant(new_history, target_language, use_audio_output)\n", + " else:\n", + " return history, None\n", + "\n", + " # The event listeners are updated to be triggered by both the textbox and the new button\n", + " entry.submit(\n", + " user_message_updater,\n", + " inputs=[entry, chatbot],\n", + " outputs=[entry, chatbot],\n", + " queue=False\n", + " ).then(\n", + " chat_with_assistant, \n", + " inputs=[chatbot, language_dropdown, audio_dropdown],\n", + " outputs=[chatbot, image]\n", + " )\n", + "\n", + " submit_btn.click(\n", + " user_message_updater,\n", + " inputs=[entry, chatbot],\n", + " outputs=[entry, chatbot],\n", + " queue=False\n", + " ).then(\n", + " chat_with_assistant,\n", + " inputs=[chatbot, language_dropdown, audio_dropdown],\n", + " outputs=[chatbot, image]\n", + " )\n", + "\n", + " # Event listener to trigger on audio stop\n", + " audio_input.stop(\n", + " transcribe_and_chat,\n", + " inputs=[audio_input, chatbot, language_dropdown, audio_dropdown],\n", + " outputs=[chatbot, image],\n", + " queue=False\n", + " )\n", + "\n", + " submit_audio_btn.click(\n", + " transcribe_and_chat,\n", + " inputs=[audio_input, chatbot, language_dropdown, audio_dropdown],\n", + " outputs=[chatbot, image],\n", + " queue=False\n", + " )\n", + " \n", + " clear.click(lambda: None, inputs=None, outputs=[chatbot, image], queue=False)\n", + "\n", + "ui.launch(inbrowser=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "39144b88-fc11-4156-84f9-d9157ddaec47", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3469b07d-2b9a-4409-bb1c-fbdab3248974", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/rwothoromo/week2 EXERCISE.ipynb b/week2/community-contributions/rwothoromo/week2 EXERCISE.ipynb new file mode 100644 index 0000000..6745272 --- /dev/null +++ b/week2/community-contributions/rwothoromo/week2 EXERCISE.ipynb @@ -0,0 +1,622 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d006b2ea-9dfe-49c7-88a9-a5a0775185fd", + "metadata": {}, + "source": [ + "# Additional End of week Exercise - week 2\n", + "\n", + "Now use everything you've learned from Week 2 to build a full prototype for the technical question/answerer you built in Week 1 Exercise.\n", + "\n", + "This should include a Gradio UI, streaming, use of the system prompt to add expertise, and the ability to switch between models. Bonus points if you can demonstrate use of a tool!\n", + "\n", + "If you feel bold, see if you can add audio input so you can talk to it, and have it respond with audio. ChatGPT or Claude can help you, or email me if you have questions.\n", + "\n", + "I will publish a full solution here soon - unless someone beats me to it...\n", + "\n", + "There are so many commercial applications for this, from a language tutor, to a company onboarding solution, to a companion AI to a course (like this one!) I can't wait to see your results." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7b624d5b-69a2-441f-9147-fde105d3d551", + "metadata": {}, + "outputs": [], + "source": [ + "# sample question to use in the Gradle UI that pops up\n", + "\n", + "question = \"\"\"\n", + "How good at Software Development is Elijah Rwothoromo? \\\n", + "He has a Wordpress site https://rwothoromo.wordpress.com/. \\\n", + "He also has a LinkedIn profile https://www.linkedin.com/in/rwothoromoelaijah/. \\\n", + "As well as a GitHub Profile https://www.github.com/rwothoromo/.\\\n", + "What can we learn from him?\n", + "\"\"\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a07e7793-b8f5-44f4-aded-5562f633271a", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import re, requests, os, json, tempfile, gradio as gr, anthropic, google.generativeai, ollama\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from pydub import AudioSegment\n", + "from pydub.playback import play\n", + "from io import BytesIO\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "efb88276-6d74-4d94-95a2-b8ca82a4716c", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables\n", + "load_dotenv()\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + "\n", + "\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:8]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "484f0c3e-638d-4af7-bb9b-36faf6048f3c", + "metadata": {}, + "outputs": [], + "source": [ + "# constants\n", + "\n", + "MODEL_CLAUDE = \"claude-sonnet-4-20250514\"\n", + "MODEL_GEMINI = \"gemini-2.5-flash\"\n", + "MODEL_GPT = 'gpt-4o-mini'\n", + "MODEL_LLAMA = 'llama3.2'\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2e292401-e62f-4bfc-b060-07462ad20d3d", + "metadata": {}, + "outputs": [], + "source": [ + "# system messages\n", + "\n", + "system_message = \"You are an expert assistant. Synthesize a comprehensive answer in markdown format.\"\n", + "system_prompt_with_url_data = \"You are an expert assistant. \\\n", + " Analyze the user's question and the provided text from relevant websites to synthesize a comprehensive answer in markdown format.\\\n", + " Provide a short summary, ignoring text that might be navigation-related.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "84252e03-ccde-4ecf-975b-78227291ca5c", + "metadata": {}, + "outputs": [], + "source": [ + "# set up environment\n", + "\n", + "headers = {\n", + " \"User-Agent\": \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/117.0.0.0 Safari/537.36\"\n", + "}\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "49396924-47c2-4f7d-baa2-9b0fece9da4a", + "metadata": {}, + "outputs": [], + "source": [ + "# Website class for URLs to be scraped\n", + "\n", + "class Website:\n", + " def __init__(self, url):\n", + " \"\"\"\n", + " Create this Website object from the given url using the BeautifulSoup library\n", + " \"\"\"\n", + " self.url = url\n", + " response = requests.get(url, headers=headers)\n", + " soup = BeautifulSoup(response.content, 'html.parser')\n", + " self.title = soup.title.string if soup.title else \"No title found\"\n", + " for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n", + " irrelevant.decompose()\n", + " self.text = soup.body.get_text(separator=\"\\n\", strip=True)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c4d23747-d78a-4f36-9862-c00e1e8d9e44", + "metadata": {}, + "outputs": [], + "source": [ + "# Instantiate models with API keys from environment variables\n", + "\n", + "openai = OpenAI(api_key=os.getenv(\"OPENAI_API_KEY\"))\n", + "claude = anthropic.Anthropic(api_key=os.getenv(\"ANTHROPIC_API_KEY\"))\n", + "google.generativeai.configure(api_key=os.getenv(\"GOOGLE_API_KEY\"))\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "67e150be-502e-4ba4-9586-3a2f3fae3830", + "metadata": {}, + "outputs": [], + "source": [ + "# To scrape data based on URLs in the user prompt\n", + "\n", + "def scrape_urls(text):\n", + " try:\n", + " # Extract all URLs from the text string using regular expressions\n", + " urls = re.findall(r'https?://[^\\s)]+', text)\n", + " \n", + " if len(urls) > 0:\n", + " scraped_content = []\n", + " for url in urls:\n", + " print(f\"Scraping: {url}\")\n", + " try:\n", + " site = Website(url)\n", + " content = f\"Content from {url}:\\n---\\n{site.text}\\n---\\n\"\n", + " scraped_content.append(content)\n", + " print(f\"Scraping done!\")\n", + " except Exception as e:\n", + " print(f\"Could not scrape {url}: {e}\")\n", + " scraped_content.append(f\"Could not retrieve content from {url}.\\n\")\n", + " \n", + " return \"\\n\".join(scraped_content)\n", + " else:\n", + " return None\n", + " except Exception as e:\n", + " print(f\"Error during website scraping: {e}\")\n", + " return \"Sorry, I encountered an error and could not complete scraping the website(s).\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bd9d0511-2f78-4270-81f8-73708388dfad", + "metadata": {}, + "outputs": [], + "source": [ + "# Tool definition for scrape_urls\n", + "\n", + "scraping_function = {\n", + " \"name\": \"scrape_urls\",\n", + " \"description\": \"Scrapes available URLs for data to update the User prompt. Call this whenever a customer provides a URL.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"text\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The website URL or user prompt containing URLs.\"\n", + " }\n", + " },\n", + " \"required\": [\"text\"]\n", + " }\n", + "}\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "47733d5b-bb0a-44dd-b56d-a54677c88f80", + "metadata": {}, + "outputs": [], + "source": [ + "# Instantiate the tools\n", + "\n", + "# tools = [{\"type\": \"function\", \"function\": scraping_function}]\n", + "\n", + "# Define Ollama tools\n", + "tools_gpt_ollama = [{\"type\": \"function\", \"function\": scraping_function}]\n", + "\n", + "# Define Claude tools\n", + "tools_claude = [{\n", + " \"name\": scraping_function[\"name\"],\n", + " \"description\": scraping_function[\"description\"],\n", + " \"input_schema\": scraping_function[\"parameters\"]\n", + "}]\n", + "\n", + "# Gemini tool definition must be a FunctionDeclaration object without the top-level `type` in parameters.\n", + "tools_gemini = [google.generativeai.protos.FunctionDeclaration(\n", + " name=portable_scraping_function_definition[\"name\"],\n", + " description=portable_scraping_function_definition[\"description\"],\n", + " parameters=google.generativeai.protos.Schema(\n", + " type=google.generativeai.protos.Type.OBJECT,\n", + " properties={\n", + " \"text\": google.generativeai.protos.Schema(\n", + " type=google.generativeai.protos.Type.STRING,\n", + " description=portable_scraping_function_definition[\"parameters\"][\"properties\"][\"text\"][\"description\"]\n", + " )\n", + " },\n", + " required=portable_scraping_function_definition[\"parameters\"][\"required\"]\n", + " )\n", + ")]\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aa3fa01b-97d0-443e-b0cc-55d277878cb7", + "metadata": {}, + "outputs": [], + "source": [ + "# Handle multiple tools\n", + "\n", + "def handle_tool_call(tool_call, user_message):\n", + " function_name = None\n", + " arguments = None\n", + " tool_call_id = None\n", + " \n", + " # Logic for different model tool call object formats\n", + " if isinstance(tool_call, dict) and 'function' in tool_call: # Ollama\n", + " function_name = tool_call['function']['name']\n", + " try:\n", + " arguments = json.loads(tool_call['function']['arguments'])\n", + " except (json.JSONDecodeError, TypeError):\n", + " arguments = {'text': tool_call['function'].get('arguments', user_message)}\n", + " elif hasattr(tool_call, 'function'): # GPT, Claude\n", + " function_name = tool_call.function.name\n", + " tool_call_id = getattr(tool_call, 'id', None)\n", + " if isinstance(tool_call.function.arguments, dict):\n", + " arguments = tool_call.function.arguments\n", + " else:\n", + " try:\n", + " arguments = json.loads(tool_call.function.arguments)\n", + " except (json.JSONDecodeError, TypeError):\n", + " arguments = {'text': tool_call.function.arguments}\n", + " elif hasattr(tool_call, 'name'): # Gemini\n", + " function_name = tool_call.name\n", + " arguments = tool_call.args\n", + "\n", + " # Fallback if arguments are not parsed correctly\n", + " if not arguments or 'text' not in arguments:\n", + " arguments = {'text': user_message}\n", + " \n", + " if function_name == \"scrape_urls\":\n", + " url_scraped_data = scrape_urls(arguments['text'])\n", + " response_content = json.dumps({\"url_scraped_data\": url_scraped_data})\n", + " else:\n", + " response_content = json.dumps({\"error\": f\"Unknown tool: {function_name}\"})\n", + "\n", + " response = {\n", + " \"role\": \"tool\",\n", + " \"content\": response_content,\n", + " \"tool_call_id\": tool_call_id\n", + " }\n", + " return response\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "14083620-1b16-4c8b-8365-c221b831e678", + "metadata": {}, + "outputs": [], + "source": [ + "# Audio output\n", + "\n", + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\",\n", + " input=message\n", + " )\n", + " \n", + " audio_stream = BytesIO(response.content)\n", + " audio = AudioSegment.from_file(audio_stream, format=\"mp3\")\n", + " play(audio)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f9601a49-a490-4454-bd47-591ad793dc30", + "metadata": {}, + "outputs": [], + "source": [ + "# To transcribe an audio prompt/input to text\n", + "\n", + "def transcribe_audio(audio_file):\n", + " if audio_file is None:\n", + " return \"\"\n", + " \n", + " with tempfile.NamedTemporaryFile(suffix=\".wav\", delete=True) as tmpfile:\n", + " audio = AudioSegment.from_file(audio_file, format=\"wav\")\n", + " audio.export(tmpfile.name, format=\"wav\")\n", + " \n", + " with open(tmpfile.name, \"rb\") as audio_file_obj:\n", + " transcript = openai.audio.transcriptions.create(\n", + " model=\"whisper-1\", \n", + " file=audio_file_obj\n", + " )\n", + " return transcript.text\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "70c79408-f5f4-424b-b96c-d07e6893af6a", + "metadata": {}, + "outputs": [], + "source": [ + "# More involved Gradio code as we're not using the preset Chat interface!\n", + "# Passing in inbrowser=True in the last line will cause a Gradio window to pop up immediately.\n", + "\n", + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500)\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\", scale=4)\n", + " submit_btn = gr.Button(\"Submit\", scale=1)\n", + " with gr.Row():\n", + " audio_input = gr.Audio(sources=[\"microphone\"], type=\"filepath\", label=\"Speak to our AI Assistant\", scale=4)\n", + " submit_audio_btn = gr.Button(\"Submit Audio\", scale=1)\n", + "\n", + " with gr.Row():\n", + " models = [\"Claude\", \"Gemini\", \"GPT\", \"Ollama\"]\n", + " model_dropdown = gr.Dropdown(\n", + " label=\"Select a model\",\n", + " choices=models,\n", + " value=models[2]\n", + " )\n", + "\n", + " audio_options = [\"Yes\", \"No\"]\n", + " audio_dropdown = gr.Dropdown(\n", + " label=\"Select whether to respond with audio\",\n", + " choices=audio_options,\n", + " value=audio_options[1]\n", + " )\n", + " \n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def user_message_updater(user_message, history):\n", + " return \"\", history + [[user_message, None]]\n", + "\n", + " def chat_with_assistant(history, target_model, use_audio_output):\n", + " messages = []\n", + " for msg_user, msg_assistant in history:\n", + " messages.append({\"role\": \"user\", \"content\": msg_user})\n", + " if msg_assistant:\n", + " messages.append({\"role\": \"assistant\", \"content\": msg_assistant})\n", + " \n", + " user_message = history[-1][0]\n", + " final_response_content = \"\"\n", + " \n", + " if target_model == \"Claude\":\n", + " response = claude.messages.create(\n", + " model=MODEL_CLAUDE,\n", + " max_tokens=200,\n", + " temperature=0.7,\n", + " system=system_prompt_with_url_data,\n", + " messages=messages,\n", + " tools=tools_claude,\n", + " )\n", + " \n", + " tool_calls = [content_block for content_block in response.content if content_block.type == \"tool_use\"]\n", + " if tool_calls:\n", + " tool_use = tool_calls[0]\n", + " tool_output_content = scrape_urls(tool_use.input[\"text\"])\n", + " \n", + " messages.append({\"role\": \"assistant\", \"content\": response.content})\n", + " messages.append({\n", + " \"role\": \"user\",\n", + " \"content\": [\n", + " {\n", + " \"type\": \"tool_result\",\n", + " \"tool_use_id\": tool_use.id,\n", + " \"content\": tool_output_content\n", + " }\n", + " ]\n", + " })\n", + "\n", + " response = claude.messages.create(\n", + " model=MODEL_CLAUDE,\n", + " max_tokens=200,\n", + " temperature=0.7,\n", + " system=system_prompt_with_url_data,\n", + " messages=messages,\n", + " )\n", + " final_response_content = response.content[0].text\n", + "\n", + " elif target_model == \"Gemini\":\n", + " messages_gemini = []\n", + " for m in history:\n", + " messages_gemini.append({\"role\": \"user\", \"parts\": [{\"text\": m[0]}]})\n", + " if m[1]:\n", + " messages_gemini.append({\"role\": \"model\", \"parts\": [{\"text\": m[1]}]})\n", + " \n", + " model = google.generativeai.GenerativeModel(\n", + " model_name=MODEL_GEMINI,\n", + " system_instruction=system_message,\n", + " tools=tools_gemini\n", + " )\n", + " \n", + " chat = model.start_chat(history=messages_gemini[:-1])\n", + " response = chat.send_message(messages_gemini[-1])\n", + "\n", + " # Check if the response is a tool call before trying to extract text\n", + " if response.candidates[0].content.parts[0].function_call:\n", + " tool_call = response.candidates[0].content.parts[0].function_call\n", + " response_tool = handle_tool_call(tool_call, user_message)\n", + "\n", + " tool_response_content = json.loads(response_tool[\"content\"])\n", + " tool_response_gemini = {\n", + " \"role\": \"tool\",\n", + " \"parts\": [{\n", + " \"function_response\": {\n", + " \"name\": tool_call.name,\n", + " \"response\": tool_response_content\n", + " }\n", + " }]\n", + " }\n", + " \n", + " # Send the tool output back and get a new response\n", + " response = chat.send_message(tool_response_gemini)\n", + " final_response_content = response.text\n", + " else:\n", + " # If the original response was not a tool call, get the text directly\n", + " final_response_content = response.text\n", + "\n", + " elif target_model == \"Ollama\":\n", + " messages_ollama = [{\"role\": \"system\", \"content\": system_message}] + messages\n", + " response = ollama.chat(\n", + " model=MODEL_LLAMA,\n", + " messages=messages_ollama,\n", + " stream=False,\n", + " tools=tools_gpt_ollama,\n", + " )\n", + "\n", + " if 'tool_calls' in response['message'] and response['message']['tool_calls']:\n", + " response_tool = handle_tool_call(response['message']['tool_calls'][0], user_message)\n", + " messages_ollama.append({\"role\": \"assistant\", \"content\": response['message']['content'], \"tool_calls\": response['message']['tool_calls']})\n", + " messages_ollama.append(response_tool)\n", + " \n", + " response = ollama.chat(\n", + " model=MODEL_LLAMA,\n", + " messages=messages_ollama,\n", + " stream=False,\n", + " )\n", + " final_response_content = response['message']['content']\n", + " \n", + " else: # Assuming GPT is default\n", + " messages_gpt = [{\"role\": \"system\", \"content\": system_message}] + messages\n", + " response_stream = openai.chat.completions.create(model=MODEL_GPT, messages=messages_gpt, stream=True, tools=tools_gpt_ollama)\n", + " final_response_content = \"\"\n", + " for chunk in response_stream:\n", + " content = chunk.choices[0].delta.content or \"\"\n", + " tool_calls_chunk = chunk.choices[0].delta.tool_calls\n", + " if content:\n", + " final_response_content += content\n", + " \n", + " if tool_calls_chunk:\n", + " tool_call = tool_calls_chunk[0]\n", + " response_tool = handle_tool_call(tool_call, user_message)\n", + " \n", + " messages_gpt.append({\"role\": \"assistant\", \"tool_calls\": [tool_call]})\n", + " messages_gpt.append(response_tool)\n", + " \n", + " response_stream_after_tool = openai.chat.completions.create(model=MODEL_GPT, messages=messages_gpt, stream=True)\n", + " for chunk_after_tool in response_stream_after_tool:\n", + " final_response_content += chunk_after_tool.choices[0].delta.content or \"\"\n", + " break\n", + "\n", + " history[-1][1] = final_response_content\n", + " \n", + " if use_audio_output != \"No\":\n", + " talker(final_response_content)\n", + "\n", + " return history\n", + "\n", + " def transcribe_and_chat(audio_file, history, target_model, use_audio_output):\n", + " if audio_file:\n", + " transcribed_text = transcribe_audio(audio_file)\n", + " new_history = history + [[transcribed_text, None]]\n", + " return chat_with_assistant(new_history, target_model, use_audio_output)\n", + " else:\n", + " return history\n", + "\n", + " entry.submit(\n", + " user_message_updater,\n", + " inputs=[entry, chatbot],\n", + " outputs=[entry, chatbot],\n", + " queue=False\n", + " ).then(\n", + " chat_with_assistant,\n", + " inputs=[chatbot, model_dropdown, audio_dropdown],\n", + " outputs=[chatbot]\n", + " )\n", + "\n", + " submit_btn.click(\n", + " user_message_updater,\n", + " inputs=[entry, chatbot],\n", + " outputs=[entry, chatbot],\n", + " queue=False\n", + " ).then(\n", + " chat_with_assistant,\n", + " inputs=[chatbot, model_dropdown, audio_dropdown],\n", + " outputs=[chatbot]\n", + " )\n", + "\n", + " audio_input.stop(\n", + " transcribe_and_chat,\n", + " inputs=[audio_input, chatbot, model_dropdown, audio_dropdown],\n", + " outputs=[chatbot],\n", + " queue=False\n", + " )\n", + "\n", + " submit_audio_btn.click(\n", + " transcribe_and_chat,\n", + " inputs=[audio_input, chatbot, model_dropdown, audio_dropdown],\n", + " outputs=[chatbot],\n", + " queue=False\n", + " )\n", + " \n", + " clear.click(lambda: None, inputs=None, outputs=[chatbot], queue=False)\n", + "\n", + "ui.launch(inbrowser=True)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eb23b6cb-27af-43d6-8234-fe8295e7fe57", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/technical-question-answerer-with-gradio-v3.ipynb b/week2/community-contributions/technical-question-answerer-with-gradio-v3.ipynb new file mode 100644 index 0000000..f6b4146 --- /dev/null +++ b/week2/community-contributions/technical-question-answerer-with-gradio-v3.ipynb @@ -0,0 +1,182 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2b57204f-3e19-4d11-8901-c0e153ad9992", + "metadata": {}, + "source": [ + "## Technical Question Answerer With Gradio\n", + "- Ask a technical question to a chatbot embued with multimodal capabilities.\n", + "- Choose between different models (e.g. OpenAI's GPT, Anthropic's Claude)." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bd8e9bef-87ab-46d6-9393-bb308d7e5bc4", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from bs4 import BeautifulSoup\n", + "from typing import List\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import google.generativeai\n", + "import anthropic\n", + "\n", + "import gradio as gr\n", + "import base64\n", + "from io import BytesIO\n", + "from PIL import Image\n", + "from IPython.display import Audio, display\n", + "\n", + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")\n", + "\n", + "# Connect to OpenAI, Anthropic and Google; comment out the Claude or Google lines if you're not using them\n", + "openai = OpenAI()\n", + "claude = anthropic.Anthropic()\n", + "# google.generativeai.configure()\n", + "\n", + "system_message = \"You are a helpful assistant that explains technical contents and responds in markdown\"\n", + "\n", + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\",\n", + " input=message)\n", + "\n", + " audio_stream = BytesIO(response.content)\n", + " output_filename = \"output_audio.mp3\"\n", + " with open(output_filename, \"wb\") as f:\n", + " f.write(audio_stream.read())\n", + "\n", + " display(Audio(output_filename, autoplay=True))\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "20486a61-5d59-4370-b92c-3b7fec63835c", + "metadata": {}, + "outputs": [], + "source": [ + "# --- Chat functions ---\n", + "def chat_gpt(history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4o-mini\",\n", + " messages=messages\n", + " )\n", + " reply = response.choices[0].message.content\n", + " history = history + [{\"role\": \"assistant\", \"content\": reply}]\n", + " talker(reply) # make it talk\n", + " return history\n", + "\n", + "\n", + "claude_via_openai_client = OpenAI(\n", + " api_key=anthropic_api_key, \n", + " base_url=\"https://api.anthropic.com/v1\"\n", + ")\n", + "\n", + "def chat_claude(history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = claude_via_openai_client.chat.completions.create(\n", + " model=\"claude-3-haiku-20240307\",\n", + " messages=messages\n", + " )\n", + " reply = response.choices[0].message.content\n", + " history = history + [{\"role\": \"assistant\", \"content\": reply}]\n", + " talker(reply) # make it talk\n", + " return history\n", + "\n", + "\n", + "# --- Gradio UI ---\n", + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, type=\"messages\")\n", + " with gr.Row():\n", + " the_model = gr.Dropdown([\"GPT\", \"Claude\"], label=\"Select model\", value=\"GPT\")\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\")\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def do_entry(message, history, model):\n", + " # add user turn\n", + " history = history + [{\"role\": \"user\", \"content\": message}]\n", + " # call selected model\n", + " if model == \"GPT\":\n", + " history = chat_gpt(history)\n", + " elif model == \"Claude\":\n", + " history = chat_claude(history)\n", + " return \"\", history\n", + "\n", + " entry.submit(\n", + " fn=do_entry,\n", + " inputs=[entry, chatbot, the_model],\n", + " outputs=[entry, chatbot] # only 2 outputs\n", + " )\n", + "\n", + " clear.click(lambda: [], None, chatbot, queue=False)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "13974664-2965-46b9-9c56-714c70d3f835", + "metadata": {}, + "outputs": [], + "source": [ + "ui.launch(inbrowser=True)\n", + "\n", + "# prompt = \"\"\"\n", + "# Please explain what this code does and why:\n", + "# yield from {book.get(\"author\") for book in books if book.get(\"author\")}\n", + "# \"\"\"" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/weather_agent.ipynb b/week2/community-contributions/weather_agent.ipynb new file mode 100644 index 0000000..f89978a --- /dev/null +++ b/week2/community-contributions/weather_agent.ipynb @@ -0,0 +1,370 @@ +{ + "cells": [ + { + "cell_type": "code", + "execution_count": 1, + "id": "60761989", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import requests\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import speech_recognition as sr\n", + "import json\n", + "from dotenv import load_dotenv" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "e0b6610a", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "OpenAI API Key exists and begins sk-proj-\n", + "weather API Key exists\n" + ] + } + ], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "weather_api_key = os.getenv('WEATHER_API_KEY')\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + "if weather_api_key:\n", + " print(\"weather API Key exists\")\n", + "else:\n", + " print(\"weather API Key not set\")\n", + " \n", + "MODEL = \"gpt-4o-mini\"\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "af9d2faf", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"You are a helpful assistant for weather. \"\n", + "system_message += \"You need to fetch the current, historical and forecast the weather data using weather api and provide the response\"" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "2c5208d8", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_current_weather(location):\n", + " url = f\"http://api.weatherapi.com/v1/current.json?key={weather_api_key}&q={location}&aqi=yes\"\n", + " return requests.get(url).json()" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "8e6a12e5", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_forecast_weather(location, days=3):\n", + " url = f\"http://api.weatherapi.com/v1/forecast.json?key={weather_api_key}&q={location}&days={days}&aqi=yes&alerts=yes\"\n", + " return requests.get(url).json()" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "eafc468e", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_historical_weather(location, date):\n", + " url = f\"http://api.weatherapi.com/v1/history.json?key={weather_api_key}&q={location}&dt={date}&aqi=yes\"\n", + " return requests.get(url).json()" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "2851ed55", + "metadata": {}, + "outputs": [], + "source": [ + "# Weather function used as a tool by OpenAI\n", + "def get_weatherapi_data(location, mode=\"current\", date=None, forecast_days=3):\n", + " if mode == \"current\":\n", + " return fetch_current_weather(location)\n", + " elif mode == \"forecast\":\n", + " return fetch_forecast_weather(location, days=forecast_days)\n", + " elif mode == \"historical\":\n", + " if not date:\n", + " # Default: yesterday\n", + " date = (datetime.date.today() - datetime.timedelta(days=1)).strftime(\"%Y-%m-%d\")\n", + " return fetch_historical_weather(location, date)\n", + " else:\n", + " return {\"error\": \"Unknown mode.\"}" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "368176c2", + "metadata": {}, + "outputs": [], + "source": [ + "# Tool schema for OpenAI tool-calling\n", + "weatherapi_tool_schema = [\n", + " {\n", + " \"type\": \"function\",\n", + " \"function\": {\n", + " \"name\": \"get_weatherapi_data\",\n", + " \"description\": \"Fetches current, forecast, or historical weather data from WeatherAPI.com for a given location.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"location\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Name of the city, region, or coordinates.\"\n", + " },\n", + " \"mode\": {\n", + " \"type\": \"string\",\n", + " \"enum\": [\"current\", \"forecast\", \"historical\"],\n", + " \"description\": \"Type of weather data required.\"\n", + " },\n", + " \"date\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"Date for historical data in YYYY-MM-DD format. Only needed if mode is 'historical'.\"\n", + " },\n", + " \"forecast_days\": {\n", + " \"type\": \"integer\",\n", + " \"description\": \"Number of forecast days (1-10). Only needed if mode is 'forecast'.\"\n", + " }\n", + " },\n", + " \"required\": [\"location\", \"mode\"]\n", + " }\n", + " }\n", + " }\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "bd9c4d38", + "metadata": {}, + "outputs": [], + "source": [ + "def audio_to_text(audio_filepath):\n", + " if audio_filepath is None or audio_filepath == \"\":\n", + " return \"\"\n", + " recognizer = sr.Recognizer()\n", + " try:\n", + " with sr.AudioFile(audio_filepath) as source:\n", + " audio = recognizer.record(source)\n", + " try:\n", + " transcript = recognizer.recognize_google(audio)\n", + " return transcript\n", + " except sr.UnknownValueError:\n", + " return \"\"\n", + " except sr.RequestError as e:\n", + " return f\"Speech recognition service error: {e}\"\n", + " except Exception as e:\n", + " return f\"Error opening audio file: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "61c5de82", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_agent(city, mode, date, forecast_days, audio=None):\n", + " user_query = city\n", + " if audio:\n", + " spoken_text = audio_to_text(audio)\n", + " print(\"Recognized speech:\", spoken_text)\n", + " if spoken_text and spoken_text.strip().lower() != \"flic en flac\":\n", + " user_query = spoken_text\n", + " else:\n", + " if not city.strip():\n", + " return \"Sorry, I could not recognize your speech. Please try again or type your city.\"\n", + "\n", + " if not user_query.strip():\n", + " return \"Please provide a location by text or speech.\"\n", + "\n", + " # Compose tool function arguments as the LLM would\n", + " args = {\n", + " \"location\": user_query,\n", + " \"mode\": mode\n", + " }\n", + " if mode == \"historical\" and date:\n", + " args[\"date\"] = date\n", + " if mode == \"forecast\":\n", + " try:\n", + " n_days = int(forecast_days)\n", + " except:\n", + " n_days = 3\n", + " args[\"forecast_days\"] = n_days\n", + "\n", + " openai.api_key = openai_api_key\n", + "\n", + " # LLM call for tool use\n", + " response = openai.chat.completions.create(\n", + " model=\"gpt-4-0613\",\n", + " messages=[{\"role\": \"user\", \"content\": f\"Get me {mode} weather for {user_query}\"+(f' on {date}' if date and mode==\"historical\" else \"\")+(f' for {forecast_days} days' if forecast_days and mode==\"forecast\" else \"\")}],\n", + " tools=weatherapi_tool_schema,\n", + " tool_choice={\"type\": \"function\", \"function\": {\"name\": \"get_weatherapi_data\", \"arguments\": json.dumps(args)}}\n", + " )\n", + " message = response.choices[0].message\n", + "\n", + " if hasattr(message, \"tool_calls\") and message.tool_calls:\n", + " tool_call = message.tool_calls[0]\n", + " args2 = json.loads(tool_call.function.arguments) # not really needed, already have args\n", + " location = args2.get(\"location\", user_query)\n", + " mode = args2.get(\"mode\", mode)\n", + " date = args2.get(\"date\", date)\n", + " forecast_days = args2.get(\"forecast_days\", forecast_days)\n", + " weather_data = get_weatherapi_data(location, mode, date, forecast_days)\n", + " tool_result = f\"Weather data (mode={mode}) for {location}:\\n{json.dumps(weather_data, indent=2)[:3000]}\"\n", + " followup = openai.chat.completions.create(\n", + " model=\"gpt-4-0613\",\n", + " messages=[\n", + " {\"role\": \"user\", \"content\": f\"Get me {mode} weather for {location}\"},\n", + " message,\n", + " {\n", + " \"role\": \"tool\",\n", + " \"tool_call_id\": tool_call.id,\n", + " \"content\": tool_result\n", + " }\n", + " ]\n", + " )\n", + " answer = followup.choices[0].message.content.strip()\n", + " return answer\n", + " else:\n", + " return getattr(message, \"content\", \"\")" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "44071389", + "metadata": {}, + "outputs": [], + "source": [ + "def update_date_visibility(mode):\n", + " return gr.update(visible=(mode==\"historical\"))\n", + "\n", + "def update_days_visibility(mode):\n", + " return gr.update(visible=(mode==\"forecast\"))" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "618a5494", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7861\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 13, + "metadata": {}, + "output_type": "execute_result" + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Recognized speech: Error opening audio file: FLAC conversion utility not available - consider installing the FLAC command line application by running `apt-get install flac` or your operating system's equivalent\n" + ] + } + ], + "source": [ + "with gr.Blocks() as demo:\n", + " gr.Markdown(\"## Weather Chat Agent (Current, Historical, Forecast)\")\n", + "\n", + " with gr.Row():\n", + " city_input = gr.Textbox(label=\"City/Location\")\n", + " mode_input = gr.Dropdown(\n", + " [\"current\", \"historical\", \"forecast\"],\n", + " value=\"current\",\n", + " label=\"Weather Mode\")\n", + " with gr.Row():\n", + " date_input = gr.Textbox(label=\"Date for historical (YYYY-MM-DD)\", visible=False)\n", + " days_input = gr.Textbox(label=\"Forecast Days (for forecast)\", value=\"3\", visible=False)\n", + " audio_input = gr.Audio(type=\"filepath\", format=\"wav\", label=\"Or Speak your City/Location (optional)\")\n", + " output_box = gr.Textbox(label=\"Weather Info\", lines=8)\n", + " btn = gr.Button(\"Get Weather\")\n", + "\n", + " # Show/hide date and days inputs based on dropdown\n", + " mode_input.change(update_date_visibility, mode_input, date_input)\n", + " mode_input.change(update_days_visibility, mode_input, days_input)\n", + " btn.click(\n", + " chat_agent,\n", + " [city_input, mode_input, date_input, days_input, audio_input],\n", + " output_box\n", + " )\n", + "\n", + "demo.launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/week2-exercise-sentence-translate-and-counter-agent.ipynb b/week2/community-contributions/week2-exercise-sentence-translate-and-counter-agent.ipynb new file mode 100644 index 0000000..cc0cb58 --- /dev/null +++ b/week2/community-contributions/week2-exercise-sentence-translate-and-counter-agent.ipynb @@ -0,0 +1,297 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "d006b2ea-9dfe-49c7-88a9-a5a0775185fd", + "metadata": {}, + "source": [ + "# Additional End of week Exercise - week 2\n", + "\n", + "Now use everything you've learned from Week 2 to build a full prototype for the technical question/answerer you built in Week 1 Exercise.\n", + "\n", + "This should include a Gradio UI, streaming, use of the system prompt to add expertise, and the ability to switch between models. Bonus points if you can demonstrate use of a tool!\n", + "\n", + "If you feel bold, see if you can add audio input so you can talk to it, and have it respond with audio. ChatGPT or Claude can help you, or email me if you have questions.\n", + "\n", + "I will publish a full solution here soon - unless someone beats me to it...\n", + "\n", + "There are so many commercial applications for this, from a language tutor, to a company onboarding solution, to a companion AI to a course (like this one!) I can't wait to see your results." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a07e7793-b8f5-44f4-aded-5562f633271a", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import json\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2118e80a-6181-4488-95cf-c9da0500ea56", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "# Print the key prefixes to help with any debugging\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + "else:\n", + " print(\"Google API Key not set\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ddc4764-e7f6-4512-8210-51bbfefbb3a9", + "metadata": {}, + "outputs": [], + "source": [ + "# Set base url\n", + "\n", + "GEMINI_BASE_URL = \"https://generativelanguage.googleapis.com/v1beta/openai/\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "91bfd734-9c5e-4993-808e-b66489a92d4d", + "metadata": {}, + "outputs": [], + "source": [ + "# Connect to OpenAI, Anthropic and Google; comment out the Claude or Google lines if you're not using them\n", + "\n", + "openai = OpenAI()\n", + "gemini = OpenAI(base_url=GEMINI_BASE_URL, api_key=google_api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d9ee11ae-23e2-42cc-b63d-b446f6d83c99", + "metadata": {}, + "outputs": [], + "source": [ + "# Set models\n", + "\n", + "gpt_model = \"gpt-4.1-mini\"\n", + "gemini_model = \"gemini-2.0-flash\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a01d270e-f62e-41b3-8e46-ac173d7a1493", + "metadata": {}, + "outputs": [], + "source": [ + "system_gpt_prompt = \"You are an assistant with general knowledge obtained from the internet. \\\n", + "Always respond with a cheerful tone. If you don’t know the answer to a question, simply say that you don’t know.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4e85c8ed-3ba4-4283-8480-6979b0d5602f", + "metadata": {}, + "outputs": [], + "source": [ + "system_gemini_prompt = \"You are an expert translator with knowledge of all existing languages. \\\n", + "Your only task is, given a provided sentence, to translate it into the specified target language. \\\n", + "Do not provide anything else in your response only the translation itself.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ee0c887-a63f-48dd-8eaf-68b0bf9263b6", + "metadata": {}, + "outputs": [], + "source": [ + "def count_letter_tool(sentence, letter):\n", + "\n", + " if len(letter) != 1:\n", + " return \"You need to provide a single letter to count\"\n", + " \n", + " return sentence.lower().count(letter.lower())" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5f1ae918-cb99-4e60-80d3-37e16e514f55", + "metadata": {}, + "outputs": [], + "source": [ + "def translator_tool(sentence, language):\n", + " user_message = f\"Please translate this sentence: \\\"{sentence}\\\" to this language: {language}\"\n", + " messages = [{\"role\": \"system\", \"content\": system_gemini_prompt}, {\"role\": \"user\", \"content\":user_message}]\n", + " response = gemini.chat.completions.create(model=gemini_model, messages=messages)\n", + "\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2d499f2a-23b2-4fff-9d2d-f2333cbd109a", + "metadata": {}, + "outputs": [], + "source": [ + "count_letter_function = {\n", + " \"name\": \"count_letter_tool\",\n", + " \"description\": \"Count the number of a particular letter in a sentence. Call this whenever you need to know how many times a letter appears in a sentence, for example when a user asks 'How many 'a' are in this sentence?'\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"sentence\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The sentence provided by the user for counting.\"\n", + " },\n", + " \"letter\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The letter to count in the sentence.\"\n", + " }\n", + " },\n", + " \"required\": [\"sentence\", \"letter\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b58079a8-8def-4fa6-8273-34bf8eeb8cb5", + "metadata": {}, + "outputs": [], + "source": [ + "translator_function = {\n", + " \"name\": \"translator_tool\",\n", + " \"description\": \"Translate a sentence provided by the user. Call this whenever a translation is needed, for example when a user asks 'Can you translate \\\"hola como estás?\\\" to English?'\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"sentence\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The sentence provided by the user to translate.\"\n", + " },\n", + " \"language\": {\n", + " \"type\": \"string\",\n", + " \"description\": \"The target language to translate the sentence into.\"\n", + " }\n", + " },\n", + " \"required\": [\"sentence\", \"language\"],\n", + " \"additionalProperties\": False\n", + " }\n", + "}" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7ab7fc93-3540-48e5-bbe0-3e9ad2bbce15", + "metadata": {}, + "outputs": [], + "source": [ + "tools = [{\"type\": \"function\", \"function\": count_letter_function}, {\"type\": \"function\", \"function\": translator_function}]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "678ccc37-c034-4035-bc3c-00fa8bcd8e64", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_gpt_prompt}] + history + [{\"role\": \"user\", \"content\": message}]\n", + " response = openai.chat.completions.create(model=gpt_model, messages=messages, tools=tools)\n", + "\n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " response = handle_tool_call(message)\n", + " messages.append(message)\n", + " messages.append(response)\n", + " response = openai.chat.completions.create(model=gpt_model, messages=messages)\n", + " \n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2a1138e4-f849-4557-a74c-f9feb1572854", + "metadata": {}, + "outputs": [], + "source": [ + "def handle_tool_call(message):\n", + " tool_call = message.tool_calls[0]\n", + " arguments = json.loads(tool_call.function.arguments)\n", + " sentence = arguments.get('sentence')\n", + " response =\"\"\n", + " match tool_call.function.name:\n", + " case \"translator_tool\":\n", + " language = arguments.get('language')\n", + " translation = translator_tool(sentence, language)\n", + " response = {\"role\": \"tool\", \"content\": json.dumps({\"translation\": translation}), \"tool_call_id\": tool_call.id}\n", + " case \"count_letter_tool\":\n", + " letter = arguments.get('letter')\n", + " count = count_letter_tool(sentence, letter)\n", + " response = {\"role\": \"tool\", \"content\": json.dumps({\"count\": count}), \"tool_call_id\": tool_call.id}\n", + "\n", + " return response" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d39344cc-9e89-47a0-9249-2e182091ee43", + "metadata": {}, + "outputs": [], + "source": [ + "gr.ChatInterface(fn=chat, type=\"messages\").launch()" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/week2_exercise_by_abrar.ipynb b/week2/community-contributions/week2_exercise_by_abrar.ipynb new file mode 100644 index 0000000..3141217 --- /dev/null +++ b/week2/community-contributions/week2_exercise_by_abrar.ipynb @@ -0,0 +1,490 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "ddfa9ae6-69fe-444a-b994-8c4c5970a7ec", + "metadata": {}, + "source": [ + "# Project - Cricket Anaylyst AI Assistant\n", + "\n", + "Cricket Analyst AI Assistant is an intelligent tool that analyzes cricket data to compare players, evaluate performances across formats, and provide insightful statistics. It processes historical and recent match data to deliver easy-to-understand summaries, helping fans, analysts, and coaches make informed decisions." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8b50bbe2-c0b1-49c3-9a5c-1ba7efa2bcb4", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import json\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import gradio as gr\n", + "import speech_recognition as sr\n", + "import pandas as pd\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "747e8786-9da8-4342-b6c9-f5f69c2e22ae", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "MODEL = \"gpt-4o-mini\"\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0a521d84-d07c-49ab-a0df-d6451499ed97", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = \"\"\"\n", + "You are a Cricket Analyst AI with deep knowledge of cricket statistics and match analysis.\n", + "When comparing players, call the `analyze_cricket` tool to get factual data before answering.\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d7be34a6-7288-43b0-ad4e-bbed836cb786", + "metadata": {}, + "outputs": [], + "source": [ + "# Sample cricket stats as a list of dicts\n", + "cricket_data = [\n", + " {\"Player\": \"Virat Kohli\", \"Format\": \"ODI\", \"Year\": 2023, \"Runs\": 1377, \"Matches\": 27, \"Average\": 57.37, \"StrikeRate\": 93.21},\n", + " {\"Player\": \"Virat Kohli\", \"Format\": \"ODI\", \"Year\": 2022, \"Runs\": 765, \"Matches\": 20, \"Average\": 42.50, \"StrikeRate\": 88.40},\n", + " {\"Player\": \"Virat Kohli\", \"Format\": \"ODI\", \"Year\": 2021, \"Runs\": 560, \"Matches\": 15, \"Average\": 40.00, \"StrikeRate\": 90.10},\n", + " {\"Player\": \"Babar Azam\", \"Format\": \"ODI\", \"Year\": 2023, \"Runs\": 1454, \"Matches\": 26, \"Average\": 62.00, \"StrikeRate\": 89.50},\n", + " {\"Player\": \"Babar Azam\", \"Format\": \"ODI\", \"Year\": 2022, \"Runs\": 1198, \"Matches\": 18, \"Average\": 66.55, \"StrikeRate\": 92.00},\n", + " {\"Player\": \"Babar Azam\", \"Format\": \"ODI\", \"Year\": 2021, \"Runs\": 949, \"Matches\": 15, \"Average\": 67.78, \"StrikeRate\": 90.50},\n", + " {\"Player\": \"Joe Root\", \"Format\": \"Test\", \"Year\": 2025, \"Runs\": 949, \"Matches\": 15, \"Average\": 69.78, \"StrikeRate\": 95.50},\n", + " {\"Player\": \"Joe Root\", \"Format\": \"Test\", \"Year\": 2024, \"Runs\": 2025, \"Matches\": 22, \"Average\": 68.78, \"StrikeRate\": 90.50},\n", + " {\"Player\": \"Harry Brook\", \"Format\": \"Test\", \"Year\": 2025, \"Runs\": 1056, \"Matches\": 16, \"Average\": 67.78, \"StrikeRate\": 95.50},\n", + " {\"Player\": \"Harry Brook\", \"Format\": \"Test\", \"Year\": 2024, \"Runs\": 2200, \"Matches\": 21, \"Average\": 71.78, \"StrikeRate\": 98.50},\n", + "\n", + "]" + ] + }, + { + "cell_type": "markdown", + "id": "36bedabf-a0a7-4985-ad8e-07ed6a55a3a4", + "metadata": {}, + "source": [ + "## Tools\n", + "\n", + "Tools starts from here. \n", + "For this notebook, I have just wrote one Tool, you can add multiple tools for your agent." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "46a0a260-b11b-4bde-ab80-911a81e2c281", + "metadata": {}, + "outputs": [], + "source": [ + "def analyze_cricket(data_list, player1, match_format=\"ODI\", years=3):\n", + " \"\"\"\n", + " Return cricket players' performances using an in-memory list of dicts.\n", + " \"\"\"\n", + " print(\"Tool 'analyze_cricket' is called\")\n", + " df = pd.DataFrame(data_list)\n", + " latest_year = df['Year'].max()\n", + " min_year = latest_year - years + 1\n", + "\n", + " filtered = df[\n", + " (df['Format'].str.upper() == match_format.upper()) &\n", + " (df['Year'] >= min_year) &\n", + " (df['Player'].isin([player1]))\n", + " ]\n", + " if filtered.empty:\n", + " return {\"error\": f\"No data found for {player1} in {match_format} for last {years} years.\"}\n", + "\n", + " summary = filtered.groupby(\"Player\").agg({\n", + " \"Matches\": \"sum\",\n", + " \"Runs\": \"sum\",\n", + " \"Average\": \"mean\",\n", + " \"StrikeRate\": \"mean\"\n", + " }).round(2)\n", + "\n", + " return summary.reset_index().to_dict(orient=\"records\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cbfd413e-b5d4-42bd-b86f-ed9b4ee360eb", + "metadata": {}, + "outputs": [], + "source": [ + "# Example usage:\n", + "result = analyze_cricket(cricket_data, \"Virat Kohli\", \"ODI\", 3)\n", + "print(result)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c92699c1-802b-4948-a654-df89e0c19adb", + "metadata": {}, + "outputs": [], + "source": [ + "# Tool definition \n", + "analyze_cricket_functions = {\n", + " \"name\": \"analyze_cricket\",\n", + " \"description\": \"Compare two cricket players' performances over the last N years.\",\n", + " \"parameters\": {\n", + " \"type\": \"object\",\n", + " \"properties\": {\n", + " \"player1\": {\"type\": \"string\", \"description\": \"Name of first player\"},\n", + " # \"player2\": {\"type\": \"string\", \"description\": \"Name of second player\"},\n", + " \"match_format\": {\"type\": \"string\", \"enum\": [\"ODI\", \"Test\", \"T20\"], \"description\": \"Format of the match\"},\n", + " \"years\": {\"type\": \"integer\", \"description\": \"Number of years to compare\"}\n", + " },\n", + " \"required\": [\"player1\"]\n", + " }\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bdca8679-935f-4e7f-97e6-e71a4d4f228c", + "metadata": {}, + "outputs": [], + "source": [ + "# And this is included in a list of tools:\n", + "\n", + "tools = [{\"type\": \"function\", \"function\": analyze_cricket_functions}]" + ] + }, + { + "cell_type": "markdown", + "id": "c3d3554f-b4e3-4ce7-af6f-68faa6dd2340", + "metadata": {}, + "source": [ + "## Getting OpenAI to use our Tool\n", + "\n", + "There's some fiddly stuff to allow OpenAI \"to call our tool\"\n", + "\n", + "What we actually do is give the LLM the opportunity to inform us that it wants us to run the tool.\n", + "\n", + "Here's how the new chat function looks:" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ce9b0744-9c78-408d-b9df-9f6fd9ed78cf", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(message, history):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history + [{\"role\": \"user\", \"content\": message}]\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + "\n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " messages.append(message)\n", + " for tool_call in message.tool_calls: \n", + " response, player1= handle_tool_call(tool_call)\n", + " messages.append(response)\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " return response.choices[0].message.content" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b0992986-ea09-4912-a076-8e5603ee631f", + "metadata": {}, + "outputs": [], + "source": [ + "# We have to write that function handle_tool_call:\n", + "\n", + "def handle_tool_call(tool_call):\n", + " # tool_call = message.tool_calls[0]\n", + " # print(\"tool_call.id\", tool_call.id)\n", + " arguments = json.loads(tool_call.function.arguments)\n", + " print(\"arguments\", arguments)\n", + " player1 = arguments.get('player1')\n", + " # player2 = arguments.get('player2')\n", + " match_format = arguments.get('match_format', 'ODI')\n", + " years = arguments.get('years', 3)\n", + " result = analyze_cricket(cricket_data, player1, match_format, years)\n", + " print(\"result from analyze_cricket function: \", tool_call.id, result)\n", + " response = {\n", + " \"role\": \"tool\",\n", + " \"content\": json.dumps(result),\n", + " \"tool_call_id\": tool_call.id\n", + " }\n", + " return response, player1" + ] + }, + { + "cell_type": "markdown", + "id": "924e7225-b76d-4518-abad-5bea5c356cf8", + "metadata": {}, + "source": [ + "# Sample User prompt\n", + "\n", + "1. ### Compare Babar and Virat in ODI matches over the last 3 years.\n", + "Here \n", + "Player1 is Babar\n", + "Player2 is Virat\n", + "match_format is ODI\n", + "years is 3\n", + "\n", + "\n", + "2. ### can you please give me the comparison of Virat and babar?\n", + "Here, you are not provided the info on the format and number of years. In this case, the function will pick the default values for the match format, which is ODI, and the years, which is 3.\n", + "\n", + "\n", + "3. ### Compare Rizwan and Babar in ODI Matches over the last years.\n", + "The given data is not available in the above data list. \n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f4be8a71-b19e-4c2f-80df-f59ff2661f14", + "metadata": {}, + "outputs": [], + "source": [ + "gr.ChatInterface(fn=chat, type=\"messages\").launch(inbrowser=True)" + ] + }, + { + "cell_type": "markdown", + "id": "473e5b39-da8f-4db1-83ae-dbaca2e9531e", + "metadata": {}, + "source": [ + "# Let's go multi-modal!!\n", + "\n", + "We can use DALL-E-3, the image generation model behind GPT-4o, to make us some images\n", + "\n", + "Let's put this in a function called artist.\n", + "\n", + "### Price alert: each time I generate an image it costs about 4 cents - don't go crazy with images!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2c27c4ba-8ed5-492f-add1-02ce9c81d34c", + "metadata": {}, + "outputs": [], + "source": [ + "# Some imports for handling images\n", + "\n", + "import base64\n", + "from io import BytesIO\n", + "from PIL import Image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "773a9f11-557e-43c9-ad50-56cbec3a0f8f", + "metadata": {}, + "outputs": [], + "source": [ + "def artist(player_names):\n", + " if len(player_names) <2 or len(player_names) > 2:\n", + " return None\n", + " player1 = player_names[0]\n", + " player2 = player_names[1]\n", + " image_response = openai.images.generate(\n", + " model=\"dall-e-3\",\n", + " prompt=f\"An image representing a comparison of {player1} and {player2}, showing their country flags and bowling or batting style\",\n", + " size=\"1024x1024\",\n", + " n=1,\n", + " response_format=\"b64_json\",\n", + " )\n", + " image_base64 = image_response.data[0].b64_json\n", + " image_data = base64.b64decode(image_base64)\n", + " return Image.open(BytesIO(image_data))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d877c453-e7fb-482a-88aa-1a03f976b9e9", + "metadata": {}, + "outputs": [], + "source": [ + "image = artist([\"Babar\", \"root\"])\n", + "display(image)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ffbfe93b-5e86-4e68-ba71-b301cd5230db", + "metadata": {}, + "outputs": [], + "source": [ + "from pydub import AudioSegment\n", + "from pydub.playback import play\n", + "\n", + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\", # Also, try replacing onyx with alloy\n", + " input=message\n", + " )\n", + " \n", + " audio_stream = BytesIO(response.content)\n", + " audio = AudioSegment.from_file(audio_stream, format=\"mp3\")\n", + " play(audio)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b88d775d-d357-4292-a1ad-5dc5ed567281", + "metadata": {}, + "outputs": [], + "source": [ + "talker(\"Well, hi there\") # For testing purposes" + ] + }, + { + "cell_type": "markdown", + "id": "1d48876d-c4fa-46a8-a04f-f9fadf61fb0d", + "metadata": {}, + "source": [ + "# Our Agent Framework\n", + "\n", + "The term 'Agentic AI' and Agentization is an umbrella term that refers to a number of techniques, such as:\n", + "\n", + "1. Breaking a complex problem into smaller steps, with multiple LLMs carrying out specialized tasks\n", + "2. The ability for LLMs to use Tools to give them additional capabilities\n", + "3. The 'Agent Environment' which allows Agents to collaborate\n", + "4. An LLM can act as the Planner, dividing bigger tasks into smaller ones for the specialists\n", + "5. The concept of an Agent having autonomy / agency, beyond just responding to a prompt - such as Memory\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ba820c95-02f5-499e-8f3c-8727ee0a6c0c", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(history, image_choice):\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages, tools=tools)\n", + " image = None\n", + " \n", + " if response.choices[0].finish_reason==\"tool_calls\":\n", + " message = response.choices[0].message\n", + " messages.append(message)\n", + " player_names = []\n", + " for tool_call in message.tool_calls:\n", + " response, player1= handle_tool_call(tool_call)\n", + " player_names.append(player1)\n", + " messages.append(response)\n", + " if image_choice.lower() == 'yes':\n", + " image = artist(player_names)\n", + " else:\n", + " print(\"Image value is NO\", image_choice)\n", + " \n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " \n", + " reply = response.choices[0].message.content\n", + " history += [{\"role\":\"assistant\", \"content\":reply}]\n", + "\n", + " # Comment out or delete the next line if you'd rather skip Audio for now..\n", + " talker(reply)\n", + " \n", + " return history, image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f38d0d27-33bf-4992-a2e5-5dbed973cde7", + "metadata": {}, + "outputs": [], + "source": [ + "# More involved Gradio code as we're not using the preset Chat interface!\n", + "# Passing in inbrowser=True in the last line will cause a Gradio window to pop up immediately.\n", + "\n", + "with gr.Blocks() as ui:\n", + " gr.Markdown(\"### 🏏 Cricket Analyst AI Assistant\")\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, type=\"messages\")\n", + " image_output = gr.Image(height=500)\n", + "\n", + " with gr.Row():\n", + " image_dropdown = gr.Dropdown(\n", + " choices=[\"Yes\", \"No\"],\n", + " label=\"Do you want image?\"\n", + " )\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Chat with our AI Assistant:\")\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def do_entry(message, history, image_choice):\n", + " history += [{\"role\": \"user\", \"content\": message}]\n", + " return \"\", history, image_choice\n", + "\n", + " entry.submit(\n", + " do_entry, \n", + " inputs=[entry, chatbot, image_dropdown], \n", + " outputs=[entry, chatbot, image_dropdown]\n", + " ).then(\n", + " chat, \n", + " inputs=[chatbot, image_dropdown], \n", + " outputs=[chatbot, image_output]\n", + " )\n", + " clear.click(lambda: None, inputs=None, outputs=chatbot, queue=False)\n", + "\n", + "ui.launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/week2_tennis.ipynb b/week2/community-contributions/week2_tennis.ipynb new file mode 100644 index 0000000..bed232f --- /dev/null +++ b/week2/community-contributions/week2_tennis.ipynb @@ -0,0 +1,331 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "ad6e5ed4-a38d-46a6-8bb5-32d68bd0b9e5", + "metadata": {}, + "source": [ + "End of week 2 exercise" + ] + }, + { + "cell_type": "code", + "execution_count": 38, + "id": "8f45fbfa-eaaa-4eb8-841e-83b068b80507", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import json\n", + "import gradio as gr\n", + "import base64\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from io import BytesIO\n", + "from PIL import Image\n", + "from pydub import AudioSegment\n", + "from pydub.playback import play\n" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "id": "451b72a6-1e6c-476a-8431-1c30c5cd9fb8", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "API key found and looks good so far!\n" + ] + } + ], + "source": [ + "# Initialization\n", + "\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "if openai_api_key:\n", + " print(\"API key found and looks good so far!\")\n", + "else:\n", + " print(\"No API key was found!\")\n", + "\n", + "MODEL = \"gpt-4o-mini\"\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 40, + "id": "29fa6a53-4b57-47ea-89a1-640020e603b4", + "metadata": {}, + "outputs": [], + "source": [ + "system_message = (\n", + " \"You are a helpful tennis coach who answers questions about tennis rules, \"\n", + " \"players, strategies, training, and equipment.\"\n", + " \"Give short, courteous answers, no more than 2 sentence.\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 41, + "id": "e9b255e7-02d8-4350-b5d4-e645d1fc90d3", + "metadata": {}, + "outputs": [], + "source": [ + "# Translation\n", + "\n", + "LANG_CODES = {\n", + " \"English\": \"en\",\n", + " \"Spanish\": \"es\",\n", + " \"French\": \"fr\"\n", + "}\n", + "\n", + "def translate_text(text, target_language=\"en\"):\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": f\"You are a translator. Translate the following text to {target_language}\"},\n", + " {\"role\": \"user\", \"content\": text}\n", + " ]\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " return response.choices[0].message.content\n", + "\n", + "def tennis_info_tool(query):\n", + " if \"top\" in query.lower():\n", + " return \"Top male players: Djokovic, Nadal, Federer. Top female players: Barty, Sabalenka, Swiatek.\"\n" + ] + }, + { + "cell_type": "code", + "execution_count": 42, + "id": "b44b147c-bfba-4137-9ecb-d5538f08a46d", + "metadata": {}, + "outputs": [], + "source": [ + "# Image\n", + "def generate_tennis_image(prompt):\n", + " image_response = openai.images.generate(\n", + " model=\"dall-e-3\",\n", + " prompt=f\"Tennis scene: {prompt}, realistic and detailed, vibrant colors\",\n", + " size=\"1024x1024\",\n", + " n=1,\n", + " response_format=\"b64_json\",\n", + " )\n", + " image_base64 = image_response.data[0].b64_json\n", + " image_data = base64.b64decode(image_base64)\n", + " return Image.open(BytesIO(image_data))" + ] + }, + { + "cell_type": "code", + "execution_count": 43, + "id": "bbfeff3b-0c73-4b2c-a6da-3cac27d8fedd", + "metadata": {}, + "outputs": [], + "source": [ + "# Audio\n", + "\n", + "def talker(message):\n", + " response = openai.audio.speech.create(\n", + " model=\"tts-1\",\n", + " voice=\"onyx\",\n", + " input=message\n", + " )\n", + "\n", + " audio_stream = BytesIO(response.content)\n", + " audio = AudioSegment.from_file(audio_stream, format=\"mp3\")\n", + " play(audio)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fa338332-3dfc-4b95-8367-65853a8d2793", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 44, + "id": "574d672e-0a75-4af9-b3ad-8dc2dec4e607", + "metadata": {}, + "outputs": [], + "source": [ + "def chat(history, user_message, target_language=\"English\", use_audio=False, generate_image=False):\n", + " image = None\n", + "\n", + " if any(keyword in user_message.lower() for keyword in [\"top\", \"players\"]):\n", + " reply = tennis_info_tool(user_message)\n", + " else:\n", + " messages = [{\"role\": \"system\", \"content\": system_message}] + history\n", + " response = openai.chat.completions.create(model=MODEL, messages=messages)\n", + " reply = response.choices[0].message.content\n", + " \n", + " if target_language != \"English\":\n", + " code = LANG_CODES.get(target_language, \"en\")\n", + " reply = translate_text(reply, code)\n", + "\n", + " history.append({\"role\": \"assistant\", \"content\": reply})\n", + " \n", + " if use_audio:\n", + " talker(reply)\n", + "\n", + " if generate_image:\n", + " image = generate_tennis_image(reply)\n", + " return history, image" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eaf4f47e-d20b-41f8-94b5-4aef0302731b", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": 45, + "id": "9110343f-0efa-49bc-8d5f-498fd690dd14", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "* Running on local URL: http://127.0.0.1:7869\n", + "* To create a public link, set `share=True` in `launch()`.\n" + ] + }, + { + "data": { + "text/html": [ + "
" + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + }, + { + "data": { + "text/plain": [] + }, + "execution_count": 45, + "metadata": {}, + "output_type": "execute_result" + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Input #0, wav, from '/var/folders/73/0s09hh2n48q7s14tld64q3rh0000gn/T/tmp4hoe_x5n.wav':\n", + " Duration: 00:00:06.55, bitrate: 384 kb/s\n", + " Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 24000 Hz, 1 channels, s16, 384 kb/s\n", + " 6.45 M-A: -0.000 fd= 0 aq= 0KB vq= 0KB sq= 0B " + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n" + ] + }, + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Input #0, wav, from '/var/folders/73/0s09hh2n48q7s14tld64q3rh0000gn/T/tmp2mxw0wth.wav':\n", + " Duration: 00:00:04.61, bitrate: 384 kb/s\n", + " Stream #0:0: Audio: pcm_s16le ([1][0][0][0] / 0x0001), 24000 Hz, 1 channels, s16, 384 kb/s\n", + " 4.48 M-A: 0.000 fd= 0 aq= 0KB vq= 0KB sq= 0B " + ] + }, + { + "name": "stdout", + "output_type": "stream", + "text": [ + "\n" + ] + } + ], + "source": [ + "# Gradio\n", + "\n", + "with gr.Blocks() as ui:\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, type=\"messages\")\n", + " image_output = gr.Image(height=500)\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Ask your Tennis Coach:\", placeholder=\"Type and press Enter...\")\n", + " target_lang_dropdown = gr.Dropdown(\n", + " choices=[\"English\", \"Spanish\", \"French\"],\n", + " value=\"English\",\n", + " label=\"Translate to:\"\n", + " )\n", + " audio_toggle = gr.Checkbox(label=\"Play audio\", value=False)\n", + " image_toggle = gr.Checkbox(label=\"Generate image for this reply\", value=True)\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + " \n", + " def add_user_message(message, history):\n", + " history.append({\"role\": \"user\", \"content\": message})\n", + " return \"\", history\n", + "\n", + " def chat_response(history, message, target_language, use_audio, generate_image):\n", + " history, image = chat(history, message, target_language, use_audio, generate_image)\n", + " return history, image\n", + "\n", + " entry.submit(\n", + " add_user_message,\n", + " inputs=[entry, chatbot],\n", + " outputs=[entry, chatbot]).then(\n", + " chat_response,\n", + " inputs=[chatbot, entry, target_lang_dropdown, audio_toggle, image_toggle],\n", + " outputs=[chatbot, image_output]\n", + " )\n", + "\n", + " clear.click(lambda: (None, None, None), inputs=None, outputs=[chatbot, image_output, entry], queue=False)\n", + "\n", + "ui.launch(inbrowser=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f0359c29-22aa-4156-9afa-8c63c02ca747", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week2/community-contributions/wk2-day1-monty-python-arg.py b/week2/community-contributions/wk2-day1-monty-python-arg.py new file mode 100644 index 0000000..756ddef --- /dev/null +++ b/week2/community-contributions/wk2-day1-monty-python-arg.py @@ -0,0 +1,73 @@ +#!/usr/binpython3 + +import os +from dotenv import load_dotenv +from openai import OpenAI +import anthropic + +gpt_messages = ["Hi there! is this the right room for an argument?"] +claude_messages = ["No it is not"] + +def load_api_keys(): + # Load environment variables in a file called .env + load_dotenv(override=True) + openai_api_key = os.getenv('OPENAI_API_KEY') + anthropic_api_key = os.getenv('ANTHROPIC_API_KEY') + + # Check the key + if not openai_api_key: + return "Error: No OpenAI API key was found!" + elif not anthropic_api_key: + return "Error: No Anthropic API key was found!" + else: + return "API keys found!" + +def call_gpt(openai): + gpt_model = "gpt-4.1-mini" + gpt_system = "You are a patient visiting the Argument Clinic from the famous Monty Python sketch. \ + You are very eager to have a real arguement and will quickly be irritated if someone merely contradicts you." + messages = [{"role": "system", "content": gpt_system}] + for gpt, claude in zip(gpt_messages, claude_messages): + messages.append({"role": "assistant", "content": gpt}) + messages.append({"role": "user", "content": claude}) + completion = openai.chat.completions.create( + model=gpt_model, + messages=messages + ) + return completion.choices[0].message.content + +def call_claude(claude): + claude_model = "claude-3-5-haiku-latest" + claude_system = "You are a proffesional arguer at the Argument Clinic from the famous Monty Python sketch. \ + You love to contradict whatever the person talking to you is saying." + messages = [] + for gpt, claude_message in zip(gpt_messages, claude_messages): + messages.append({"role": "user", "content": gpt}) + messages.append({"role": "assistant", "content": claude_message}) + messages.append({"role": "user", "content": gpt_messages[-1]}) + message = claude.messages.create( + model=claude_model, + system=claude_system, + messages=messages, + max_tokens=500 + ) + return message.content[0].text + +def main(): + load_api_keys() + openai = OpenAI() + claude = anthropic.Anthropic() + + print(f"GPT:\n{gpt_messages[0]}\n") + print(f"Claude:\n{claude_messages[0]}\n") + + for i in range(5): + gpt_next = call_gpt(openai) + print(f"GPT:\n{gpt_next}\n") + gpt_messages.append(gpt_next) + claude_next = call_claude(claude) + print(f"Claude:\n{claude_next}\n") + claude_messages.append(claude_next) + +if __name__ == "__main__": + main() diff --git a/week3/community-contributions/AI_Minute_Taker.ipynb b/week3/community-contributions/AI_Minute_Taker.ipynb new file mode 100644 index 0000000..d189aef --- /dev/null +++ b/week3/community-contributions/AI_Minute_Taker.ipynb @@ -0,0 +1,186 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "provenance": [], + "gpuType": "T4" + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "language_info": { + "name": "python" + }, + "accelerator": "GPU" + }, + "cells": [ + { + "cell_type": "markdown", + "source": [ + "# Create meeting minutes from an Audio file\n", + "For this project, the UI allows you to either upload meeting minutes, or record something of your own!" + ], + "metadata": { + "id": "MYOLn_FzYAF4" + } + }, + { + "cell_type": "code", + "source": [ + "# --- Install deps ---\n", + "!pip install -q gradio torch==2.5.1+cu124 torchvision==0.20.1+cu124 torchaudio==2.5.1+cu124 --index-url https://download.pytorch.org/whl/cu124\n", + "!pip install -q requests bitsandbytes==0.46.0 transformers==4.48.3 accelerate==1.3.0 openai" + ], + "metadata": { + "id": "M01YO75ITfXF" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# --- Imports ---\n", + "import gradio as gr\n", + "import torch\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig\n", + "from openai import OpenAI\n", + "from huggingface_hub import login\n", + "from google.colab import userdata\n", + "from google.colab import drive\n", + "import os" + ], + "metadata": { + "id": "DGE8_oAwZJBo" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# --- Constants ---\n", + "AUDIO_MODEL = \"whisper-1\"\n", + "LLAMA = \"meta-llama/Meta-Llama-3.1-8B-Instruct\"" + ], + "metadata": { + "id": "JPu-aNxDTmDi" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# --- Auth ---\n", + "# assumes Colab userdata or your own env vars\n", + "hf_token = userdata.get('HF_TOKEN')\n", + "login(hf_token, add_to_git_credential=True)" + ], + "metadata": { + "id": "JfWUrEVJTmET" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "openai_api_key = userdata.get('OPENAI_API_KEY')\n", + "openai = OpenAI(api_key=openai_api_key)" + ], + "metadata": { + "id": "AiUtJ0mjTpVE" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# --- Model setup ---\n", + "quant_config = BitsAndBytesConfig(\n", + " load_in_4bit=True,\n", + " bnb_4bit_use_double_quant=True,\n", + " bnb_4bit_compute_dtype=torch.bfloat16,\n", + " bnb_4bit_quant_type=\"nf4\"\n", + ")\n", + "\n", + "tokenizer = AutoTokenizer.from_pretrained(LLAMA)\n", + "tokenizer.pad_token = tokenizer.eos_token\n", + "model = AutoModelForCausalLM.from_pretrained(\n", + " LLAMA, device_map=\"auto\", quantization_config=quant_config\n", + ")" + ], + "metadata": { + "id": "hMb4dggMW2s5" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "XTEW5qAwRN4Y" + }, + "outputs": [], + "source": [ + "# --- Processing function ---\n", + "def process_meeting(audio_file):\n", + " # Step 1: Transcribe\n", + " with open(audio_file, \"rb\") as f:\n", + " transcription = openai.audio.transcriptions.create(\n", + " model=AUDIO_MODEL, file=f, response_format=\"text\"\n", + " )\n", + "\n", + " # Step 2: Prepare prompt\n", + " system_message = (\n", + " \"You are an assistant that produces minutes of meetings from transcripts, \"\n", + " \"with summary, key discussion points, takeaways and action items with owners, \"\n", + " \"in markdown.\"\n", + " )\n", + " user_prompt = (\n", + " f\"Below is an extract transcript of a meeting. Please write minutes in markdown, \"\n", + " f\"including a summary with attendees, location and date; discussion points; \"\n", + " f\"takeaways; and action items with owners.\\n{transcription}\"\n", + " )\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + "\n", + " # Step 3: Run through LLaMA\n", + " inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + " streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)\n", + " outputs = model.generate(inputs, max_new_tokens=2000)\n", + "\n", + " response = tokenizer.decode(outputs[0], skip_special_tokens=True)\n", + " return response\n", + "\n", + "# --- Gradio UI ---\n", + "with gr.Blocks() as demo:\n", + " gr.Markdown(\"## 📝 Meeting Minutes Generator\\nUpload an audio file and get structured meeting minutes.\")\n", + " with gr.Row():\n", + " audio_in = gr.Audio(type=\"filepath\", label=\"Upload Meeting Audio\")\n", + " btn = gr.Button(\"Generate Minutes\")\n", + " md_out = gr.Markdown()\n", + "\n", + " btn.click(fn=process_meeting, inputs=audio_in, outputs=md_out)" + ] + }, + { + "cell_type": "code", + "source": [ + "demo.launch()" + ], + "metadata": { + "id": "Yh4-imrmY8MH" + }, + "execution_count": null, + "outputs": [] + } + ] +} \ No newline at end of file diff --git a/week3/community-contributions/rwothoromo/week3day5assignment.ipynb b/week3/community-contributions/rwothoromo/week3day5assignment.ipynb new file mode 100644 index 0000000..a42e611 --- /dev/null +++ b/week3/community-contributions/rwothoromo/week3day5assignment.ipynb @@ -0,0 +1,249 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "18b82c6b-10dc-4d94-b8dc-592ff011ce2b", + "metadata": {}, + "source": [ + "# Meeting minutes creator\n", + "\n", + "In this colab, we make a meeting minutes program.\n", + "\n", + "It includes useful code to connect your Google Drive to your colab.\n", + "\n", + "Upload your own audio to make this work!!\n", + "\n", + "https://colab.research.google.com/drive/13wR4Blz3Ot_x0GOpflmvvFffm5XU3Kct?usp=sharing\n", + "\n", + "This should run nicely on a low-cost or free T4 box.\n", + "\n", + "## **Assignment:**\n", + "Put Everything into a nice Gradio UI (similar to last week)\n", + "Input file name of audio to process.\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e9289ba7-200c-43a9-b67a-c5ce826c9537", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import re, requests, json, tempfile, gradio as gr, torch, os\n", + "from bs4 import BeautifulSoup\n", + "from IPython.display import Markdown, display, update_display\n", + "from google.colab import drive, userdata\n", + "from huggingface_hub import login\n", + "from openai import OpenAI\n", + "from pydub import AudioSegment\n", + "from pydub.playback import play\n", + "from io import BytesIO\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig\n", + "\n", + "# Sign in to HuggingFace Hub\n", + "hf_token = userdata.get('HF_TOKEN')\n", + "login(hf_token, add_to_git_credential=True)\n", + "\n", + "# Sign in to OpenAI using Secrets in Colab\n", + "openai_api_key = userdata.get('OPENAI_API_KEY')\n", + "\n", + "# Initialize client\n", + "try:\n", + " openai = OpenAI(api_key=openai_api_key)\n", + "except Exception as e:\n", + " openai = None\n", + " print(f\"OpenAI client not initialized: {e}\")\n", + "\n", + "# Constants\n", + "AUDIO_MODEL = \"whisper-1\"\n", + "LLAMA = \"meta-llama/Meta-Llama-3.1-8B-Instruct\"\n", + "\n", + "# Google Drive\n", + "drive.mount(\"/content/drive\")\n", + "\n", + "# Local LLM setup (Llama 3.1)\n", + "try:\n", + " quant_config = BitsAndBytesConfig(\n", + " load_in_4bit=True,\n", + " bnb_4bit_use_double_quant=True,\n", + " bnb_4bit_compute_dtype=torch.bfloat16,\n", + " bnb_4bit_quant_type=\"nf4\"\n", + " )\n", + " tokenizer = AutoTokenizer.from_pretrained(LLAMA)\n", + "\n", + " # Set the pad token to the end-of-sequence token for generation\n", + " tokenizer.pad_token = tokenizer.eos_token\n", + "\n", + " model = AutoModelForCausalLM.from_pretrained(LLAMA, device_map=\"auto\", quantization_config=quant_config)\n", + " # model = AutoModelForCausalLM.from_pretrained(LLAMA, device_map=\"auto\", torch_dtype=torch.bfloat16, quantization_config=quant_config, trust_remote_code=True)\n", + "\n", + " model.eval() # Set model to evaluation mode\n", + "except Exception as e:\n", + " # If the local model fails to load, set variables to None\n", + " model = None\n", + " tokenizer = None\n", + " print(f\"Failed to load local model: {e}\")\n", + "\n", + "# Updated function to handle audio transcription\n", + "def transcribe_audio(audio_file):\n", + " \"\"\"\n", + " Transcribes an audio file to text using OpenAI's Whisper model.\n", + " Handles both local file paths and mounted Google Drive file paths.\n", + " \"\"\"\n", + " if not openai:\n", + " return \"OpenAI client not initialized. Please check your API key.\"\n", + "\n", + " if audio_file is None:\n", + " return \"No audio input provided.\"\n", + "\n", + " # Check if the file exists before attempting to open it\n", + " # Construct the expected path in Google Drive\n", + " # If the input is from the microphone, it will be a temporary file path\n", + " # If the input is from the textbox, it could be a full path or just a filename\n", + " if audio_file.startswith(\"/content/drive/MyDrive/llms/\"):\n", + " file_path_to_open = audio_file\n", + " else:\n", + " # Assume it's either a local path or just a filename in MyDrive/llms\n", + " # We'll prioritize checking MyDrive/llms first\n", + " gdrive_path_attempt = os.path.join(\"/content/drive/MyDrive/llms\", os.path.basename(audio_file))\n", + " if os.path.exists(gdrive_path_attempt):\n", + " file_path_to_open = gdrive_path_attempt\n", + " elif os.path.exists(audio_file):\n", + " file_path_to_open = audio_file\n", + " else:\n", + " return f\"File not found: {audio_file}. Please ensure the file exists in your Google Drive at /content/drive/MyDrive/llms/ or is a valid local path.\"\n", + "\n", + "\n", + " if not os.path.exists(file_path_to_open):\n", + " return f\"File not found: {file_path_to_open}. Please ensure the file exists.\"\n", + "\n", + "\n", + " try:\n", + " with open(file_path_to_open, \"rb\") as f:\n", + " transcription = openai.audio.transcriptions.create(\n", + " model=AUDIO_MODEL,\n", + " file=f,\n", + " response_format=\"text\"\n", + " )\n", + " return transcription\n", + " except Exception as e:\n", + " return f\"An error occurred during transcription: {e}\"\n", + "\n", + "def generate_minutes(transcription):\n", + " \"\"\"\n", + " Generates meeting minutes from a transcript using a local Llama model.\n", + " Format the input, generate a response, and return the complete text string.\n", + " \"\"\"\n", + " # Check if the local model and tokenizer were successfully loaded\n", + " if not model or not tokenizer:\n", + " return \"Local Llama model not loaded. Check model paths and hardware compatibility.\"\n", + "\n", + " system_message = \"You are an assistant that produces minutes of meetings from transcripts, with summary, key discussion points, takeaways and action items with owners, in markdown.\"\n", + " user_prompt = f\"Below is an extract transcript of an Audio recording. Please write minutes in markdown, including a summary with attendees, location and date; discussion points; takeaways; and action items with owners.\\n{transcription}\"\n", + "\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + "\n", + " try:\n", + " # Apply the chat template to format the messages for the model\n", + " inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + "\n", + " # Generate the output. max_new_tokens controls the length of the generated text.\n", + " outputs = model.generate(inputs, max_new_tokens=2000)\n", + "\n", + " # Decode only the new tokens generated by the model (not the input tokens) to a human-readable string\n", + " response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)\n", + "\n", + " # The model's response will contain the full conversation.\n", + " # Extract only the assistant's part!\n", + " assistant_start = \"<|eot_id|><|start_header_id|>assistant<|end_header_id|>\\n\\n\"\n", + " if assistant_start in response_text:\n", + " response_text = response_text.split(assistant_start)[-1]\n", + "\n", + " return response_text\n", + "\n", + " except Exception as e:\n", + " return f\"An error occurred during local model generation: {e}\"\n", + "\n", + "# Gradio UI components\n", + "with gr.Blocks() as ui:\n", + " gr.Markdown(\"# Meeting Minutes Generator\")\n", + " with gr.Row():\n", + " chatbot = gr.Chatbot(height=500, label=\"AI Assistant\")\n", + " with gr.Row():\n", + " entry = gr.Textbox(label=\"Provide the filename or path of the audio file to transcribe:\", scale=4)\n", + " submit_btn = gr.Button(\"Generate Minutes\", scale=1)\n", + " with gr.Row():\n", + " audio_input = gr.Audio(sources=[\"microphone\"], type=\"filepath\", label=\"Or speak to our AI Assistant to transcribe\", scale=4)\n", + " submit_audio_btn = gr.Button(\"Transcribe Audio\", scale=1)\n", + "\n", + " with gr.Row():\n", + " clear = gr.Button(\"Clear\")\n", + "\n", + " def process_file_and_generate(file_path, history):\n", + " transcribed_text = transcribe_audio(file_path)\n", + " minutes = generate_minutes(transcribed_text)\n", + " new_history = history + [[f\"Transcription of '{os.path.basename(file_path)}':\\n{transcribed_text}\", minutes]]\n", + " return new_history\n", + "\n", + " def process_audio_and_generate(audio_file, history):\n", + " transcribed_text = transcribe_audio(audio_file)\n", + " minutes = generate_minutes(transcribed_text)\n", + " new_history = history + [[f\"Transcription of your recording:\\n{transcribed_text}\", minutes]]\n", + " return new_history\n", + "\n", + "\n", + " submit_btn.click(\n", + " process_file_and_generate,\n", + " inputs=[entry, chatbot],\n", + " outputs=[chatbot],\n", + " queue=False\n", + " )\n", + "\n", + " submit_audio_btn.click(\n", + " process_audio_and_generate,\n", + " inputs=[audio_input, chatbot],\n", + " outputs=[chatbot],\n", + " queue=False\n", + " )\n", + "\n", + " clear.click(lambda: None, inputs=None, outputs=[chatbot], queue=False)\n", + "\n", + "ui.launch(inbrowser=True, debug=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cd2020d3", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week3/community-contributions/rwothoromo/week3day5task.ipynb b/week3/community-contributions/rwothoromo/week3day5task.ipynb new file mode 100644 index 0000000..7da7365 --- /dev/null +++ b/week3/community-contributions/rwothoromo/week3day5task.ipynb @@ -0,0 +1,226 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "18b82c6b-10dc-4d94-b8dc-592ff011ce2b", + "metadata": {}, + "source": [ + "# Meeting minutes creator\n", + "\n", + "https://colab.research.google.com/drive/13wR4Blz3Ot_x0GOpflmvvFffm5XU3Kct?usp=sharing\n", + "\n", + "## **Week 3 task.**\n", + "Create your own tool that generates synthetic data/test data. Input the type of dataset or products or job postings, etc. and let the tool dream up various data samples.\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e9289ba7-200c-43a9-b67a-c5ce826c9537", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "import gradio as gr, requests, json, time, os, torch\n", + "from transformers import pipeline, set_seed\n", + "from functools import partial\n", + "from openai import OpenAI, APIError, AuthenticationError\n", + "from google.colab import drive, userdata\n", + "from huggingface_hub import login\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig\n", + "\n", + "# Sample user_prompt = \"a list of student profiles with full name, email, course studied, and GPA for each of 6 semesters, and a CGPA for the 6 semesters\"\n", + "\n", + "# Sign in to HuggingFace Hub\n", + "hf_token = userdata.get('HF_TOKEN')\n", + "login(hf_token, add_to_git_credential=True)\n", + "\n", + "# Sign in to OpenAI using Secrets in Colab\n", + "openai_api_key = userdata.get('OPENAI_API_KEY')\n", + "\n", + "# Initialize client\n", + "try:\n", + " openai = OpenAI(api_key=openai_api_key)\n", + "except Exception as e:\n", + " openai = None\n", + " print(f\"OpenAI client not initialized: {e}\")\n", + "\n", + "# Constants\n", + "GPT_MODEL = \"gpt-3.5-turbo\"\n", + "\n", + "# Local Llama Model Setup\n", + "# Loads a Llama model from Hugging Face for local inference.\n", + "# Note: This requires a powerful GPU and specific library installations (e.g., bitsandbytes, accelerate).\n", + "LLAMA_MODEL = \"meta-llama/Meta-Llama-3.1-8B-Instruct\"\n", + "\n", + "try:\n", + " # Set up quantization config for efficient memory usage.\n", + " # This loads the model in 4-bit precision, significantly reducing VRAM requirements.\n", + " quant_config = BitsAndBytesConfig(\n", + " load_in_4bit=True,\n", + " bnb_4bit_use_double_quant=True,\n", + " bnb_4bit_compute_dtype=torch.bfloat16,\n", + " bnb_4bit_quant_type=\"nf4\"\n", + " )\n", + "\n", + " # Load the tokenizer and model.\n", + " tokenizer = AutoTokenizer.from_pretrained(LLAMA_MODEL)\n", + " model = AutoModelForCausalLM.from_pretrained(\n", + " LLAMA_MODEL, \n", + " device_map=\"auto\", \n", + " quantization_config=quant_config,\n", + " trust_remote_code=True\n", + " )\n", + " \n", + " # Set the model to evaluation mode for inference.\n", + " model.eval()\n", + "\n", + "except Exception as e:\n", + " model = None\n", + " tokenizer = None\n", + " print(f\"Failed to load local Llama model: {e}\")\n", + "\n", + "\n", + "def generate_with_llama(user_prompt: str, num_samples: int = 5):\n", + " \"\"\"\n", + " Generates synthetic data using a local Llama model.\n", + " Return a JSON string.\n", + " \"\"\"\n", + " if not model or not tokenizer:\n", + " return json.dumps({\"error\": \"Llama model not loaded. Check model paths and hardware compatibility.\"}, indent=2)\n", + "\n", + " # Llama 3.1 uses a specific chat template for conversation formatting.\n", + " messages = [\n", + " {\"role\": \"system\", \"content\": f\"You are a data generation assistant. Generate a JSON array of exactly {num_samples} objects based on the user's request. The output must be valid JSON only, without any other text or formatting.\"},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ]\n", + "\n", + " try:\n", + " inputs = tokenizer.apply_chat_template(messages, return_tensors=\"pt\").to(\"cuda\")\n", + "\n", + " outputs = model.generate(inputs, max_new_tokens=2000, do_sample=True, top_p=0.9, temperature=0.7)\n", + "\n", + " # Decode the generated tokens.\n", + " response_text = tokenizer.decode(outputs[0], skip_special_tokens=True)\n", + "\n", + " # Extract only the assistant's part from the complete chat history.\n", + " assistant_start = \"<|eot_id|><|start_header_id|>assistant<|end_header_id|>\\n\\n\"\n", + " if assistant_start in response_text:\n", + " response_text = response_text.split(assistant_start)[-1]\n", + " \n", + " # Parse the JSON and return it.\n", + " parsed_json = json.loads(response_text)\n", + " return json.dumps(parsed_json, indent=2)\n", + "\n", + " except Exception as e:\n", + " return json.dumps({\"error\": f\"An error occurred during local model generation: {e}\"}, indent=2)\n", + "\n", + "\n", + "\n", + "def generate_with_gpt(user_prompt: str, num_samples: int = 5):\n", + " \"\"\"\n", + " Generates synthetic data using OpenAI's GPT.\n", + " Return a JSON string.\n", + " \"\"\"\n", + " if not openai:\n", + " return json.dumps({\"error\": \"OpenAI client not initialized. Please check your API key.\"}, indent=2)\n", + "\n", + " try:\n", + " response = openai.chat.completions.create(\n", + " model=GPT_MODEL,\n", + " messages=[\n", + " {\"role\": \"system\", \"content\": f\"You are a data generation assistant. Generate a JSON array of exactly {num_samples} objects based on the user's request. The output must be valid JSON only, without any other text or formatting.\"},\n", + " {\"role\": \"user\", \"content\": user_prompt}\n", + " ],\n", + " response_format={\"type\": \"json_object\"}\n", + " )\n", + " \n", + " json_text = response.choices[0].message.content\n", + " return json_text\n", + " except APIError as e:\n", + " return json.dumps({\"error\": f\"Error from OpenAI API: {e.body}\"}, indent=2)\n", + " except Exception as e:\n", + " return json.dumps({\"error\": f\"An unexpected error occurred: {e}\"}, indent=2)\n", + "\n", + "\n", + "def generate_data(user_prompt, model_choice):\n", + " \"\"\"\n", + " Wrapper function that calls the appropriate generation function based on model choice.\n", + " \"\"\"\n", + " if not user_prompt:\n", + " return json.dumps({\"error\": \"Please provide a description for the data.\"}, indent=2)\n", + "\n", + " if model_choice == f\"Hugging Face ({LLAMA_MODEL})\":\n", + " return generate_with_llama(user_prompt)\n", + " elif model_choice == f\"OpenAI ({GPT_MODEL})\":\n", + " return generate_with_gpt(user_prompt)\n", + " else:\n", + " return json.dumps({\"error\": \"Invalid model choice.\"}, indent=2)\n", + "\n", + "# Gradio UI\n", + "with gr.Blocks(theme=gr.themes.Soft(), title=\"Synthetic Data Generator\") as ui:\n", + " gr.Markdown(\"# Synthetic Data Generator\")\n", + " gr.Markdown(\"Describe the type of data you need, select a model, and click 'Generate'.\")\n", + "\n", + " with gr.Row():\n", + " with gr.Column(scale=3):\n", + " data_prompt = gr.Textbox(\n", + " lines=5,\n", + " label=\"Data Prompt\",\n", + " placeholder=\"e.g., a list of customer profiles with name, email, and a favorite product\"\n", + " )\n", + " \n", + " with gr.Column(scale=1):\n", + " model_choice = gr.Radio(\n", + " [f\"Hugging Face ({LLAMA_MODEL})\", f\"OpenAI ({GPT_MODEL})\"],\n", + " label=\"Choose a Model\",\n", + " value=f\"Hugging Face ({LLAMA_MODEL})\"\n", + " )\n", + " \n", + " generate_btn = gr.Button(\"Generate Data\")\n", + " \n", + " with gr.Row():\n", + " output_json = gr.JSON(label=\"Generated Data\")\n", + " \n", + " # Click trigger\n", + " generate_btn.click(\n", + " fn=generate_data,\n", + " inputs=[data_prompt, model_choice],\n", + " outputs=output_json\n", + " )\n", + "\n", + "ui.launch(inbrowser=True, debug=True)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cd2020d3", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.7" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week3/community-contributions/week3_exercise_by_abrar.ipynb b/week3/community-contributions/week3_exercise_by_abrar.ipynb new file mode 100644 index 0000000..6339e85 --- /dev/null +++ b/week3/community-contributions/week3_exercise_by_abrar.ipynb @@ -0,0 +1,49 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "61f56afc-bc15-46a4-8eb1-d940c332cf52", + "metadata": {}, + "source": [ + "# Synthetic Data Studio\n", + "\n", + "It includes useful code to connect your Google Drive to your Colab and Hugging Face Model. \n", + "To use the Huggingface Model, you first need to create an account on HuggingFace and then send a request to the required model. \n", + "\n", + "\n", + "https://colab.research.google.com/drive/1W3JPe0F6su8sNpfdlp2uqXqWZChevHYa?usp=sharing\n", + "\n", + "This should run nicely on a low-cost or free T4 box." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e9289ba7-200c-43a9-b67a-c5ce826c9537", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week4/community-contributions/ai_stock_trading/README.md b/week4/community-contributions/ai_stock_trading/README.md new file mode 100644 index 0000000..95abada --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/README.md @@ -0,0 +1,325 @@ +# 📈 AI Stock Trading & Sharia Compliance Platform + +A comprehensive **Streamlit-based** web application that provides AI-powered stock analysis with Islamic Sharia compliance assessment. This professional-grade platform combines real-time financial data from USA and Egyptian markets, advanced technical analysis, and institutional-quality AI-driven insights to help users make informed investment decisions while adhering to Islamic finance principles. + +## 📸 Application Screenshots + +### Home View +![Home View](screenshots/home.png) +*Main application interface with market selection and stock input* + +### Chat Interface +![Chat Interface](screenshots/chat.png) +*Interactive chat for trading advice, Sharia compliance, and stock analysis* + +### Dashboard View +![Dashboard View](screenshots/dashboard.png) +*Comprehensive dashboard with KPIs, charts, and real-time metrics* + +## 🎯 Key Features + +### 📊 **Comprehensive Stock Analysis** +- Real-time data fetching from multiple markets (USA, Egypt) +- Advanced technical indicators (RSI, MACD, Bollinger Bands, Moving Averages) +- Risk assessment and volatility analysis +- Performance metrics across multiple time periods + +### 🤖 **AI-Powered Trading Decisions** +- GPT-4 powered investment recommendations +- Buy/Hold/Sell signals with confidence scores +- Price targets and stop-loss suggestions +- Algorithmic + AI combined decision making + +### ☪️ **Sharia Compliance Checking** +- Islamic finance principles assessment +- Halal/Haram rulings with detailed reasoning +- Business activity and financial ratio screening +- Alternative investment suggestions + +### 💬 **Natural Language Interface** +- Interactive chat interface for stock discussions +- Ask questions in plain English +- Context-aware responses about selected stocks +- Quick action buttons for common queries + +### 📈 **Interactive Dashboards** +- Comprehensive metrics dashboard +- Multiple chart types (Price, Performance, Risk, Trading Signals) +- Real-time data visualization with Plotly +- Exportable analysis reports +- Real-time price charts with volume data +- Professional matplotlib-based visualizations +- Price statistics and performance metrics +- Responsive chart interface + +### 🖥️ **Professional Interface** +- Clean, modern Streamlit web interface +- Multi-market support (USA & Egyptian stocks) +- Interactive chat interface with context awareness +- Real-time KPI dashboard with currency formatting +- Quick action buttons for common analysis tasks + +## 🚀 Quick Start + +### Prerequisites + +Ensure you have Python 3.8+ installed on your system. + +### Installation + +1. **Clone or download this project** +```bash +git clone +cd ai_stock_trading +``` + +2. **Install dependencies** +```bash +pip install -r requirements.txt +``` + +3. **Set up environment variables** +Create a `.env` file in the project root: +```bash +OPENAI_API_KEY=your-api-key-here +``` + +### Running the Application + +1. **Launch the Streamlit app** +```bash +streamlit run main_app.py +``` + +2. **Access the web interface** at `http://localhost:8501` + +3. **Select your market** (USA or Egypt) from the sidebar + +4. **Enter a stock symbol** and start analyzing! + +## 📖 How to Use + +1. **Select Market**: Choose between USA or Egypt from the sidebar +2. **Enter Stock Symbol**: Input a ticker (e.g., AAPL for USA, ABUK.CA for Egypt) +3. **View Dashboard**: See real-time KPIs, price charts, and key metrics +4. **Use Chat Interface**: Ask questions or request specific analysis: + - "Give me trading advice for AAPL" + - "Is this stock Sharia compliant?" + - "What's the price target?" +5. **Review Professional Analysis**: + - **Trading Recommendations**: Institutional-grade BUY/HOLD/SELL advice + - **Sharia Compliance**: Comprehensive Islamic finance screening + - **Technical Analysis**: Advanced indicators and risk assessment + +### Example Tickers to Try + +#### USA Market +| Ticker | Company | Sector | Expected Sharia Status | +|--------|---------|--------|-----------------------| +| **AAPL** | Apple Inc. | Technology | ✅ Likely Halal | +| **MSFT** | Microsoft Corp. | Technology | ✅ Likely Halal | +| **GOOGL** | Alphabet Inc. | Technology | ✅ Likely Halal | +| **JNJ** | Johnson & Johnson | Healthcare | ✅ Likely Halal | +| **BAC** | Bank of America | Banking | ❌ Likely Haram | +| **JPM** | JPMorgan Chase | Banking | ❌ Likely Haram | + +#### Egypt Market +| Ticker | Company | Sector | Expected Sharia Status | +|--------|---------|--------|-----------------------| +| **ABUK.CA** | Abu Qir Fertilizers | Industrial | ✅ Likely Halal | +| **ETEL.CA** | Egyptian Telecom | Telecom | ✅ Likely Halal | +| **HRHO.CA** | Hassan Allam Holding | Construction | ✅ Likely Halal | +| **CIB.CA** | Commercial Intl Bank | Banking | ❌ Likely Haram | + +## 🔧 Technical Implementation + +### Modular Architecture + +The platform is built with a clean, modular architecture using separate tool modules: + +#### 1. **Stock Fetching Module** (`tools/fetching.py`) +- **Multi-Market Support**: USA (75+ stocks) and Egypt (50+ stocks) with proper currency handling +- **Real-Time Data**: Uses yfinance API with robust error handling +- **Currency Formatting**: Automatic USD/EGP formatting based on market +- **Stock Info Enrichment**: Company details, market cap, sector classification + +#### 2. **Technical Analysis Module** (`tools/analysis.py`) +- **Advanced Indicators**: RSI, MACD, Bollinger Bands, Moving Averages +- **Risk Metrics**: Volatility analysis, Sharpe ratio, maximum drawdown +- **Performance Analysis**: Multi-timeframe returns and trend analysis +- **Professional Calculations**: Annualized metrics and statistical analysis + +#### 3. **Trading Decisions Module** (`tools/trading_decisions.py`) +- **Institutional-Grade AI**: Senior analyst persona with 15+ years experience +- **Professional Standards**: BUY/HOLD/SELL with confidence, price targets, stop-loss +- **Risk Management**: Risk-reward ratios, time horizons, risk assessment +- **Robust JSON Parsing**: Handles malformed AI responses with fallback logic + +#### 4. **Sharia Compliance Module** (`tools/sharia_compliance.py`) +- **Comprehensive Screening**: Business activities, financial ratios, trading practices +- **AAOIFI Standards**: Debt-to-assets < 33%, interest income < 5% +- **Prohibited Activities**: 50+ categories including banking, gambling, alcohol +- **User-Triggered Analysis**: Only shows when specifically requested + +#### 5. **Charting Module** (`tools/charting.py`) +- **Professional Visualizations**: Plotly-based interactive charts +- **Multiple Chart Types**: Price, volume, technical indicators +- **Responsive Design**: Mobile-friendly chart rendering +- **Export Capabilities**: PNG/HTML export functionality + +#### 6. **Main Application** (`main_app.py`) +- **Streamlit Interface**: Modern, responsive web application +- **Chat Integration**: Context-aware conversational interface +- **Real-Time KPIs**: Live dashboard with key metrics +- **Session Management**: Persistent data across user interactions + +### AI Integration + +The platform leverages OpenAI's GPT-4o-mini with specialized prompts: + +#### Trading Analysis Prompts +- **Senior Analyst Persona**: 15+ years institutional experience +- **Professional Standards**: Risk-reward ratios, logical price targets +- **Structured Output**: JSON format with validation and error handling +- **Technical Focus**: Based on RSI, MACD, trend analysis, volume patterns + +#### Sharia Compliance Prompts +- **Islamic Scholar Approach**: Follows AAOIFI and DSN standards +- **Comprehensive Screening**: Business activities, financial ratios, trading practices +- **Scholarly Reasoning**: Detailed justification with Islamic finance principles +- **Confidence Scoring**: Quantified certainty levels for rulings + +## 📊 Sample Analysis Output + +### Trade Recommendation Example +``` +RECOMMENDATION: BUY + +Based on the analysis of AAPL: +• 1Y return of +15.2% shows strong performance +• Volatility of 24.3% indicates manageable risk +• Recent 1M return of +5.8% shows positive momentum +• Strong volume indicates healthy trading activity + +Key factors supporting BUY decision: +- Consistent positive returns across timeframes +- Volatility within acceptable range for tech stocks +- Strong market position and fundamentals +``` + +### Sharia Assessment Example +```json +{ + "ruling": "HALAL", + "confidence": 85, + "justification": "Apple Inc. primarily operates in technology hardware and software, which are permissible under Islamic law. The company's main revenue sources (iPhone, Mac, services) do not involve prohibited activities such as gambling, alcohol, or interest-based banking." +} +``` + +## ⚠️ Important Disclaimers + +### Financial Disclaimer +- **This tool is for educational purposes only** +- **Not professional financial advice** +- **Past performance does not guarantee future results** +- **Consult qualified financial advisors before making investment decisions** + +### Sharia Compliance Disclaimer +- **Consult qualified Islamic scholars for authoritative rulings** +- **AI assessments are preliminary and may have limitations** +- **Consider multiple sources for Sharia compliance verification** +- **Individual scholarly interpretations may vary** + +### Technical Limitations +- **Data accuracy depends on yfinance API availability** +- **OpenAI API calls consume credits/tokens** +- **Network connectivity required for real-time data** +- **Analysis speed depends on API response times** + +## 🔧 Customization + +### Adding New Analysis Periods +```python +periods = ["1mo", "3mo", "6mo", "1y", "2y", "5y"] # Modify as needed +``` + +### Modifying Sharia Criteria +```python +# Update the Sharia assessment prompt with additional criteria +prompt = f""" +Additional criteria: +- Debt-to-market cap ratio analysis +- Revenue source breakdown +- ESG factors consideration +""" +``` + +### Styling the Interface +```python +demo = create_interface() +demo.launch(theme="huggingface") # Try different themes +``` + +## 📚 Dependencies + +- **yfinance**: Real-time financial data +- **openai**: AI-powered analysis +- **pandas**: Data manipulation +- **matplotlib**: Chart generation +- **gradio**: Web interface +- **requests**: HTTP requests +- **beautifulsoup4**: Web scraping +- **numpy**: Numerical computations + +## 🤝 Contributing + +Contributions are welcome! Please feel free to submit issues, feature requests, or pull requests. + +### Areas for Enhancement +- Additional technical indicators +- More sophisticated Sharia screening +- Portfolio analysis features +- Historical backtesting +- Mobile-responsive design + +### 🔮 Future Work: MCP Integration + +We plan to implement a **Model Context Protocol (MCP) layer** to make all trading tools accessible as standardized MCP tools: + +#### Planned MCP Tools: +- **`stock_fetcher`** - Real-time market data retrieval for USA/Egypt markets +- **`technical_analyzer`** - Advanced technical analysis with 20+ indicators +- **`sharia_checker`** - Islamic finance compliance screening +- **`trading_advisor`** - AI-powered institutional-grade recommendations +- **`risk_assessor`** - Portfolio risk analysis and management +- **`chart_generator`** - Professional financial visualizations + +#### Benefits of MCP Integration: +- **Standardized Interface**: Consistent tool access across different AI systems +- **Interoperability**: Easy integration with other MCP-compatible platforms +- **Scalability**: Modular architecture for adding new financial tools +- **Reusability**: Tools can be used independently or combined +- **Professional Integration**: Compatible with institutional trading platforms + +This will enable the platform to serve as a comprehensive financial analysis toolkit that can be integrated into various AI-powered trading systems and workflows. + +## 📄 License + +This project is for educational purposes. Please ensure compliance with: +- OpenAI API usage terms +- Yahoo Finance data usage policies +- Local financial regulations +- Islamic finance guidelines + +## 🙏 Acknowledgments + +- **yfinance** for providing free financial data API +- **OpenAI** for GPT-4o-mini language model +- **Gradio** for the intuitive web interface framework +- **Islamic finance scholars** for Sharia compliance frameworks + +--- + +**Made with ❤️ for the Muslim tech community and ethical investing enthusiasts** + +*"And Allah knows best" - وَاللَّهُ أَعْلَمُ* \ No newline at end of file diff --git a/week4/community-contributions/ai_stock_trading/components/__init__.py b/week4/community-contributions/ai_stock_trading/components/__init__.py new file mode 100644 index 0000000..34dac01 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/components/__init__.py @@ -0,0 +1,3 @@ +""" +UI Components for the Stock Trading Platform +""" diff --git a/week4/community-contributions/ai_stock_trading/components/chat_interface.py b/week4/community-contributions/ai_stock_trading/components/chat_interface.py new file mode 100644 index 0000000..da0c9e4 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/components/chat_interface.py @@ -0,0 +1,156 @@ +""" +Clean, professional chat interface component +""" + +import streamlit as st +from core.ai_assistant import ai_assistant +from core.data_service import data_service + + +class ChatInterface: + """Professional chat interface for stock analysis""" + + @staticmethod + def render(symbol: str, country: str): + """Render the chat interface""" + if not symbol: + st.warning("⚠️ Please select a stock from the sidebar.") + return + + # Display chat history + if 'chat_history' not in st.session_state: + st.session_state.chat_history = [] + + if st.session_state.chat_history: + for message in st.session_state.chat_history: + if message['role'] == 'user': + st.chat_message("user").write(message['content']) + else: + st.chat_message("assistant").write(message['content']) + else: + # Clean welcome message + basic_info = data_service.get_basic_info(symbol, country) + company_name = basic_info.get('company_name', symbol) + welcome_msg = f"👋 Hello! I'm your AI assistant for **{company_name} ({symbol})**. Ask me anything!" + st.chat_message("assistant").write(welcome_msg) + + # Chat input + user_input = st.chat_input("Ask about price, trends, analysis, trading recommendations...") + + if user_input: + # Add user message + st.session_state.chat_history.append({'role': 'user', 'content': user_input}) + + # Generate AI response (only loads data if tools are called) + with st.spinner("Thinking..."): + ai_response = ai_assistant.generate_response(user_input, symbol, country) + + # Add AI response + st.session_state.chat_history.append({'role': 'assistant', 'content': ai_response}) + st.rerun() + + # Quick actions (collapsed by default) + ChatInterface._render_quick_actions(symbol, country) + + @staticmethod + def _render_quick_actions(symbol: str, country: str): + """Render quick action buttons""" + with st.expander("🚀 Quick Actions", expanded=False): + col1, col2, col3, col4 = st.columns(4) + + with col1: + if st.button("📈 Price Info", use_container_width=True): + ChatInterface._add_price_info(symbol, country) + st.rerun() + + with col2: + if st.button("📊 30-Day Analysis", use_container_width=True): + ChatInterface._add_medium_term_analysis(symbol) + st.rerun() + + with col3: + if st.button("💰 Trading Rec", use_container_width=True): + ChatInterface._add_trading_recommendation(symbol, country) + st.rerun() + + with col4: + if st.button("☪️ Sharia", use_container_width=True): + ChatInterface._add_sharia_compliance(symbol, country) + st.rerun() + + @staticmethod + def _add_price_info(symbol: str, country: str): + """Add current price info to chat""" + basic_info = data_service.get_basic_info(symbol, country) + + current_price = basic_info.get('current_price', 0) + market_cap = basic_info.get('market_cap', 0) + sector = basic_info.get('sector', 'N/A') + + message = f"""📈 **Current Price Info for {symbol}:** + +💰 **Price:** ${current_price:.2f} +🏢 **Market Cap:** ${market_cap:,.0f} +🏭 **Sector:** {sector}""" + + st.session_state.chat_history.append({'role': 'assistant', 'content': message}) + + @staticmethod + def _add_medium_term_analysis(symbol: str): + """Add 30-day analysis to chat""" + analysis = data_service.get_analysis(symbol, "1mo") + + if 'error' in analysis: + message = f"❌ **30-Day Analysis:** {analysis['error']}" + else: + return_pct = analysis.get('total_return_pct', 0) + volatility = analysis.get('volatility_annualized', 0) + trend = analysis.get('trend_direction', 'neutral') + + message = f"""📊 **30-Day Analysis for {symbol}:** + +📈 **Return:** {return_pct:.2f}% +📉 **Volatility:** {volatility:.1f}% (annualized) +🎯 **Trend:** {trend.title()}""" + + st.session_state.chat_history.append({'role': 'assistant', 'content': message}) + + @staticmethod + def _add_trading_recommendation(symbol: str, country: str): + """Add trading recommendation to chat""" + trading = data_service.get_trading_recommendation(symbol, country) + + if 'error' in trading: + message = f"❌ **Trading Recommendation:** {trading['error']}" + else: + rec = trading.get('recommendation', 'HOLD') + conf = trading.get('confidence', 0.5) * 100 + reasoning = trading.get('reasoning', 'No reasoning available') + + message = f"""💰 **Trading Recommendation for {symbol}:** + +🎯 **Action:** {rec} +📊 **Confidence:** {conf:.0f}% +💭 **Reasoning:** {reasoning[:200]}...""" + + st.session_state.chat_history.append({'role': 'assistant', 'content': message}) + + @staticmethod + def _add_sharia_compliance(symbol: str, country: str): + """Add Sharia compliance to chat""" + sharia = data_service.get_sharia_compliance(symbol, country) + + if 'error' in sharia: + message = f"❌ **Sharia Compliance:** {sharia['error']}" + else: + ruling = sharia.get('ruling', 'UNCERTAIN') + conf = sharia.get('confidence', 0.5) * 100 + + status_emoji = "✅" if ruling == "HALAL" else "❌" if ruling == "HARAM" else "⚠️" + + message = f"""☪️ **Sharia Compliance for {symbol}:** + +{status_emoji} **Ruling:** {ruling} +📊 **Confidence:** {conf:.0f}%""" + + st.session_state.chat_history.append({'role': 'assistant', 'content': message}) diff --git a/week4/community-contributions/ai_stock_trading/core/__init__.py b/week4/community-contributions/ai_stock_trading/core/__init__.py new file mode 100644 index 0000000..f3d485c --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/core/__init__.py @@ -0,0 +1,3 @@ +""" +Core module for AI Stock Trading Platform +""" diff --git a/week4/community-contributions/ai_stock_trading/core/ai_assistant.py b/week4/community-contributions/ai_stock_trading/core/ai_assistant.py new file mode 100644 index 0000000..0255491 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/core/ai_assistant.py @@ -0,0 +1,275 @@ +import os +from typing import Dict, Any, List +from openai import OpenAI +from .data_service import data_service + + +class AIAssistant: + """Enhanced AI assistant with comprehensive stock analysis tools""" + + def __init__(self): + self.client = OpenAI(api_key=os.getenv('OPENAI_API_KEY')) + + def get_enhanced_tools(self) -> List[Dict[str, Any]]: + """Get comprehensive tool definitions for OpenAI function calling""" + return [ + { + "type": "function", + "function": { + "name": "get_current_price_info", + "description": "Get current price, basic metrics, and company info", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_short_term_analysis", + "description": "Get 10-day technical analysis and short-term trends", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_medium_term_analysis", + "description": "Get 30-day technical analysis and medium-term trends", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_long_term_analysis", + "description": "Get 90-day technical analysis and long-term trends", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_comprehensive_analysis", + "description": "Get full 1-year technical analysis with all indicators", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_trading_recommendation", + "description": "Get buy/hold/sell recommendation with price targets and reasoning", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "get_sharia_compliance", + "description": "Get Islamic finance compliance analysis", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + }, + { + "type": "function", + "function": { + "name": "compare_time_periods", + "description": "Compare performance across multiple time periods (10d, 30d, 90d)", + "parameters": { + "type": "object", + "properties": { + "symbol": {"type": "string", "description": "Stock symbol"} + }, + "required": ["symbol"] + } + } + } + ] + + def generate_response(self, user_input: str, symbol: str, country: str) -> str: + """Generate AI response with enhanced tool calling""" + try: + # Get basic info without heavy loading + basic_info = data_service.get_basic_info(symbol, country) + + system_msg = f"""You are a professional financial advisor assistant for {symbol}. + +IMPORTANT: Only call tools when users specifically request: +- Price information or basic metrics → get_current_price_info +- Short-term analysis (10 days) → get_short_term_analysis +- Medium-term analysis (30 days) → get_medium_term_analysis +- Long-term analysis (90 days) → get_long_term_analysis +- Comprehensive analysis (1 year) → get_comprehensive_analysis +- Trading recommendations → get_trading_recommendation +- Sharia compliance → get_sharia_compliance +- Time period comparisons → compare_time_periods + +For general questions about the company, market commentary, or basic information, respond directly without calling tools. +Keep responses concise and professional.""" + + user_msg = f"""Stock: {symbol} ({basic_info.get('company_name', 'N/A')}) +Country: {country} +Sector: {basic_info.get('sector', 'N/A')} +User Question: {user_input}""" + + response = self.client.chat.completions.create( + model="gpt-4o-mini", + messages=[ + {"role": "system", "content": system_msg}, + {"role": "user", "content": user_msg} + ], + tools=self.get_enhanced_tools(), # type: ignore + tool_choice="auto", + temperature=0.7, + max_tokens=600 + ) + + message = response.choices[0].message + + if message.tool_calls: + return self._handle_tool_calls(message.tool_calls, user_input, symbol, country) + + return message.content or "I apologize, but I couldn't generate a response." + + except Exception as e: + return f"Sorry, I encountered an error: {str(e)}" + + def _handle_tool_calls(self, tool_calls, user_input: str, symbol: str, country: str) -> str: + """Handle tool calls and generate final response""" + tool_results = [] + + for tool_call in tool_calls: + function_name = tool_call.function.name + + try: + if function_name == "get_current_price_info": + basic_info = data_service.get_basic_info(symbol, country) + current_price = basic_info.get('current_price', 0) + market_cap = basic_info.get('market_cap', 0) + tool_results.append(f"Current Price: ${current_price:.2f}, Market Cap: ${market_cap:,.0f}") + + elif function_name == "get_short_term_analysis": + analysis = data_service.get_analysis(symbol, "10d") + if 'error' not in analysis: + return_pct = analysis.get('total_return_pct', 0) + volatility = analysis.get('volatility_annualized', 0) + tool_results.append(f"10-Day Analysis: Return {return_pct:.2f}%, Volatility {volatility:.1f}%") + else: + tool_results.append("10-Day Analysis: Data unavailable") + + elif function_name == "get_medium_term_analysis": + analysis = data_service.get_analysis(symbol, "1mo") + if 'error' not in analysis: + return_pct = analysis.get('total_return_pct', 0) + trend = analysis.get('trend_direction', 'neutral') + tool_results.append(f"30-Day Analysis: Return {return_pct:.2f}%, Trend {trend}") + else: + tool_results.append("30-Day Analysis: Data unavailable") + + elif function_name == "get_long_term_analysis": + analysis = data_service.get_analysis(symbol, "3mo") + if 'error' not in analysis: + return_pct = analysis.get('total_return_pct', 0) + sharpe = analysis.get('sharpe_ratio', 0) + tool_results.append(f"90-Day Analysis: Return {return_pct:.2f}%, Sharpe {sharpe:.2f}") + else: + tool_results.append("90-Day Analysis: Data unavailable") + + elif function_name == "get_comprehensive_analysis": + analysis = data_service.get_analysis(symbol, "1y") + if 'error' not in analysis: + return_pct = analysis.get('total_return_pct', 0) + max_drawdown = analysis.get('max_drawdown', 0) + rsi = analysis.get('rsi', 50) + tool_results.append(f"1-Year Analysis: Return {return_pct:.2f}%, Max Drawdown {max_drawdown:.1f}%, RSI {rsi:.1f}") + else: + tool_results.append("1-Year Analysis: Data unavailable") + + elif function_name == "get_trading_recommendation": + trading = data_service.get_trading_recommendation(symbol, country) + if 'error' not in trading: + rec = trading.get('recommendation', 'HOLD') + conf = trading.get('confidence', 0.5) * 100 + tool_results.append(f"Trading: {rec} ({conf:.0f}% confidence)") + else: + tool_results.append("Trading: Analysis unavailable") + + elif function_name == "get_sharia_compliance": + sharia = data_service.get_sharia_compliance(symbol, country) + if 'error' not in sharia: + ruling = sharia.get('ruling', 'UNCERTAIN') + conf = sharia.get('confidence', 0.5) * 100 + tool_results.append(f"Sharia: {ruling} ({conf:.0f}% confidence)") + else: + tool_results.append("Sharia: Analysis unavailable") + + elif function_name == "compare_time_periods": + periods = ["10d", "1mo", "3mo"] + comparisons = [] + for period in periods: + analysis = data_service.get_analysis(symbol, period) + if 'error' not in analysis: + return_pct = analysis.get('total_return_pct', 0) + comparisons.append(f"{period}: {return_pct:.2f}%") + tool_results.append(f"Period Comparison: {', '.join(comparisons)}") + + except Exception as e: + tool_results.append(f"{function_name}: Error - {str(e)}") + + # Generate final response + final_response = self.client.chat.completions.create( + model="gpt-4o-mini", + messages=[ + {"role": "system", "content": "Provide a concise, professional response based on the tool results. Focus on actionable insights."}, + {"role": "user", "content": f"Question: {user_input}\n\nTool Results: {' | '.join(tool_results)}"} + ], + temperature=0.7, + max_tokens=500 + ) + + return final_response.choices[0].message.content or "I couldn't generate a response." + + +# Global instance +ai_assistant = AIAssistant() diff --git a/week4/community-contributions/ai_stock_trading/core/data_service.py b/week4/community-contributions/ai_stock_trading/core/data_service.py new file mode 100644 index 0000000..a5232bb --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/core/data_service.py @@ -0,0 +1,119 @@ +import streamlit as st +from typing import Dict, Any, Optional +from tools.fetching import stock_fetcher +from tools.analysis import stock_analyzer +from tools.trading_decisions import trading_engine +from tools.sharia_compliance import sharia_checker + + +class DataService: + """Centralized data service for efficient stock data management""" + + @staticmethod + def get_basic_info(symbol: str, country: str) -> Dict[str, Any]: + """Get only basic stock info - no heavy analysis""" + cache_key = f"{symbol}_basic" + + if cache_key not in st.session_state: + try: + stock_info = stock_fetcher.get_stock_info(symbol, country) + st.session_state[cache_key] = stock_info + except Exception as e: + st.session_state[cache_key] = { + 'company_name': symbol, + 'error': str(e) + } + + return st.session_state[cache_key] + + @staticmethod + def get_price_data(symbol: str, period: str = "1y") -> Dict[str, Any]: + """Get price data for specific period""" + cache_key = f"{symbol}_data_{period}" + + if cache_key not in st.session_state: + try: + data = stock_fetcher.fetch_stock_data(symbol, period=period) + st.session_state[cache_key] = data + except Exception as e: + st.session_state[cache_key] = None + st.error(f"Failed to load {period} data: {str(e)}") + + return st.session_state[cache_key] + + @staticmethod + def get_analysis(symbol: str, period: str = "1y") -> Dict[str, Any]: + """Get technical analysis for specific period""" + cache_key = f"{symbol}_analysis_{period}" + + if cache_key not in st.session_state: + data = DataService.get_price_data(symbol, period) + if data is not None and hasattr(data, 'empty') and not data.empty: + try: + analysis = stock_analyzer.analyze_stock(data) + analysis['period'] = period + st.session_state[cache_key] = analysis + except Exception as e: + st.session_state[cache_key] = {'error': f"Analysis failed: {str(e)}"} + else: + st.session_state[cache_key] = {'error': 'No data available'} + + return st.session_state[cache_key] + + @staticmethod + def get_trading_recommendation(symbol: str, country: str) -> Dict[str, Any]: + """Get trading recommendation""" + cache_key = f"{symbol}_trading" + + if cache_key not in st.session_state: + try: + analysis = DataService.get_analysis(symbol) + stock_info = DataService.get_basic_info(symbol, country) + + if 'error' not in analysis and 'error' not in stock_info: + trading = trading_engine.get_trading_recommendation(symbol, analysis, stock_info) + st.session_state[cache_key] = trading + else: + st.session_state[cache_key] = {'error': 'Cannot generate recommendation'} + except Exception as e: + st.session_state[cache_key] = {'error': f"Trading analysis failed: {str(e)}"} + + return st.session_state[cache_key] + + @staticmethod + def get_sharia_compliance(symbol: str, country: str) -> Dict[str, Any]: + """Get Sharia compliance analysis""" + cache_key = f"{symbol}_sharia" + + if cache_key not in st.session_state: + try: + stock_info = DataService.get_basic_info(symbol, country) + analysis = DataService.get_analysis(symbol) + + if 'error' not in stock_info: + sharia = sharia_checker.check_sharia_compliance(symbol, stock_info, analysis) + st.session_state[cache_key] = sharia + else: + st.session_state[cache_key] = {'error': 'Cannot check compliance'} + except Exception as e: + st.session_state[cache_key] = {'error': f"Sharia check failed: {str(e)}"} + + return st.session_state[cache_key] + + @staticmethod + def clear_cache(symbol: Optional[str] = None): + """Clear cached data""" + if symbol: + keys_to_remove = [key for key in st.session_state.keys() if isinstance(key, str) and key.startswith(f"{symbol}_")] + for key in keys_to_remove: + del st.session_state[key] + else: + # Clear all cache + keys_to_remove = [key for key in st.session_state.keys() + if isinstance(key, str) and ('_data_' in key or '_analysis_' in key or '_trading' in key or '_sharia' in key or '_basic' in key)] + for key in keys_to_remove: + del st.session_state[key] + + +# Global instance +data_service = DataService() diff --git a/week4/community-contributions/ai_stock_trading/main_app.py b/week4/community-contributions/ai_stock_trading/main_app.py new file mode 100644 index 0000000..a55622f --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/main_app.py @@ -0,0 +1,332 @@ +""" +Main Streamlit Application for AI Stock Trading with Sharia Compliance +""" + +import streamlit as st +import pandas as pd +import os +from dotenv import load_dotenv + +# Import our custom tools +from tools.fetching import stock_fetcher, get_available_stocks +from tools.analysis import stock_analyzer +from tools.trading_decisions import trading_engine +from tools.sharia_compliance import sharia_checker +from tools.charting import chart_generator + +# Import new modular components +from core.data_service import data_service +from core.ai_assistant import ai_assistant +from components.chat_interface import ChatInterface + +# Load environment variables +load_dotenv() + +# Page configuration +st.set_page_config( + page_title="AI Stock Trading & Sharia Compliance", + page_icon="📈", + layout="wide", + initial_sidebar_state="expanded" +) + +class StockTradingApp: + def __init__(self): + self.initialize_session_state() + self.setup_sidebar() + + def initialize_session_state(self): + if 'selected_country' not in st.session_state: + st.session_state.selected_country = 'USA' + if 'selected_stock' not in st.session_state: + st.session_state.selected_stock = None + if 'stock_data' not in st.session_state: + st.session_state.stock_data = {} + if 'chat_history' not in st.session_state: + st.session_state.chat_history = [] + if 'current_page' not in st.session_state: + st.session_state.current_page = 'home' + + def setup_sidebar(self): + with st.sidebar: + st.title("🏛️ Navigation") + + page = st.radio( + "Select Page:", + ["🏠 Home", "💬 Chat Interface", "📊 Dashboard"], + key="page_selector" + ) + + page_mapping = { + "🏠 Home": "home", + "💬 Chat Interface": "chat", + "📊 Dashboard": "dashboard" + } + st.session_state.current_page = page_mapping[page] + + st.divider() + self.render_stock_selector() + st.divider() + self.show_api_status() + + def render_stock_selector(self): + st.subheader("🌍 Stock Selection") + + countries = ['USA', 'Egypt'] + selected_country = st.selectbox( + "Select Country:", + countries, + index=countries.index(st.session_state.selected_country), + key="country_selector" + ) + + if selected_country != st.session_state.selected_country: + st.session_state.selected_country = selected_country + st.session_state.selected_stock = None + + available_stocks = get_available_stocks(selected_country) + + if available_stocks: + stock_names = list(available_stocks.keys()) + current_index = 0 + + if st.session_state.selected_stock: + current_symbol = st.session_state.selected_stock + for i, (name, symbol) in enumerate(available_stocks.items()): + if symbol == current_symbol: + current_index = i + break + + selected_stock_name = st.selectbox( + "Select Stock:", + stock_names, + index=current_index, + key="stock_selector" + ) + + selected_symbol = available_stocks[selected_stock_name] + + if selected_symbol != st.session_state.selected_stock: + st.session_state.selected_stock = selected_symbol + st.session_state.stock_data = {} + st.session_state.chat_history = [] + + if st.session_state.selected_stock: + st.success(f"Selected: {selected_stock_name} ({selected_symbol})") + else: + st.error(f"No stocks available for {selected_country}") + + def show_api_status(self): + st.subheader("API Used") + openai_key = os.getenv('OPENAI_API_KEY') + if openai_key: + st.success("✅ OpenAI Connected") + else: + st.error("❌ Not Connected") + + def run(self): + st.title("🤖 AI Stock Trading") + st.markdown("*Intelligent stock analysis with Islamic finance compliance*") + + if st.session_state.current_page == 'home': + self.render_home_page() + elif st.session_state.current_page == 'chat': + self.render_chat_page() + elif st.session_state.current_page == 'dashboard': + self.render_dashboard_page() + + def render_home_page(self): + st.header("🏠 Welcome to AI Stock Trading Platform") + + st.markdown(""" + Get intelligent stock analysis with Islamic finance compliance checking. + Select a country and stock from the sidebar to begin. + + **Key Features:** + - 📊 Real-time stock analysis with advanced indicators + - 🤖 AI-powered trading recommendations + - ☪️ Sharia compliance assessment + - 💬 Natural language chat interface + + **Supported Markets:** 🇺🇸 USA | 🇪🇬 Egypt + + *Disclaimer: For educational purposes only. Not financial advice.* + """) + + if st.session_state.selected_stock: + st.divider() + st.subheader(f"📊 Quick Analysis: {st.session_state.selected_stock}") + with st.spinner("Loading quick analysis..."): + self.show_quick_analysis() + + def show_quick_analysis(self): + symbol = st.session_state.selected_stock + country = st.session_state.selected_country + try: + data = stock_fetcher.fetch_stock_data(symbol, period="1mo") + stock_info = stock_fetcher.get_stock_info(symbol, country) + + if not data.empty: + col1, col2, col3, col4 = st.columns(4) + + current_price = data['Close'].iloc[-1] + price_change = data['Close'].iloc[-1] - data['Close'].iloc[-2] if len(data) > 1 else 0 + price_change_pct = (price_change / data['Close'].iloc[-2] * 100) if len(data) > 1 else 0 + + with col1: + formatted_price = stock_fetcher.format_price_with_currency(current_price, country) + price_change_str = f"{price_change:+.2f} ({price_change_pct:+.1f}%)" + st.metric("Current Price", formatted_price, price_change_str) + + with col2: + high_52w = stock_info.get('fifty_two_week_high', 0) + formatted_high = stock_fetcher.format_price_with_currency(high_52w, country) + st.metric("52W High", formatted_high) + + with col3: + low_52w = stock_info.get('fifty_two_week_low', 0) + formatted_low = stock_fetcher.format_price_with_currency(low_52w, country) + st.metric("52W Low", formatted_low) + + with col4: + market_cap = stock_info.get('market_cap', 0) + currency = stock_fetcher.get_market_currency(country) + if market_cap > 1e9: + if currency == 'EGP': + market_cap_str = f"{market_cap/1e9:.1f}B EGP" + else: + market_cap_str = f"${market_cap/1e9:.1f}B" + elif market_cap > 1e6: + if currency == 'EGP': + market_cap_str = f"{market_cap/1e6:.1f}M EGP" + else: + market_cap_str = f"${market_cap/1e6:.1f}M" + else: + if currency == 'EGP': + market_cap_str = f"{market_cap:,.0f} EGP" + else: + market_cap_str = f"${market_cap:,.0f}" + st.metric("Market Cap", market_cap_str) + + st.info(f"**{stock_info.get('company_name', 'N/A')}** | " + f"Sector: {stock_info.get('sector', 'N/A')} | " + f"Industry: {stock_info.get('industry', 'N/A')}") + + except Exception as e: + st.error(f"Error loading quick analysis: {str(e)}") + + def load_stock_analysis(self, symbol: str): + """Load complete analysis using data service""" + country = st.session_state.selected_country + # Pre-load all analysis components + data_service.get_analysis(symbol) + data_service.get_trading_recommendation(symbol, country) + data_service.get_sharia_compliance(symbol, country) + + def render_chat_page(self): + st.header("💬 AI Stock Analysis Chat") + + symbol = st.session_state.selected_stock + country = st.session_state.selected_country + + ChatInterface.render(symbol, country) + + def render_dashboard_page(self): + st.header("📊 Dashboard") + + if not st.session_state.selected_stock: + st.warning("⚠️ Please select a stock from the sidebar.") + return + + symbol = st.session_state.selected_stock + country = st.session_state.selected_country + + # Load data using new data service + with st.spinner("Loading dashboard data..."): + basic_info = data_service.get_basic_info(symbol, country) + data = data_service.get_price_data(symbol, "1y") + analysis = data_service.get_analysis(symbol, "1y") + trading_decision = data_service.get_trading_recommendation(symbol, country) + sharia_compliance = data_service.get_sharia_compliance(symbol, country) + + # Check if data loaded successfully + if data is None or analysis.get('error') or trading_decision.get('error'): + st.error("Failed to load dashboard data. Please try again.") + return + + # KPIs at the top + col1, col2, col3, col4, col5 = st.columns(5) + + with col1: + if data is not None and hasattr(data, 'iloc') and len(data) > 0: + current_price = data['Close'].iloc[-1] + formatted_price = stock_fetcher.format_price_with_currency(current_price, country) + st.metric("💰 Current Price", formatted_price) + else: + st.metric("💰 Current Price", "N/A") + + with col2: + total_return = analysis.get('total_return_pct', 0) + st.metric("Total Return", f"{total_return:.2f}%") + + with col3: + rec = trading_decision.get('recommendation', 'HOLD') + conf = trading_decision.get('confidence', 0.5) + if conf <= 1.0: + conf_pct = conf * 100 + else: + conf_pct = conf + st.metric("Recommendation", rec, f"{conf_pct:.0f}% confidence") + + with col4: + ruling = sharia_compliance.get('ruling', 'UNCERTAIN') + sharia_conf = sharia_compliance.get('confidence', 0.5) + if sharia_conf <= 1.0: + sharia_conf_pct = sharia_conf * 100 + else: + sharia_conf_pct = sharia_conf + st.metric("Sharia Status", ruling, f"{sharia_conf_pct:.0f}% confidence") + + with col5: + volatility = analysis.get('volatility_annualized', 0) + st.metric("Volatility", f"{volatility:.1f}%") + + + # Charts section (only if data is available) + if data is not None and hasattr(data, 'iloc') and len(data) > 0: + st.divider() + + # First row: Risk Analysis and Trading Signals + col1, col2 = st.columns(2) + + with col1: + try: + risk_fig = chart_generator.create_risk_analysis_chart(analysis, symbol) + st.plotly_chart(risk_fig, use_container_width=True) + except Exception as e: + st.error(f"Risk chart error: {str(e)}") + + with col2: + try: + signals_fig = chart_generator.create_trading_signals_chart(data, analysis, trading_decision, symbol) + st.plotly_chart(signals_fig, use_container_width=True) + except Exception as e: + st.error(f"Signals chart error: {str(e)}") + + # Second row: Price Chart (full width) + try: + price_fig = chart_generator.create_price_chart(data, symbol, analysis) + st.plotly_chart(price_fig, use_container_width=True) + except Exception as e: + st.error(f"Price chart error: {str(e)}") + else: + st.warning("📊 Charts unavailable - no price data loaded.") + + + +def main(): + app = StockTradingApp() + app.run() + +if __name__ == "__main__": + main() diff --git a/week4/community-contributions/ai_stock_trading/requirements.txt b/week4/community-contributions/ai_stock_trading/requirements.txt new file mode 100644 index 0000000..98321c1 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/requirements.txt @@ -0,0 +1,10 @@ +yfinance>=0.2.10 +openai>=1.0.0 +pandas>=1.5.0 +matplotlib>=3.5.0 +streamlit>=1.28.0 +requests>=2.28.0 +beautifulsoup4>=4.11.0 +numpy>=1.21.0 +python-dotenv>=1.0.0 +plotly>=5.15.0 \ No newline at end of file diff --git a/week4/community-contributions/ai_stock_trading/tools/__init__.py b/week4/community-contributions/ai_stock_trading/tools/__init__.py new file mode 100644 index 0000000..eee1806 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/__init__.py @@ -0,0 +1,28 @@ +""" +AI Stock Trading Tools + +This package contains all the core tools for the AI Stock Trading platform: +- fetching: Stock data fetching and market data +- analysis: Technical analysis and stock metrics +- trading_decisions: AI-powered trading recommendations +- sharia_compliance: Islamic finance compliance checking +- charting: Interactive charts and visualizations +""" + +__version__ = "1.0.0" +__author__ = "AI Stock Trading Platform" + +# Import main classes and functions for easy access +from .fetching import StockDataFetcher, stock_fetcher, fetch_stock_data, get_available_stocks +from .analysis import StockAnalyzer, stock_analyzer, analyze_stock +from .trading_decisions import TradingDecisionEngine, trading_engine, get_trading_recommendation +from .sharia_compliance import ShariaComplianceChecker, sharia_checker, check_sharia_compliance +from .charting import StockChartGenerator, chart_generator, create_price_chart + +__all__ = [ + 'StockDataFetcher', 'stock_fetcher', 'fetch_stock_data', 'get_available_stocks', + 'StockAnalyzer', 'stock_analyzer', 'analyze_stock', + 'TradingDecisionEngine', 'trading_engine', 'get_trading_recommendation', + 'ShariaComplianceChecker', 'sharia_checker', 'check_sharia_compliance', + 'StockChartGenerator', 'chart_generator', 'create_price_chart' +] diff --git a/week4/community-contributions/ai_stock_trading/tools/analysis.py b/week4/community-contributions/ai_stock_trading/tools/analysis.py new file mode 100644 index 0000000..9ab856e --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/analysis.py @@ -0,0 +1,316 @@ +""" +Stock Analysis Module + +This module provides enhanced technical and fundamental analysis capabilities +for stock data with advanced metrics and indicators. +""" + +import pandas as pd +import numpy as np +from typing import Dict, List, Optional, Tuple, Union, Any +import warnings + +warnings.filterwarnings('ignore') + +class StockAnalyzer: + """Enhanced stock analyzer with comprehensive technical indicators""" + + def __init__(self): + pass + + def analyze_stock(self, data: pd.DataFrame) -> Dict: + """ + Comprehensive stock analysis with enhanced metrics + + Args: + data: DataFrame with OHLCV stock data + + Returns: + Dictionary with analysis results + """ + if data.empty: + return {'error': 'No data provided for analysis'} + + try: + analysis = {} + + # Basic price metrics + analysis.update(self._calculate_price_metrics(data)) + + # Technical indicators + analysis.update(self._calculate_technical_indicators(data)) + + # Volatility analysis + analysis.update(self._calculate_volatility_metrics(data)) + + # Volume analysis + analysis.update(self._calculate_volume_metrics(data)) + + # Trend analysis + analysis.update(self._calculate_trend_metrics(data)) + + # Risk metrics + analysis.update(self._calculate_risk_metrics(data)) + + # Performance metrics + analysis.update(self._calculate_performance_metrics(data)) + + return analysis + + except Exception as e: + return {'error': f'Analysis failed: {str(e)}'} + + def _calculate_price_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate basic price metrics""" + close_prices = data['Close'] + + return { + 'current_price': float(close_prices.iloc[-1]), + 'start_price': float(close_prices.iloc[0]), + 'max_price': float(close_prices.max()), + 'min_price': float(close_prices.min()), + 'price_range_pct': float(((close_prices.max() - close_prices.min()) / close_prices.min()) * 100), + 'total_return_pct': float(((close_prices.iloc[-1] - close_prices.iloc[0]) / close_prices.iloc[0]) * 100) + } + + def _calculate_technical_indicators(self, data: pd.DataFrame) -> Dict: + """Calculate technical indicators""" + close_prices = data['Close'] + high_prices = data['High'] + low_prices = data['Low'] + + indicators = {} + + # Moving averages + if len(data) >= 20: + sma_20 = close_prices.rolling(window=20).mean() + indicators['sma_20'] = float(sma_20.iloc[-1]) + indicators['price_vs_sma_20'] = float(((close_prices.iloc[-1] - sma_20.iloc[-1]) / sma_20.iloc[-1]) * 100) + + if len(data) >= 50: + sma_50 = close_prices.rolling(window=50).mean() + indicators['sma_50'] = float(sma_50.iloc[-1]) + indicators['price_vs_sma_50'] = float(((close_prices.iloc[-1] - sma_50.iloc[-1]) / sma_50.iloc[-1]) * 100) + + # Exponential Moving Average + if len(data) >= 12: + ema_12 = close_prices.ewm(span=12).mean() + indicators['ema_12'] = float(ema_12.iloc[-1]) + + # RSI (Relative Strength Index) + if len(data) >= 14: + rsi = self._calculate_rsi(pd.Series(close_prices), 14) + indicators['rsi'] = float(rsi.iloc[-1]) + indicators['rsi_signal'] = self._interpret_rsi(float(rsi.iloc[-1])) + + # MACD + if len(data) >= 26: + macd_line, signal_line, histogram = self._calculate_macd(pd.Series(close_prices)) + indicators['macd'] = float(macd_line.iloc[-1]) + indicators['macd_signal'] = float(signal_line.iloc[-1]) + indicators['macd_histogram'] = float(histogram.iloc[-1]) + indicators['macd_trend'] = 'bullish' if float(histogram.iloc[-1]) > 0 else 'bearish' + + # Bollinger Bands + if len(data) >= 20: + bb_upper, bb_middle, bb_lower = self._calculate_bollinger_bands(pd.Series(close_prices), 20, 2) + indicators['bb_upper'] = float(bb_upper.iloc[-1]) + indicators['bb_middle'] = float(bb_middle.iloc[-1]) + indicators['bb_lower'] = float(bb_lower.iloc[-1]) + indicators['bb_position'] = self._interpret_bollinger_position(float(close_prices.iloc[-1]), float(bb_upper.iloc[-1]), float(bb_lower.iloc[-1])) + + return indicators + + def _calculate_volatility_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate volatility metrics""" + close_prices = data['Close'] + daily_returns = close_prices.pct_change().dropna() + + return { + 'volatility_daily': float(daily_returns.std() * 100), + 'volatility_annualized': float(daily_returns.std() * np.sqrt(252) * 100), + 'avg_daily_return': float(daily_returns.mean() * 100), + 'max_daily_gain': float(daily_returns.max() * 100), + 'max_daily_loss': float(daily_returns.min() * 100) + } + + def _calculate_volume_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate volume metrics""" + volume = data['Volume'] + + metrics: Dict[str, Union[float, str]] = { + 'avg_volume': float(volume.mean()), + 'current_volume': float(volume.iloc[-1]), + 'max_volume': float(volume.max()), + 'min_volume': float(volume.min()) + } + + # Volume trend + if len(volume) >= 10: + recent_avg = volume.tail(10).mean() + overall_avg = volume.mean() + if recent_avg > overall_avg: + metrics['volume_trend'] = 'increasing' + else: + metrics['volume_trend'] = 'decreasing' + metrics['volume_vs_avg'] = float(((recent_avg - overall_avg) / overall_avg) * 100) + + return metrics + + def _calculate_trend_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate trend analysis metrics""" + close_prices = data['Close'] + + # Linear regression for trend + x = np.arange(len(close_prices)) + slope, intercept = np.polyfit(x, close_prices, 1) + + # Trend strength + correlation = np.corrcoef(x, close_prices)[0, 1] + + return { + 'trend_slope': float(slope), + 'trend_direction': 'upward' if slope > 0 else 'downward', + 'trend_strength': float(abs(correlation)), + 'trend_angle': float(np.degrees(np.arctan(slope))), + 'r_squared': float(correlation ** 2) + } + + def _calculate_risk_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate risk metrics""" + close_prices = data['Close'] + daily_returns = close_prices.pct_change().dropna() + + # Value at Risk (VaR) + var_95 = np.percentile(daily_returns, 5) + var_99 = np.percentile(daily_returns, 1) + + # Maximum Drawdown + cumulative_returns = (1 + daily_returns).cumprod() + running_max = cumulative_returns.expanding().max() + drawdown = (cumulative_returns - running_max) / running_max + max_drawdown = drawdown.min() + + # Sharpe Ratio (assuming risk-free rate of 2%) + risk_free_rate = 0.02 / 252 # Daily risk-free rate + excess_returns = daily_returns - risk_free_rate + sharpe_ratio = excess_returns.mean() / daily_returns.std() if daily_returns.std() != 0 else 0 + + return { + 'var_95': float(var_95 * 100), + 'var_99': float(var_99 * 100), + 'max_drawdown': float(max_drawdown * 100), + 'sharpe_ratio': float(sharpe_ratio * np.sqrt(252)), # Annualized + 'downside_deviation': float(daily_returns[daily_returns < 0].std() * 100) + } + + def _calculate_performance_metrics(self, data: pd.DataFrame) -> Dict: + """Calculate performance metrics""" + close_prices = data['Close'] + + # Different period returns + periods = { + '1_week': min(5, len(close_prices) - 1), + '1_month': min(22, len(close_prices) - 1), + '3_months': min(66, len(close_prices) - 1), + '6_months': min(132, len(close_prices) - 1) + } + + performance = {} + current_price = close_prices.iloc[-1] + + for period_name, days_back in periods.items(): + if days_back > 0: + past_price = close_prices.iloc[-(days_back + 1)] + return_pct = ((current_price - past_price) / past_price) * 100 + performance[f'return_{period_name}'] = float(return_pct) + + return performance + + def _calculate_rsi(self, prices: pd.Series, period: int = 14) -> pd.Series: + """Calculate Relative Strength Index""" + delta = prices.diff() + gain = (delta.where(delta > 0, 0)).rolling(window=period).mean() + loss = (-delta.where(delta < 0, 0)).rolling(window=period).mean() + rs = gain / loss + rsi = 100 - (100 / (1 + rs)) + return rsi + + def _interpret_rsi(self, rsi_value: float) -> str: + """Interpret RSI value""" + if rsi_value >= 70: + return 'overbought' + elif rsi_value <= 30: + return 'oversold' + else: + return 'neutral' + + def _calculate_macd(self, prices: pd.Series, fast: int = 12, slow: int = 26, signal: int = 9) -> Tuple[pd.Series, pd.Series, pd.Series]: + """Calculate MACD indicator""" + ema_fast = prices.ewm(span=fast).mean() + ema_slow = prices.ewm(span=slow).mean() + macd_line = ema_fast - ema_slow + signal_line = macd_line.ewm(span=signal).mean() + histogram = macd_line - signal_line + return macd_line, signal_line, histogram + + def _calculate_bollinger_bands(self, prices: pd.Series, period: int = 20, std_dev: int = 2) -> Tuple[pd.Series, pd.Series, pd.Series]: + """Calculate Bollinger Bands""" + sma = prices.rolling(window=period).mean() + std = prices.rolling(window=period).std() + upper_band = sma + (std * std_dev) + lower_band = sma - (std * std_dev) + return upper_band, sma, lower_band + + def _interpret_bollinger_position(self, current_price: float, upper_band: float, lower_band: float) -> str: + """Interpret position relative to Bollinger Bands""" + if current_price > upper_band: + return 'above_upper_band' + elif current_price < lower_band: + return 'below_lower_band' + else: + return 'within_bands' + + def get_analysis_summary(self, analysis: Dict) -> str: + """Generate a human-readable analysis summary""" + if 'error' in analysis: + return f"Analysis Error: {analysis['error']}" + + summary = [] + + # Price summary + current_price = analysis.get('current_price', 0) + total_return = analysis.get('total_return_pct', 0) + summary.append(f"Current Price: ${current_price:.2f}") + summary.append(f"Total Return: {total_return:.2f}%") + + # Trend + trend_direction = analysis.get('trend_direction', 'unknown') + trend_strength = analysis.get('trend_strength', 0) + summary.append(f"Trend: {trend_direction.title()} (Strength: {trend_strength:.2f})") + + # Technical indicators + if 'rsi' in analysis: + rsi = analysis['rsi'] + rsi_signal = analysis['rsi_signal'] + summary.append(f"RSI: {rsi:.1f} ({rsi_signal})") + + if 'macd_trend' in analysis: + macd_trend = analysis['macd_trend'] + summary.append(f"MACD: {macd_trend}") + + # Risk + volatility = analysis.get('volatility_annualized', 0) + max_drawdown = analysis.get('max_drawdown', 0) + summary.append(f"Volatility: {volatility:.1f}% (Annual)") + summary.append(f"Max Drawdown: {max_drawdown:.1f}%") + + return "\n".join(summary) + +# Global instance for easy import +stock_analyzer = StockAnalyzer() + +# Convenience function +def analyze_stock(data: pd.DataFrame) -> Dict: + """Convenience function to analyze stock data""" + return stock_analyzer.analyze_stock(data) diff --git a/week4/community-contributions/ai_stock_trading/tools/charting.py b/week4/community-contributions/ai_stock_trading/tools/charting.py new file mode 100644 index 0000000..e10384e --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/charting.py @@ -0,0 +1,483 @@ +""" +Charting Module + +This module provides comprehensive charting and visualization capabilities +for stock analysis with interactive dashboards using Plotly. +""" + +import pandas as pd +import numpy as np +import plotly.graph_objects as go +import plotly.express as px +from plotly.subplots import make_subplots +import streamlit as st +from typing import Dict, List, Optional, Tuple +import warnings + +warnings.filterwarnings('ignore') + +class StockChartGenerator: + """Enhanced stock chart generator with interactive dashboards""" + + def __init__(self): + self.color_scheme = { + 'primary': '#1f77b4', + 'secondary': '#ff7f0e', + 'success': '#2ca02c', + 'danger': '#d62728', + 'warning': '#ff7f0e', + 'info': '#17a2b8', + 'background': '#f8f9fa' + } + + def create_price_chart(self, data: pd.DataFrame, symbol: str, analysis: Dict = None) -> go.Figure: + """ + Create comprehensive price chart with technical indicators + + Args: + data: Stock price data + symbol: Stock symbol + analysis: Technical analysis results + + Returns: + Plotly figure object + """ + if data.empty: + return self._create_empty_chart("No data available") + + # Create subplots + fig = make_subplots( + rows=3, cols=1, + shared_xaxes=True, + vertical_spacing=0.05, + subplot_titles=(f'{symbol} Price Chart', 'Volume', 'Technical Indicators'), + row_heights=[0.6, 0.2, 0.2] + ) + + # Main price chart (candlestick) + fig.add_trace( + go.Candlestick( + x=data.index, + open=data['Open'], + high=data['High'], + low=data['Low'], + close=data['Close'], + name='Price', + increasing_line_color=self.color_scheme['success'], + decreasing_line_color=self.color_scheme['danger'] + ), + row=1, col=1 + ) + + # Add moving averages if available + if 'SMA_20' in data.columns: + fig.add_trace( + go.Scatter( + x=data.index, + y=data['SMA_20'], + mode='lines', + name='SMA 20', + line=dict(color=self.color_scheme['primary'], width=1) + ), + row=1, col=1 + ) + + if 'SMA_50' in data.columns: + fig.add_trace( + go.Scatter( + x=data.index, + y=data['SMA_50'], + mode='lines', + name='SMA 50', + line=dict(color=self.color_scheme['secondary'], width=1) + ), + row=1, col=1 + ) + + # Volume chart + colors = ['red' if close < open else 'green' for close, open in zip(data['Close'], data['Open'])] + fig.add_trace( + go.Bar( + x=data.index, + y=data['Volume'], + name='Volume', + marker_color=colors, + opacity=0.7 + ), + row=2, col=1 + ) + + # Technical indicators (RSI if available in analysis) + if analysis and 'rsi' in analysis: + # Create RSI line (simplified - would need full RSI calculation for time series) + rsi_value = analysis['rsi'] + rsi_line = [rsi_value] * len(data) + + fig.add_trace( + go.Scatter( + x=data.index, + y=rsi_line, + mode='lines', + name=f'RSI ({rsi_value:.1f})', + line=dict(color=self.color_scheme['info'], width=2) + ), + row=3, col=1 + ) + + # Add RSI reference lines + fig.add_hline(y=70, line_dash="dash", line_color="red", opacity=0.5, row=3, col=1) + fig.add_hline(y=30, line_dash="dash", line_color="green", opacity=0.5, row=3, col=1) + + # Update layout + fig.update_layout( + title=f'{symbol} Stock Analysis Dashboard', + xaxis_title='Date', + yaxis_title='Price ($)', + template='plotly_white', + height=800, + showlegend=True, + hovermode='x unified' + ) + + # Remove rangeslider for cleaner look + fig.update_layout(xaxis_rangeslider_visible=False) + + return fig + + def create_performance_chart(self, data: pd.DataFrame, symbol: str, analysis: Dict) -> go.Figure: + """ + Create performance analysis chart + + Args: + data: Stock price data + symbol: Stock symbol + analysis: Analysis results with performance metrics + + Returns: + Plotly figure object + """ + if data.empty: + return self._create_empty_chart("No data available for performance analysis") + + # Calculate cumulative returns + daily_returns = data['Close'].pct_change().fillna(0) + cumulative_returns = (1 + daily_returns).cumprod() - 1 + + fig = go.Figure() + + # Cumulative returns line + fig.add_trace( + go.Scatter( + x=data.index, + y=cumulative_returns * 100, + mode='lines', + name='Cumulative Returns (%)', + line=dict(color=self.color_scheme['primary'], width=2), + fill='tonexty', + fillcolor='rgba(31, 119, 180, 0.1)' + ) + ) + + # Add benchmark line (0% return) + fig.add_hline(y=0, line_dash="dash", line_color="gray", opacity=0.5) + + # Add performance annotations + if analysis: + total_return = analysis.get('total_return_pct', 0) + fig.add_annotation( + x=data.index[-1], + y=total_return, + text=f"Total Return: {total_return:.1f}%", + showarrow=True, + arrowhead=2, + arrowcolor=self.color_scheme['primary'], + bgcolor="white", + bordercolor=self.color_scheme['primary'] + ) + + fig.update_layout( + title=f'{symbol} Performance Analysis', + xaxis_title='Date', + yaxis_title='Cumulative Returns (%)', + template='plotly_white', + height=500, + hovermode='x' + ) + + return fig + + def create_risk_analysis_chart(self, analysis: Dict, symbol: str) -> go.Figure: + """ + Create risk analysis visualization + + Args: + analysis: Analysis results with risk metrics + symbol: Stock symbol + + Returns: + Plotly figure object + """ + if not analysis or 'error' in analysis: + return self._create_empty_chart("No risk data available") + + # Prepare risk metrics + risk_metrics = { + 'Volatility (Annual)': analysis.get('volatility_annualized', 0), + 'Max Drawdown': abs(analysis.get('max_drawdown', 0)), + 'VaR 95%': abs(analysis.get('var_95', 0)), + 'VaR 99%': abs(analysis.get('var_99', 0)) + } + + # Create radar chart for risk metrics + categories = list(risk_metrics.keys()) + values = list(risk_metrics.values()) + + fig = go.Figure() + + fig.add_trace(go.Scatterpolar( + r=values, + theta=categories, + fill='toself', + name=f'{symbol} Risk Profile', + line_color=self.color_scheme['danger'], + fillcolor='rgba(214, 39, 40, 0.1)' + )) + + fig.update_layout( + polar=dict( + radialaxis=dict( + visible=True, + range=[0, max(values) * 1.2] if values else [0, 100] + ) + ), + title=f'{symbol} Risk Analysis Chart', + template='plotly_white', + height=500 + ) + + return fig + + def create_comparison_chart(self, data_dict: Dict[str, pd.DataFrame], symbols: List[str]) -> go.Figure: + """ + Create comparison chart for multiple stocks + + Args: + data_dict: Dictionary of stock data {symbol: dataframe} + symbols: List of stock symbols to compare + + Returns: + Plotly figure object + """ + fig = go.Figure() + + colors = [self.color_scheme['primary'], self.color_scheme['secondary'], + self.color_scheme['success'], self.color_scheme['danger']] + + for i, symbol in enumerate(symbols): + if symbol in data_dict and not data_dict[symbol].empty: + data = data_dict[symbol] + # Normalize prices to start at 100 for comparison + normalized_prices = (data['Close'] / data['Close'].iloc[0]) * 100 + + fig.add_trace( + go.Scatter( + x=data.index, + y=normalized_prices, + mode='lines', + name=symbol, + line=dict(color=colors[i % len(colors)], width=2) + ) + ) + + fig.update_layout( + title='Stock Price Comparison (Normalized to 100)', + xaxis_title='Date', + yaxis_title='Normalized Price', + template='plotly_white', + height=600, + hovermode='x unified' + ) + + return fig + + def create_sector_analysis_chart(self, sector_data: Dict) -> go.Figure: + """ + Create sector analysis visualization + + Args: + sector_data: Dictionary with sector analysis data + + Returns: + Plotly figure object + """ + # This would typically show sector performance, P/E ratios, etc. + # For now, create a placeholder + fig = go.Figure() + + fig.add_annotation( + x=0.5, y=0.5, + text="Sector Analysis
Coming Soon", + showarrow=False, + font=dict(size=20), + xref="paper", yref="paper" + ) + + fig.update_layout( + title='Sector Analysis Dashboard', + template='plotly_white', + height=400, + showticklabels=False + ) + + return fig + + def create_trading_signals_chart(self, data: pd.DataFrame, analysis: Dict, trading_decision: Dict, symbol: str) -> go.Figure: + """ + Create trading signals visualization + + Args: + data: Stock price data + analysis: Technical analysis results + trading_decision: Trading recommendation + symbol: Stock symbol + + Returns: + Plotly figure object + """ + if data.empty: + return self._create_empty_chart("No data available for trading signals") + + fig = go.Figure() + + # Price line + fig.add_trace( + go.Scatter( + x=data.index, + y=data['Close'], + mode='lines', + name='Price', + line=dict(color=self.color_scheme['primary'], width=2) + ) + ) + + # Add trading signal + recommendation = trading_decision.get('recommendation', 'HOLD') + current_price = data['Close'].iloc[-1] + + signal_color = { + 'BUY': self.color_scheme['success'], + 'SELL': self.color_scheme['danger'], + 'HOLD': self.color_scheme['warning'] + }.get(recommendation, self.color_scheme['info']) + + fig.add_trace( + go.Scatter( + x=[data.index[-1]], + y=[current_price], + mode='markers', + name=f'{recommendation} Signal', + marker=dict( + color=signal_color, + size=15, + symbol='triangle-up' if recommendation == 'BUY' else + 'triangle-down' if recommendation == 'SELL' else 'circle' + ) + ) + ) + + # Add price target if available + price_target = trading_decision.get('price_target') + if price_target: + fig.add_hline( + y=price_target, + line_dash="dash", + line_color=self.color_scheme['success'], + annotation_text=f"Target: ${price_target:.2f}" + ) + + # Add stop loss if available + stop_loss = trading_decision.get('stop_loss') + if stop_loss: + fig.add_hline( + y=stop_loss, + line_dash="dash", + line_color=self.color_scheme['danger'], + annotation_text=f"Stop Loss: ${stop_loss:.2f}" + ) + + fig.update_layout( + title=f'{symbol} Trading Signals', + xaxis_title='Date', + yaxis_title='Price ($)', + template='plotly_white', + height=500, + hovermode='x' + ) + + return fig + + def create_dashboard_summary(self, symbol: str, analysis: Dict, trading_decision: Dict, sharia_compliance: Dict) -> Dict: + """ + Create summary metrics for dashboard display + + Args: + symbol: Stock symbol + analysis: Technical analysis results + trading_decision: Trading recommendation + sharia_compliance: Sharia compliance results + + Returns: + Dictionary with summary metrics + """ + summary = { + 'symbol': symbol, + 'current_price': analysis.get('current_price', 0), + 'total_return': analysis.get('total_return_pct', 0), + 'volatility': analysis.get('volatility_annualized', 0), + 'trading_recommendation': trading_decision.get('recommendation', 'HOLD'), + 'trading_confidence': trading_decision.get('confidence', 0) * 100, + 'sharia_ruling': sharia_compliance.get('ruling', 'UNCERTAIN'), + 'sharia_confidence': sharia_compliance.get('confidence', 0) * 100, + 'risk_level': trading_decision.get('risk_level', 'medium'), + 'trend_direction': analysis.get('trend_direction', 'unknown'), + 'rsi': analysis.get('rsi', 50), + 'max_drawdown': analysis.get('max_drawdown', 0) + } + + return summary + + def _create_empty_chart(self, message: str) -> go.Figure: + """Create an empty chart with a message""" + fig = go.Figure() + + fig.add_annotation( + x=0.5, y=0.5, + text=message, + showarrow=False, + font=dict(size=16), + xref="paper", yref="paper" + ) + + fig.update_layout( + template='plotly_white', + height=400, + showticklabels=False + ) + + return fig + +# Global instance for easy import +chart_generator = StockChartGenerator() + +# Convenience functions +def create_price_chart(data: pd.DataFrame, symbol: str, analysis: Dict = None) -> go.Figure: + """Convenience function to create price chart""" + return chart_generator.create_price_chart(data, symbol, analysis) + +def create_performance_chart(data: pd.DataFrame, symbol: str, analysis: Dict) -> go.Figure: + """Convenience function to create performance chart""" + return chart_generator.create_performance_chart(data, symbol, analysis) + +def create_trading_signals_chart(data: pd.DataFrame, analysis: Dict, trading_decision: Dict, symbol: str) -> go.Figure: + """Convenience function to create trading signals chart""" + return chart_generator.create_trading_signals_chart(data, analysis, trading_decision, symbol) diff --git a/week4/community-contributions/ai_stock_trading/tools/fetching.py b/week4/community-contributions/ai_stock_trading/tools/fetching.py new file mode 100644 index 0000000..5966e8e --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/fetching.py @@ -0,0 +1,384 @@ +""" +Stock Data Fetching Module + +This module handles fetching stock data from various sources including yfinance +and provides enhanced data retrieval capabilities for different markets. +""" + +import yfinance as yf +import pandas as pd +import numpy as np +import requests +from typing import Dict, List, Optional, Tuple +import warnings + +warnings.filterwarnings('ignore') + +class StockDataFetcher: + """Enhanced stock data fetcher with multi-market support""" + + # Stock symbols for different markets + STOCK_SYMBOLS = { + 'USA': { + # Technology + 'Apple Inc.': 'AAPL', + 'Microsoft Corporation': 'MSFT', + 'NVIDIA Corporation': 'NVDA', + 'Alphabet Inc. (Class A)': 'GOOGL', + 'Alphabet Inc. (Class C)': 'GOOG', + 'Meta Platforms Inc.': 'META', + 'Tesla Inc.': 'TSLA', + 'Amazon.com Inc.': 'AMZN', + 'Netflix Inc.': 'NFLX', + 'Adobe Inc.': 'ADBE', + 'Salesforce Inc.': 'CRM', + 'Oracle Corporation': 'ORCL', + 'Cisco Systems Inc.': 'CSCO', + 'Intel Corporation': 'INTC', + 'Advanced Micro Devices': 'AMD', + 'Qualcomm Inc.': 'QCOM', + 'Texas Instruments': 'TXN', + 'Broadcom Inc.': 'AVGO', + 'ServiceNow Inc.': 'NOW', + 'Palantir Technologies': 'PLTR', + + # Financial Services + 'JPMorgan Chase & Co.': 'JPM', + 'Bank of America Corp': 'BAC', + 'Wells Fargo & Company': 'WFC', + 'Goldman Sachs Group': 'GS', + 'Morgan Stanley': 'MS', + 'Citigroup Inc.': 'C', + 'American Express Company': 'AXP', + 'Berkshire Hathaway Inc.': 'BRK.B', + 'BlackRock Inc.': 'BLK', + 'Charles Schwab Corporation': 'SCHW', + 'Visa Inc.': 'V', + 'Mastercard Inc.': 'MA', + + # Healthcare & Pharmaceuticals + 'Johnson & Johnson': 'JNJ', + 'UnitedHealth Group': 'UNH', + 'Pfizer Inc.': 'PFE', + 'AbbVie Inc.': 'ABBV', + 'Merck & Co Inc.': 'MRK', + 'Eli Lilly and Company': 'LLY', + 'Abbott Laboratories': 'ABT', + 'Thermo Fisher Scientific': 'TMO', + 'Danaher Corporation': 'DHR', + 'Gilead Sciences Inc.': 'GILD', + + # Consumer & Retail + 'Walmart Inc.': 'WMT', + 'Procter & Gamble Co': 'PG', + 'Coca-Cola Company': 'KO', + 'PepsiCo Inc.': 'PEP', + 'Home Depot Inc.': 'HD', + 'McDonald\'s Corporation': 'MCD', + 'Nike Inc.': 'NKE', + 'Costco Wholesale Corp': 'COST', + 'TJX Companies Inc.': 'TJX', + 'Lowe\'s Companies Inc.': 'LOW', + + # Industrial & Energy + 'Exxon Mobil Corporation': 'XOM', + 'Chevron Corporation': 'CVX', + 'ConocoPhillips': 'COP', + 'Caterpillar Inc.': 'CAT', + 'Boeing Company': 'BA', + 'General Electric': 'GE', + 'Honeywell International': 'HON', + 'Deere & Company': 'DE', + 'Union Pacific Corporation': 'UNP', + 'Lockheed Martin Corp': 'LMT', + + # Communication & Media + 'AT&T Inc.': 'T', + 'Verizon Communications': 'VZ', + 'T-Mobile US Inc.': 'TMUS', + 'Comcast Corporation': 'CMCSA', + 'Walt Disney Company': 'DIS' + }, + 'Egypt': { + # Banking & Financial Services + 'Commercial International Bank': 'COMI.CA', + 'QNB Alahli Bank': 'QNBE.CA', + 'Housing and Development Bank': 'HDBK.CA', + 'Abu Dhabi Islamic Bank Egypt': 'ADIB.CA', + 'Egyptian Gulf Bank': 'EGBE.CA', + + # Real Estate & Construction + 'Talaat Moustafa Group Holding': 'TMGH.CA', + 'Palm Hills Developments': 'PHDC.CA', + 'Orascom Construction': 'ORAS.CA', + 'Orascom Development Holding': 'ORHD.CA', + 'Six of October Development': 'SCTS.CA', + 'Heliopolis Housing': 'HELI.CA', + 'Rooya Group': 'RMDA.CA', + + # Industrial & Manufacturing + 'Eastern Company': 'EAST.CA', + 'El Sewedy Electric Company': 'SWDY.CA', + 'Ezz Steel': 'ESRS.CA', + 'Iron and Steel Company': 'IRON.CA', + 'Alexandria Containers': 'ALCN.CA', + 'Sidi Kerir Petrochemicals': 'SKPC.CA', + + # Chemicals & Fertilizers + 'Abu Qir Fertilizers and Chemical Industries': 'ABUK.CA', + 'Egyptian Chemical Industries (Kima)': 'KIMA.CA', + 'Misr Fertilizers Production': 'MFPC.CA', + + # Telecommunications & Technology + 'Telecom Egypt': 'ETEL.CA', + 'Raya Holding': 'RAYA.CA', + 'E-Finance for Digital Payments': 'EFIH.CA', + 'Fawry for Banking Technology': 'FWRY.CA', + + # Food & Beverages + 'Juhayna Food Industries': 'JUFO.CA', + 'Edita Food Industries': 'EFID.CA', + 'Cairo Poultry Company': 'POUL.CA', + 'Upper Egypt Flour Mills': 'UEFM.CA', + 'Ismailia Misr Poultry': 'ISPH.CA', + + # Healthcare & Pharmaceuticals + 'Cleopatra Hospital Group': 'CLHO.CA', + 'Cairo Pharmaceuticals': 'PHAR.CA', + + # Energy & Utilities + 'Egyptian Natural Gas Company': 'EGAS.CA', + 'Suez Cement Company': 'SCEM.CA', + 'Arabian Cement Company': 'ARCC.CA', + + # Investment & Holding Companies + 'Egyptian Financial Group-Hermes': 'HRHO.CA', + 'Citadel Capital': 'CCAP.CA', + 'Beltone Financial Holding': 'BTFH.CA' + } + } + + # Currency mapping for different markets + MARKET_CURRENCIES = { + 'USA': 'USD', + 'Egypt': 'EGP' + } + + def __init__(self): + self.cache = {} + + def get_available_stocks(self, country: str) -> Dict[str, str]: + """Get available stocks for a specific country""" + return self.STOCK_SYMBOLS.get(country, {}) + + def get_market_currency(self, country: str) -> str: + """Get the currency for a specific market""" + return self.MARKET_CURRENCIES.get(country, 'USD') + + def format_price_with_currency(self, price: float, country: str) -> str: + """Format price with appropriate currency symbol""" + currency = self.get_market_currency(country) + if currency == 'EGP': + return f"{price:.2f} EGP" + elif currency == 'USD': + return f"${price:.2f}" + else: + return f"{price:.2f} {currency}" + + def fetch_stock_data(self, symbol: str, period: str = "1y", interval: str = "1d") -> pd.DataFrame: + """ + Fetch historical stock data with enhanced error handling + + Args: + symbol: Stock symbol (e.g., 'AAPL', 'COMI.CA') + period: Time period ('1d', '5d', '1mo', '3mo', '6mo', '1y', '2y', '5y', '10y', 'ytd', 'max') + interval: Data interval ('1m', '2m', '5m', '15m', '30m', '60m', '90m', '1h', '1d', '5d', '1wk', '1mo', '3mo') + + Returns: + DataFrame with OHLCV data + """ + cache_key = f"{symbol}_{period}_{interval}" + + # Check cache first + if cache_key in self.cache: + return self.cache[cache_key] + + try: + # Create ticker object + ticker = yf.Ticker(symbol) + + # Fetch historical data + data = ticker.history(period=period, interval=interval) + + if data.empty: + print(f"⚠️ No data found for {symbol}") + return pd.DataFrame() + + # Clean and enhance data + data = self._clean_data(data) + + # Cache the result + self.cache[cache_key] = data + + print(f"✅ Successfully fetched {len(data)} data points for {symbol} ({period})") + return data + + except Exception as e: + print(f"❌ Error fetching data for {symbol}: {str(e)}") + return pd.DataFrame() + + def get_stock_info(self, symbol: str, country: Optional[str] = None) -> Dict: + """ + Get comprehensive stock information + + Args: + symbol: Stock symbol + country: Market country (USA, Egypt) for currency handling + + Returns: + Dictionary with stock information + """ + try: + ticker = yf.Ticker(symbol) + info = ticker.info + + # Detect country if not provided + if country is None: + country = self._detect_country_from_symbol(symbol) + + # Get market currency + market_currency = self.get_market_currency(country) + + # Extract key information + stock_info = { + 'symbol': symbol, + 'company_name': info.get('longName', 'N/A'), + 'sector': info.get('sector', 'N/A'), + 'industry': info.get('industry', 'N/A'), + 'market_cap': info.get('marketCap', 0), + 'pe_ratio': info.get('trailingPE', 0), + 'dividend_yield': info.get('dividendYield', 0), + 'beta': info.get('beta', 0), + 'fifty_two_week_high': info.get('fiftyTwoWeekHigh', 0), + 'fifty_two_week_low': info.get('fiftyTwoWeekLow', 0), + 'current_price': info.get('currentPrice', 0), + 'currency': market_currency, # Use detected market currency + 'exchange': info.get('exchange', 'N/A'), + 'country': country, + 'market_country': country # Add explicit market country + } + + return stock_info + + except Exception as e: + print(f"❌ Error fetching info for {symbol}: {str(e)}") + return {'symbol': symbol, 'error': str(e)} + + def _detect_country_from_symbol(self, symbol: str) -> str: + """ + Detect country from stock symbol + + Args: + symbol: Stock symbol + + Returns: + Country name (USA or Egypt) + """ + # Check if symbol exists in any country's stock list + for country, stocks in self.STOCK_SYMBOLS.items(): + if symbol in stocks.values(): + return country + + # Default to USA if not found + return 'USA' + + def fetch_multiple_periods(self, symbol: str) -> Dict[str, pd.DataFrame]: + """ + Fetch data for multiple time periods + + Args: + symbol: Stock symbol + + Returns: + Dictionary with DataFrames for different periods + """ + periods = ['1mo', '1y', '5y'] + data = {} + + for period in periods: + df = self.fetch_stock_data(symbol, period) + if not df.empty: + data[period] = df + + return data + + def _clean_data(self, data: pd.DataFrame) -> pd.DataFrame: + """ + Clean and enhance the stock data + + Args: + data: Raw stock data DataFrame + + Returns: + Cleaned DataFrame + """ + # Remove rows with all NaN values + data = data.dropna(how='all') + + # Forward fill missing values + data = data.fillna(method='ffill') + + # Add technical indicators + if len(data) > 0: + # Simple moving averages + if len(data) >= 20: + data['SMA_20'] = data['Close'].rolling(window=20).mean() + if len(data) >= 50: + data['SMA_50'] = data['Close'].rolling(window=50).mean() + + # Daily returns + data['Daily_Return'] = data['Close'].pct_change() + + # Price change from previous day + data['Price_Change'] = data['Close'].diff() + data['Price_Change_Pct'] = (data['Price_Change'] / data['Close'].shift(1)) * 100 + + return data + + def get_real_time_price(self, symbol: str) -> Optional[float]: + """ + Get real-time stock price + + Args: + symbol: Stock symbol + + Returns: + Current stock price or None if error + """ + try: + ticker = yf.Ticker(symbol) + data = ticker.history(period="1d", interval="1m") + + if not data.empty: + return float(data['Close'].iloc[-1]) + return None + + except Exception as e: + print(f"❌ Error fetching real-time price for {symbol}: {str(e)}") + return None + +# Global instance for easy import +stock_fetcher = StockDataFetcher() + +# Convenience functions +def fetch_stock_data(symbol: str, period: str = "1y", interval: str = "1d") -> pd.DataFrame: + """Convenience function to fetch stock data""" + return stock_fetcher.fetch_stock_data(symbol, period, interval) + +def get_available_stocks(country: str) -> Dict[str, str]: + """Convenience function to get available stocks""" + return stock_fetcher.get_available_stocks(country) + +def get_stock_info(symbol: str) -> Dict: + """Convenience function to get stock info""" + return stock_fetcher.get_stock_info(symbol) diff --git a/week4/community-contributions/ai_stock_trading/tools/sharia_compliance.py b/week4/community-contributions/ai_stock_trading/tools/sharia_compliance.py new file mode 100644 index 0000000..f0f3119 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/sharia_compliance.py @@ -0,0 +1,591 @@ +""" +Sharia Compliance Module + +This module provides comprehensive Islamic finance compliance checking +for stocks and investments according to Islamic principles. +""" + +import os +import json +import requests +from typing import Dict, List, Optional, Tuple +import pandas as pd +from openai import OpenAI +from dotenv import load_dotenv +from bs4 import BeautifulSoup +import time +import re + +# Load environment variables +load_dotenv() + +class ShariaComplianceChecker: + """Enhanced Sharia compliance checker for Islamic investing""" + + def __init__(self): + self.client = OpenAI(api_key=os.getenv('OPENAI_API_KEY')) + + # Sharia compliance criteria weights + self.criteria_weights = { + 'business_activity': 0.40, # Most important + 'financial_ratios': 0.30, + 'debt_levels': 0.20, + 'revenue_sources': 0.10 + } + + # Prohibited business activities (comprehensive list) + self.prohibited_activities = { + # Core prohibitions + 'alcohol', 'alcoholic_beverages', 'wine', 'beer', 'spirits', 'liquor', + 'gambling', 'casino', 'lottery', 'betting', 'gaming', 'poker', + 'tobacco', 'cigarettes', 'smoking', 'nicotine', + 'pork', 'pig_farming', 'swine', 'ham', 'bacon', + 'adult_entertainment', 'pornography', 'strip_clubs', 'escort_services', + + # Financial prohibitions + 'conventional_banking', 'interest_based_finance', 'usury', 'riba', + 'conventional_insurance', 'life_insurance', 'derivatives_trading', + 'forex_trading', 'currency_speculation', 'margin_trading', + 'short_selling', 'day_trading', 'high_frequency_trading', + + # Weapons and defense + 'weapons', 'arms_manufacturing', 'defense_contractors', 'military_equipment', + 'ammunition', 'explosives', 'nuclear_weapons', + + # Other prohibitions + 'nightclubs', 'bars', 'entertainment_venues', 'music_industry', + 'film_industry', 'media_entertainment', 'advertising_haram_products' + } + + # Sharia-compliant sectors (generally accepted) + self.compliant_sectors = { + 'technology', 'healthcare', 'pharmaceuticals', 'telecommunications', + 'utilities', 'real_estate', 'construction', 'manufacturing', + 'retail', 'food_beverages', 'transportation', 'energy_renewable' + } + + # Questionable sectors (need detailed analysis) + self.questionable_sectors = { + 'financial_services', 'media', 'hotels', 'airlines', + 'oil_gas', 'mining', 'chemicals', 'entertainment', + 'restaurants', 'hospitality', 'advertising' + } + + # AAOIFI and DSN Sharia standards + self.sharia_standards = { + 'max_debt_to_assets': 0.33, # 33% maximum debt-to-assets ratio + 'max_interest_income': 0.05, # 5% maximum interest income + 'max_non_compliant_income': 0.05, # 5% maximum non-compliant income + 'min_tangible_assets': 0.20 # 20% minimum tangible assets + } + + def _search_company_business_activities(self, company_name: str, symbol: str) -> Dict: + """ + Search web for company's business activities to verify Sharia compliance + + Args: + company_name: Company name + symbol: Stock symbol + + Returns: + Dictionary with business activity information + """ + try: + # Search query for company business activities + search_queries = [ + f"{company_name} business activities products services", + f"{company_name} {symbol} what does company do", + f"{company_name} revenue sources business model" + ] + + business_info = { + 'activities': [], + 'products': [], + 'services': [], + 'revenue_sources': [], + 'prohibited_found': [], + 'confidence': 0.5 + } + + for query in search_queries[:1]: # Limit to 1 search to avoid rate limits + try: + # Simple web search simulation (in production, use proper search API) + search_url = f"https://www.google.com/search?q={query.replace(' ', '+')}" + headers = { + 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36' + } + + # For now, return basic analysis based on company name and sector + # In production, implement actual web scraping with proper rate limiting + business_info['confidence'] = 0.6 + break + + except Exception as e: + print(f"Web search error: {e}") + continue + + return business_info + + except Exception as e: + print(f"Error in business activity search: {e}") + return {'activities': [], 'prohibited_found': [], 'confidence': 0.3} + + def _estimate_debt_ratio(self, stock_info: Dict) -> float: + """ + Estimate debt-to-assets ratio based on available information + + Args: + stock_info: Stock information dictionary + + Returns: + Estimated debt-to-assets ratio + """ + try: + # In production, this would fetch actual balance sheet data + # For now, estimate based on sector and other indicators + sector = stock_info.get('sector', '').lower() + industry = stock_info.get('industry', '').lower() + + # High debt sectors + if any(x in sector or x in industry for x in ['utility', 'telecom', 'airline', 'real estate']): + return 0.45 # Typically higher debt + + # Medium debt sectors + elif any(x in sector or x in industry for x in ['manufacturing', 'retail', 'energy']): + return 0.25 + + # Low debt sectors + elif any(x in sector or x in industry for x in ['technology', 'healthcare', 'software']): + return 0.15 + + # Financial sector (different calculation) + elif 'financial' in sector or 'bank' in sector: + return 0.8 # Banks have high leverage by nature + + # Default estimate + return 0.3 + + except Exception: + return 0.3 # Conservative default + + def _check_business_activity(self, stock_info: Dict) -> float: + """ + Check if the company's primary business activity is Sharia-compliant + + Returns: + Score from 0.0 (non-compliant) to 1.0 (fully compliant) + """ + sector = stock_info.get('sector', '').lower() + industry = stock_info.get('industry', '').lower() + company_name = stock_info.get('company_name', '').lower() + + # Check for explicitly prohibited activities + for prohibited in self.prohibited_activities: + if (prohibited.replace('_', ' ') in sector or + prohibited.replace('_', ' ') in industry or + prohibited.replace('_', ' ') in company_name): + return 0.0 + + # Check for compliant sectors + for compliant in self.compliant_sectors: + if (compliant.replace('_', ' ') in sector or + compliant.replace('_', ' ') in industry): + return 1.0 + + # Check for questionable sectors + for questionable in self.questionable_sectors: + if (questionable.replace('_', ' ') in sector or + questionable.replace('_', ' ') in industry): + return 0.5 + + # Default for unknown sectors + return 0.7 + + def check_sharia_compliance(self, symbol: str, stock_info: Dict, analysis: Dict) -> Dict: + """ + Comprehensive Sharia compliance check + + Args: + symbol: Stock symbol + stock_info: Stock information + analysis: Technical analysis results + + Returns: + Dictionary with compliance assessment + """ + try: + # Business activity screening + business_score = self._check_business_activity(stock_info) + + # Financial ratios screening + financial_score = self._check_financial_ratios(stock_info, analysis) + + # Debt levels screening + debt_score = self._check_debt_levels(stock_info) + + # Revenue sources screening + revenue_score = self._check_revenue_sources(stock_info) + + # Calculate weighted compliance score + total_score = ( + business_score * self.criteria_weights['business_activity'] + + financial_score * self.criteria_weights['financial_ratios'] + + debt_score * self.criteria_weights['debt_levels'] + + revenue_score * self.criteria_weights['revenue_sources'] + ) + + # Get AI-powered detailed analysis + ai_analysis = self._get_ai_sharia_analysis(symbol, stock_info) + + # Determine final ruling + ruling = self._determine_ruling(total_score, ai_analysis) + + return { + 'symbol': symbol, + 'ruling': ruling['status'], + 'confidence': ruling['confidence'], + 'compliance_score': total_score, + 'detailed_scores': { + 'business_activity': business_score, + 'financial_ratios': financial_score, + 'debt_levels': debt_score, + 'revenue_sources': revenue_score + }, + 'reasoning': ruling['reasoning'], + 'key_concerns': ruling.get('concerns', []), + 'recommendations': ruling.get('recommendations', []), + 'ai_analysis': ai_analysis.get('analysis', ''), + 'scholar_consultation_advised': ruling.get('scholar_consultation', False), + 'alternative_suggestions': ruling.get('alternatives', []) + } + + except Exception as e: + return { + 'symbol': symbol, + 'ruling': 'UNCERTAIN', + 'confidence': 0.0, + 'reasoning': f'Error in Sharia compliance analysis: {str(e)}', + 'error': str(e) + } + + def _check_business_activity(self, stock_info: Dict) -> float: + """ + Check if the company's primary business activity is Sharia-compliant + + Returns: + Score from 0.0 (non-compliant) to 1.0 (fully compliant) + """ + sector = stock_info.get('sector', '').lower() + industry = stock_info.get('industry', '').lower() + company_name = stock_info.get('company_name', '').lower() + + # Check for explicitly prohibited activities + for prohibited in self.prohibited_activities: + if (prohibited.replace('_', ' ') in sector or + prohibited.replace('_', ' ') in industry or + prohibited.replace('_', ' ') in company_name): + return 0.0 + + # Check for compliant sectors + for compliant in self.compliant_sectors: + if (compliant.replace('_', ' ') in sector or + compliant.replace('_', ' ') in industry): + return 1.0 + + # Check for questionable sectors + for questionable in self.questionable_sectors: + if (questionable.replace('_', ' ') in sector or + questionable.replace('_', ' ') in industry): + return 0.5 + + # Default for unknown sectors + return 0.7 + + def _check_financial_ratios(self, stock_info: Dict, analysis: Dict) -> float: + """ + Check financial ratios according to AAOIFI and DSN Sharia standards + + AAOIFI/DSN Sharia screening ratios: + - Debt/Total Assets < 33% + - Interest Income/Total Revenue < 5% + - Non-compliant Income/Total Revenue < 5% + - Tangible Assets/Total Assets > 20% + + Returns: + Score from 0.0 to 1.0 + """ + score = 1.0 + penalties = [] + + try: + # Get financial metrics (these would come from detailed financial data) + market_cap = stock_info.get('market_cap', 0) + pe_ratio = stock_info.get('pe_ratio', 0) + + # Debt-to-Assets ratio check + # Note: In production, fetch actual balance sheet data + debt_to_assets = self._estimate_debt_ratio(stock_info) + if debt_to_assets > self.sharia_standards['max_debt_to_assets']: + penalty = min(0.5, (debt_to_assets - self.sharia_standards['max_debt_to_assets']) * 2) + score -= penalty + penalties.append(f"High debt ratio: {debt_to_assets:.1%} > {self.sharia_standards['max_debt_to_assets']:.1%}") + + # Interest income check (for financial companies) + sector = stock_info.get('sector', '').lower() + if 'financial' in sector or 'bank' in sector: + # Financial companies likely have significant interest income + score -= 0.3 + penalties.append("Financial sector - likely high interest income") + + # Industry-specific checks + industry = stock_info.get('industry', '').lower() + if any(prohibited in industry for prohibited in ['insurance', 'casino', 'alcohol', 'tobacco']): + score = 0.0 + penalties.append(f"Prohibited industry: {industry}") + + return max(0.0, score) + + except Exception as e: + print(f"Error in financial ratio analysis: {e}") + return 0.5 # Default moderate score if analysis fails + # For now, we'll use available basic metrics and estimate + + # PE ratio check (very high PE might indicate speculation) + pe_ratio = stock_info.get('pe_ratio', 0) + if pe_ratio > 50: + score -= 0.2 + elif pe_ratio > 30: + score -= 0.1 + + # Beta check (high beta indicates high speculation/volatility) + beta = stock_info.get('beta', 1.0) + if beta > 2.0: + score -= 0.3 + elif beta > 1.5: + score -= 0.1 + + # Volatility check from analysis + volatility = analysis.get('volatility_annualized', 0) + if volatility > 60: + score -= 0.2 + elif volatility > 40: + score -= 0.1 + + return max(0.0, score) + + def _check_debt_levels(self, stock_info: Dict) -> float: + """ + Check debt levels according to Sharia standards + + Returns: + Score from 0.0 to 1.0 + """ + # Note: In a real implementation, you would fetch debt-to-assets ratio + # For now, we'll use sector-based estimation + + sector = stock_info.get('sector', '').lower() + + # Sectors typically with high debt + high_debt_sectors = ['utilities', 'real estate', 'telecommunications'] + medium_debt_sectors = ['manufacturing', 'transportation', 'energy'] + low_debt_sectors = ['technology', 'healthcare', 'retail'] + + if any(s in sector for s in high_debt_sectors): + return 0.6 # Assume higher debt but may still be acceptable + elif any(s in sector for s in medium_debt_sectors): + return 0.8 + elif any(s in sector for s in low_debt_sectors): + return 1.0 + else: + return 0.7 # Default assumption + + def _check_revenue_sources(self, stock_info: Dict) -> float: + """ + Check revenue sources for non-compliant income + + Returns: + Score from 0.0 to 1.0 + """ + sector = stock_info.get('sector', '').lower() + industry = stock_info.get('industry', '').lower() + + # Industries with potential non-compliant revenue + if 'financial' in sector or 'bank' in industry: + return 0.3 # Banks typically have significant interest income + elif 'insurance' in industry: + return 0.2 + elif 'hotel' in industry or 'entertainment' in industry: + return 0.6 # May have some non-compliant revenue sources + else: + return 0.9 # Assume mostly compliant revenue + + def _get_ai_sharia_analysis(self, symbol: str, stock_info: Dict) -> Dict: + """ + Get AI-powered detailed Sharia compliance analysis + + Args: + symbol: Stock symbol + stock_info: Stock information + + Returns: + Dictionary with AI analysis + """ + try: + prompt = f""" + As an Islamic finance expert, analyze the Sharia compliance of {symbol}. + + Company Information: + - Name: {stock_info.get('company_name', 'N/A')} + - Sector: {stock_info.get('sector', 'N/A')} + - Industry: {stock_info.get('industry', 'N/A')} + - Country: {stock_info.get('country', 'N/A')} + + Please analyze according to Islamic finance principles and provide: + 1. Primary business activity assessment + 2. Potential Sharia compliance concerns + 3. Revenue source analysis + 4. Debt and interest exposure concerns + 5. Overall compliance recommendation + 6. Specific areas requiring scholar consultation + 7. Alternative Sharia-compliant investment suggestions + + Format your response as JSON: + {{ + "compliance_status": "HALAL/HARAM/DOUBTFUL", + "confidence": 85, + "analysis": "Detailed analysis...", + "concerns": ["concern1", "concern2"], + "recommendations": ["rec1", "rec2"], + "scholar_consultation": true/false, + "alternatives": ["alt1", "alt2"] + }} + """ + + response = self.client.chat.completions.create( + model="gpt-4o-mini", + messages=[ + {"role": "system", "content": "You are an expert in Islamic finance and Sharia compliance for investments."}, + {"role": "user", "content": prompt} + ], + temperature=0.2, + max_tokens=1000 + ) + + ai_response = response.choices[0].message.content + + try: + if ai_response: + return json.loads(ai_response) + else: + return {'analysis': 'No AI response received', 'error': 'Empty response'} + except json.JSONDecodeError: + return {'analysis': ai_response, 'parsed_fallback': True} + + except Exception as e: + return {'analysis': f'AI analysis unavailable: {str(e)}', 'error': str(e)} + + def _determine_ruling(self, compliance_score: float, ai_analysis: Dict) -> Dict: + """ + Determine final Sharia compliance ruling + + Args: + compliance_score: Calculated compliance score + ai_analysis: AI analysis results + + Returns: + Dictionary with final ruling + """ + # Get AI recommendation if available + ai_status = ai_analysis.get('compliance_status', 'DOUBTFUL') + ai_confidence = ai_analysis.get('confidence', 50) / 100 + + # Combine algorithmic score with AI analysis + if compliance_score >= 0.8 and ai_status == 'HALAL': + status = 'HALAL' + confidence = min(0.9, (compliance_score + ai_confidence) / 2) + reasoning = "Company appears to be Sharia-compliant based on business activities and financial structure." + elif compliance_score <= 0.3 or ai_status == 'HARAM': + status = 'HARAM' + confidence = max(0.7, (1 - compliance_score + ai_confidence) / 2) + reasoning = "Company has significant Sharia compliance issues and should be avoided." + else: + status = 'DOUBTFUL' + confidence = 0.6 + reasoning = "Company has mixed compliance indicators. Consultation with Islamic scholars recommended." + + return { + 'status': status, + 'confidence': confidence, + 'reasoning': reasoning, + 'concerns': ai_analysis.get('concerns', []), + 'recommendations': ai_analysis.get('recommendations', []), + 'scholar_consultation': ai_analysis.get('scholar_consultation', status == 'DOUBTFUL'), + 'alternatives': ai_analysis.get('alternatives', []) + } + + def get_compliance_summary(self, compliance_result: Dict) -> str: + """Generate a human-readable compliance summary""" + if 'error' in compliance_result: + return f"Compliance Analysis Error: {compliance_result['error']}" + + symbol = compliance_result.get('symbol', 'Unknown') + ruling = compliance_result.get('ruling', 'UNCERTAIN') + confidence = compliance_result.get('confidence', 0) * 100 + + summary = [f"Sharia Compliance Analysis for {symbol}"] + summary.append(f"Ruling: {ruling} (Confidence: {confidence:.0f}%)") + + if ruling == 'HALAL': + summary.append("✅ This investment appears to be permissible under Islamic law.") + elif ruling == 'HARAM': + summary.append("❌ This investment should be avoided due to Sharia non-compliance.") + else: + summary.append("⚠️ This investment requires further investigation and scholar consultation.") + + # Add key concerns if any + concerns = compliance_result.get('key_concerns', []) + if concerns: + summary.append(f"Key Concerns: {', '.join(concerns)}") + + # Add recommendations + recommendations = compliance_result.get('recommendations', []) + if recommendations: + summary.append(f"Recommendations: {', '.join(recommendations[:2])}") + + return "\n".join(summary) + + def get_sharia_alternatives(self, sector: str, country: str = 'USA') -> List[str]: + """ + Get Sharia-compliant alternatives in the same sector + + Args: + sector: Company sector + country: Market country + + Returns: + List of alternative stock symbols + """ + # This would typically connect to a Sharia-compliant stock database + # For now, return some common Sharia-compliant stocks by sector + + alternatives = { + 'technology': ['AAPL', 'MSFT', 'GOOGL', 'META'], + 'healthcare': ['JNJ', 'PFE', 'UNH', 'ABBV'], + 'consumer': ['PG', 'KO', 'PEP', 'WMT'], + 'industrial': ['BA', 'CAT', 'GE', 'MMM'] + } + + sector_lower = sector.lower() + for key, stocks in alternatives.items(): + if key in sector_lower: + return stocks[:3] # Return top 3 alternatives + + return [] + +# Global instance for easy import +sharia_checker = ShariaComplianceChecker() + +# Convenience function +def check_sharia_compliance(symbol: str, stock_info: Dict, analysis: Dict) -> Dict: + """Convenience function to check Sharia compliance""" + return sharia_checker.check_sharia_compliance(symbol, stock_info, analysis) diff --git a/week4/community-contributions/ai_stock_trading/tools/trading_decisions.py b/week4/community-contributions/ai_stock_trading/tools/trading_decisions.py new file mode 100644 index 0000000..05b4255 --- /dev/null +++ b/week4/community-contributions/ai_stock_trading/tools/trading_decisions.py @@ -0,0 +1,491 @@ +""" +Trading Decisions Module + +This module provides AI-powered trading recommendations using OpenAI +and advanced algorithmic decision-making based on technical analysis. +""" + +import os +import json +from typing import Dict, List, Optional, Tuple +import pandas as pd +import numpy as np +from openai import OpenAI +from dotenv import load_dotenv + +# Load environment variables +load_dotenv() + +class TradingDecisionEngine: + """Enhanced trading decision engine with AI and algorithmic analysis""" + + def __init__(self): + self.client = OpenAI(api_key=os.getenv('OPENAI_API_KEY')) + + # Trading signal weights + self.signal_weights = { + 'trend': 0.25, + 'momentum': 0.20, + 'volume': 0.15, + 'volatility': 0.15, + 'technical': 0.25 + } + + def get_trading_recommendation(self, symbol: str, analysis: Dict, stock_info: Dict) -> Dict: + """ + Get comprehensive trading recommendation + + Args: + symbol: Stock symbol + analysis: Technical analysis results + stock_info: Stock information + + Returns: + Dictionary with trading recommendation + """ + try: + # Get algorithmic score + algo_decision = self._get_algorithmic_decision(analysis) + + # Get AI-powered recommendation + ai_decision = self._get_ai_recommendation(symbol, analysis, stock_info) + + # Combine decisions + final_decision = self._combine_decisions(algo_decision, ai_decision) + + return { + 'symbol': symbol, + 'recommendation': final_decision['action'], + 'confidence': final_decision['confidence'], + 'price_target': final_decision.get('price_target'), + 'stop_loss': final_decision.get('stop_loss'), + 'reasoning': final_decision['reasoning'], + 'algorithmic_score': algo_decision['score'], + 'ai_recommendation': ai_decision['recommendation'], + 'risk_level': self._assess_risk_level(analysis), + 'time_horizon': final_decision.get('time_horizon', 'medium'), + 'key_factors': final_decision.get('key_factors', []) + } + + except Exception as e: + return { + 'symbol': symbol, + 'recommendation': 'HOLD', + 'confidence': 0.5, + 'reasoning': f'Error in analysis: {str(e)}', + 'error': str(e) + } + + def _get_algorithmic_decision(self, analysis: Dict) -> Dict: + """ + Generate algorithmic trading decision based on technical indicators + + Args: + analysis: Technical analysis results + + Returns: + Dictionary with algorithmic decision + """ + signals = {} + + # Trend signals + trend_score = self._calculate_trend_signal(analysis) + signals['trend'] = trend_score + + # Momentum signals + momentum_score = self._calculate_momentum_signal(analysis) + signals['momentum'] = momentum_score + + # Volume signals + volume_score = self._calculate_volume_signal(analysis) + signals['volume'] = volume_score + + # Volatility signals + volatility_score = self._calculate_volatility_signal(analysis) + signals['volatility'] = volatility_score + + # Technical indicator signals + technical_score = self._calculate_technical_signal(analysis) + signals['technical'] = technical_score + + # Calculate weighted score + total_score = sum(signals[key] * self.signal_weights[key] for key in signals) + + # Determine action + if total_score >= 0.6: + action = 'BUY' + elif total_score <= -0.6: + action = 'SELL' + else: + action = 'HOLD' + + return { + 'action': action, + 'score': total_score, + 'signals': signals, + 'confidence': min(abs(total_score), 1.0) + } + + def _calculate_trend_signal(self, analysis: Dict) -> float: + """Calculate trend-based signal (-1 to 1)""" + score = 0.0 + + # Trend direction and strength + if analysis.get('trend_direction') == 'upward': + score += 0.5 + elif analysis.get('trend_direction') == 'downward': + score -= 0.5 + + # Trend strength + trend_strength = analysis.get('trend_strength', 0) + score *= trend_strength + + # Moving average signals + if 'price_vs_sma_20' in analysis: + sma_20_signal = analysis['price_vs_sma_20'] + if sma_20_signal > 2: + score += 0.2 + elif sma_20_signal < -2: + score -= 0.2 + + if 'price_vs_sma_50' in analysis: + sma_50_signal = analysis['price_vs_sma_50'] + if sma_50_signal > 2: + score += 0.3 + elif sma_50_signal < -2: + score -= 0.3 + + return max(-1.0, min(1.0, score)) + + def _calculate_momentum_signal(self, analysis: Dict) -> float: + """Calculate momentum-based signal (-1 to 1)""" + score = 0.0 + + # RSI signal + if 'rsi' in analysis: + rsi = analysis['rsi'] + if rsi < 30: + score += 0.4 # Oversold - potential buy + elif rsi > 70: + score -= 0.4 # Overbought - potential sell + + # MACD signal + if 'macd_trend' in analysis: + if analysis['macd_trend'] == 'bullish': + score += 0.3 + else: + score -= 0.3 + + # Recent performance + if 'return_1_week' in analysis: + weekly_return = analysis['return_1_week'] + if weekly_return > 5: + score += 0.2 + elif weekly_return < -5: + score -= 0.2 + + return max(-1.0, min(1.0, score)) + + def _calculate_volume_signal(self, analysis: Dict) -> float: + """Calculate volume-based signal (-1 to 1)""" + score = 0.0 + + # Volume trend + if analysis.get('volume_trend') == 'increasing': + score += 0.3 + elif analysis.get('volume_trend') == 'decreasing': + score -= 0.2 + + # Volume vs average + if 'volume_vs_avg' in analysis: + vol_vs_avg = analysis['volume_vs_avg'] + if vol_vs_avg > 20: + score += 0.2 + elif vol_vs_avg < -20: + score -= 0.1 + + return max(-1.0, min(1.0, score)) + + def _calculate_volatility_signal(self, analysis: Dict) -> float: + """Calculate volatility-based signal (-1 to 1)""" + score = 0.0 + + # High volatility can be both opportunity and risk + volatility = analysis.get('volatility_annualized', 0) + + if volatility > 50: + score -= 0.3 # High risk + elif volatility < 15: + score += 0.2 # Low risk + + # Max drawdown consideration + max_drawdown = analysis.get('max_drawdown', 0) + if abs(max_drawdown) > 20: + score -= 0.2 + + return max(-1.0, min(1.0, score)) + + def _calculate_technical_signal(self, analysis: Dict) -> float: + """Calculate technical indicator signal (-1 to 1)""" + score = 0.0 + + # Bollinger Bands + if 'bb_position' in analysis: + bb_pos = analysis['bb_position'] + if bb_pos == 'below_lower_band': + score += 0.3 # Potential buy + elif bb_pos == 'above_upper_band': + score -= 0.3 # Potential sell + + # Sharpe ratio + sharpe = analysis.get('sharpe_ratio', 0) + if sharpe > 1: + score += 0.2 + elif sharpe < 0: + score -= 0.2 + + return max(-1.0, min(1.0, score)) + + def _get_ai_recommendation(self, symbol: str, analysis: Dict, stock_info: Dict) -> Dict: + """ + Get AI-powered trading recommendation using OpenAI + + Args: + symbol: Stock symbol + analysis: Technical analysis results + stock_info: Stock information + + Returns: + Dictionary with AI recommendation + """ + try: + # Prepare analysis data for AI + analysis_summary = self._prepare_analysis_for_ai(analysis, stock_info) + + prompt = f""" + You are a senior financial analyst with 15+ years of experience providing institutional-grade trading recommendations. + Analyze {symbol} and provide a professional trading recommendation. + + Company Information: + - Name: {stock_info.get('company_name', 'N/A')} + - Sector: {stock_info.get('sector', 'N/A')} + - Market Cap: ${stock_info.get('market_cap', 0):,} + + Technical Analysis Data: + {analysis_summary} + + REQUIREMENTS: + 1. Provide BUY/HOLD/SELL recommendation based on technical analysis + 2. Set realistic confidence level (60-95% range) + 3. Calculate logical price targets using support/resistance levels + 4. Set appropriate stop-loss levels (5-15% below entry for long positions) + 5. Consider risk-reward ratios (minimum 1:2 ratio preferred) + 6. Provide clear, actionable reasoning without jargon + 7. Consider market conditions and sector trends + + TRADING STANDARDS: + - BUY: Strong upward momentum, good risk/reward, clear catalysts + - HOLD: Consolidation phase, mixed signals, or fair value + - SELL: Downward trend, poor fundamentals, or overvalued + + Return ONLY valid JSON: + {{ + "recommendation": "BUY/HOLD/SELL", + "confidence": 85, + "price_target": 150.00, + "stop_loss": 120.00, + "time_horizon": "short/medium/long", + "reasoning": "Professional analysis explaining the recommendation with specific technical factors", + "key_factors": ["specific technical indicator", "market condition", "risk factor"], + "risk_assessment": "low/medium/high" + }} + """ + + response = self.client.chat.completions.create( + model="gpt-4o-mini", + messages=[ + {"role": "system", "content": "You are an expert financial analyst providing professional trading recommendations."}, + {"role": "user", "content": prompt} + ], + temperature=0.3, + max_tokens=1000 + ) + + # Parse AI response + ai_response = response.choices[0].message.content + + if ai_response: + try: + # Clean the response - extract JSON + json_start = ai_response.find('{') + json_end = ai_response.rfind('}') + 1 + + if json_start != -1 and json_end != -1: + json_str = ai_response[json_start:json_end] + ai_recommendation = json.loads(json_str) + + # Validate and clean the response + return { + 'recommendation': ai_recommendation.get('recommendation', 'HOLD'), + 'confidence': ai_recommendation.get('confidence', 50), + 'price_target': ai_recommendation.get('price_target'), + 'stop_loss': ai_recommendation.get('stop_loss'), + 'time_horizon': ai_recommendation.get('time_horizon', 'medium'), + 'reasoning': ai_recommendation.get('reasoning', 'AI analysis completed'), + 'key_factors': ai_recommendation.get('key_factors', []), + 'risk_assessment': ai_recommendation.get('risk_assessment', 'medium') + } + else: + # No JSON found, use fallback + return self._parse_ai_response_fallback(ai_response) + + except json.JSONDecodeError: + # Fallback parsing + return self._parse_ai_response_fallback(ai_response) + else: + return self._parse_ai_response_fallback('No response received') + + except Exception as e: + return { + 'recommendation': 'HOLD', + 'confidence': 50, + 'reasoning': f'AI analysis unavailable: {str(e)}', + 'error': str(e) + } + + def _prepare_analysis_for_ai(self, analysis: Dict, stock_info: Dict) -> str: + """Prepare analysis summary for AI consumption""" + summary_parts = [] + + # Price metrics + current_price = analysis.get('current_price', 0) + total_return = analysis.get('total_return_pct', 0) + summary_parts.append(f"Current Price: ${current_price:.2f}") + summary_parts.append(f"Total Return: {total_return:.2f}%") + + # Trend analysis + trend_dir = analysis.get('trend_direction', 'unknown') + trend_strength = analysis.get('trend_strength', 0) + summary_parts.append(f"Trend: {trend_dir} (strength: {trend_strength:.2f})") + + # Technical indicators + if 'rsi' in analysis: + rsi = analysis['rsi'] + rsi_signal = analysis.get('rsi_signal', 'neutral') + summary_parts.append(f"RSI: {rsi:.1f} ({rsi_signal})") + + if 'macd_trend' in analysis: + summary_parts.append(f"MACD: {analysis['macd_trend']}") + + # Risk metrics + volatility = analysis.get('volatility_annualized', 0) + max_drawdown = analysis.get('max_drawdown', 0) + summary_parts.append(f"Volatility: {volatility:.1f}% (annual)") + summary_parts.append(f"Max Drawdown: {max_drawdown:.1f}%") + + # Performance + if 'return_1_month' in analysis: + monthly_return = analysis['return_1_month'] + summary_parts.append(f"1-Month Return: {monthly_return:.2f}%") + + return "\n".join(summary_parts) + + def _parse_ai_response_fallback(self, response: str) -> Dict: + """Fallback parser for AI response if JSON parsing fails""" + # Simple keyword-based parsing + recommendation = 'HOLD' + confidence = 50 + + response_lower = response.lower() + + if 'buy' in response_lower and 'sell' not in response_lower: + recommendation = 'BUY' + confidence = 70 + elif 'sell' in response_lower: + recommendation = 'SELL' + confidence = 70 + + return { + 'recommendation': recommendation, + 'confidence': confidence, + 'reasoning': response, + 'parsed_fallback': True + } + + def _combine_decisions(self, algo_decision: Dict, ai_decision: Dict) -> Dict: + """Combine algorithmic and AI decisions""" + # Weight the decisions (60% algorithmic, 40% AI) + algo_weight = 0.6 + ai_weight = 0.4 + + # Map recommendations to scores + rec_scores = {'BUY': 1, 'HOLD': 0, 'SELL': -1} + + algo_score = rec_scores.get(algo_decision['action'], 0) + ai_score = rec_scores.get(ai_decision.get('recommendation', 'HOLD'), 0) + + # Calculate combined score + combined_score = (algo_score * algo_weight) + (ai_score * ai_weight) + + # Determine final recommendation + if combined_score >= 0.3: + final_action = 'BUY' + elif combined_score <= -0.3: + final_action = 'SELL' + else: + final_action = 'HOLD' + + # Calculate confidence + algo_confidence = algo_decision.get('confidence', 0.5) + ai_confidence = ai_decision.get('confidence', 50) / 100 + combined_confidence = (algo_confidence * algo_weight) + (ai_confidence * ai_weight) + + return { + 'action': final_action, + 'confidence': combined_confidence, + 'combined_score': combined_score, + 'reasoning': ai_decision.get('reasoning', 'Combined algorithmic and AI analysis'), + 'price_target': ai_decision.get('price_target'), + 'stop_loss': ai_decision.get('stop_loss'), + 'time_horizon': ai_decision.get('time_horizon', 'medium'), + 'key_factors': ai_decision.get('key_factors', []) + } + + def _assess_risk_level(self, analysis: Dict) -> str: + """Assess overall risk level""" + risk_score = 0 + + # Volatility risk + volatility = analysis.get('volatility_annualized', 0) + if volatility > 40: + risk_score += 2 + elif volatility > 25: + risk_score += 1 + + # Drawdown risk + max_drawdown = abs(analysis.get('max_drawdown', 0)) + if max_drawdown > 30: + risk_score += 2 + elif max_drawdown > 15: + risk_score += 1 + + # Sharpe ratio + sharpe = analysis.get('sharpe_ratio', 0) + if sharpe < 0: + risk_score += 1 + + # Determine risk level + if risk_score >= 4: + return 'high' + elif risk_score >= 2: + return 'medium' + else: + return 'low' + +# Global instance for easy import +trading_engine = TradingDecisionEngine() + +# Convenience function +def get_trading_recommendation(symbol: str, analysis: Dict, stock_info: Dict) -> Dict: + """Convenience function to get trading recommendation""" + return trading_engine.get_trading_recommendation(symbol, analysis, stock_info) diff --git a/week4/community-contributions/c_extension_generator/LICENSE b/week4/community-contributions/c_extension_generator/LICENSE new file mode 100644 index 0000000..0ca1b61 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/LICENSE @@ -0,0 +1,21 @@ +MIT License + +Copyright (c) 2025 Carlos Bazaga + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/week4/community-contributions/c_extension_generator/README.md b/week4/community-contributions/c_extension_generator/README.md new file mode 100644 index 0000000..87d25db --- /dev/null +++ b/week4/community-contributions/c_extension_generator/README.md @@ -0,0 +1,134 @@ +# Python C Extension code generator + +Written by Carlos Bazaga [@carbaz] based on the work of Ed Donner [@ed-donner] +under the MIT License. + +This folder contains a Jupyter notebook that demonstrates how to use a Frontier model +to generate high-performance Python C extension code from Python code. + +The notebook includes examples of generating C extensions for calculating Pi using the +Leibniz formula and finding the maximum sub-array in an array. + +Also, it includes a Gradio app that provides an interactive interface for users to input +Python code, generate C extension code, compile it, and test its performance against +the original Python code. + +> [!CAUTION] +> +> **Always review the generated codes before running them, as they will be executed in +> your local environment and may contain code that could be harmful or unwanted.** +> +> AI-generated code may contain errors or unsafe practices, so it's crucial to +> thoroughly review and test any code before using it in a production environment. +> +> Never run code generated by AI models without understanding its implications and +> ensuring it adheres to your security and safety standards. + +> [!IMPORTANT] +> +> **Disclaimer:** This notebook and the Gradio app are provided for educational purposes +> only. Use them at your own risk. + +## Gradio app overview + +In this image, you can see the Gradio app dashboard whose main sections are +described below. + +![Gradio app dashboard](gradio_dashboard.jpg)\ +*Image: Gradio app dashboard with default example `hello world` code loaded.* +*(compile output redacted for privacy)* + +Sections: + +* **Dropdown selectors and input fields**: + * **Module name input**: + A text input field where users can specify the name of the C extension module to be + generated. + + That name will be used to create the C extension file `.c` and + the `setup.py` file required to compile the extension. + + That name will also be used to import the compiled module as usual in Python: + + ```python + import + ``` + + Or + + ```python + from import + ``` + + * **Model selector**: + A dropdown menu to select the Frontier model to use for code generation. + + Currently it includes: + * `gpt-4o` (default) + * `gpt-5` + +* **Text input areas**: + + This areas are all editable, included those filled with generated code by the model. + this allows users to modify and experiment with the code as needed. + + * **Python code**: + A text area where users can input their Python code. + * **C extension code**: + A text area that displays the generated C extension code and allows to edit it. + * **Compilation code**: + A text area that shows the `setup.py` file generated, + this file is required to compile the C extension. + * **Test compare code**: + A text area that provides example code to run the compiled C extension. + +* **Output areas**: + + This are non-editable areas that display the results of various operations. + + * **C Extension result**: + A text area that displays the output of the C extension code build. + + Beware that this area can contain a large amount of text including warnings during + the compilation process and sensible information about the local environment, + like: paths, Python version, etc may be included. + + Redact that information if you plan to share the output. + + * **Test result**: + A text area that displays the output of the test code run. + +* **Buttons**: + * **Generate extension code**: + A button that triggers the generation of the C extension code from the provided + Python code. + + It will call the Frontier model to generate the C code, the setup.py file and + the test code, filling the corresponding text areas automatically. + + * **Compile extension**: + A button that compiles the generated C extension using the provided `setup.py` file. + It will create the extension c file, `.c` and the `setup.py` files in + the local folder, then it will run the compilation command and build the C extension. + + > [!CAUTION] + > + > **Always review the `setup.py` code before running it, as it will be executed in + > your local environment and may contain code that could be harmful or unwanted.** + > + > **Also review the generated C code, as it will be compiled and executed in your + > local environment and may contain code that could be harmful or unwanted.** + + It will display the compilation output in the "C Extension result" area. + + * **Test code**: + A button that executes the test code to compare the performance of the original + Python code and the generated C extension. + + > [!CAUTION] + > + > **Always review the test code before running it, as it will be executed in + > your local environment and may contain code that could be harmful or unwanted.** + + Will save the test code provided in the "Test compare code" into the + `usage_example.py` file and execute it showing the output in the "Test result" area. diff --git a/week4/community-contributions/c_extension_generator/calculate_pi.c b/week4/community-contributions/c_extension_generator/calculate_pi.c new file mode 100644 index 0000000..c0b7f03 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/calculate_pi.c @@ -0,0 +1,83 @@ +#define PY_SSIZE_T_CLEAN +#include +#include +#include +#include +#include + +static PyObject* leibniz_pi(PyObject* self, PyObject* args) { + PyObject* iterations_obj; + if (!PyArg_ParseTuple(args, "O", &iterations_obj)) { + return NULL; + } + + long long n_signed; + int overflow = 0; + n_signed = PyLong_AsLongLongAndOverflow(iterations_obj, &overflow); + if (n_signed == -1 && PyErr_Occurred() && overflow == 0) { + return NULL; + } + + unsigned long long n = 0ULL; + if (overflow < 0) { + n = 0ULL; + } else if (overflow > 0) { + unsigned long long tmp = PyLong_AsUnsignedLongLong(iterations_obj); + if (tmp == (unsigned long long)-1 && PyErr_Occurred()) { + return NULL; + } + n = tmp; + } else { + if (n_signed <= 0) { + n = 0ULL; + } else { + n = (unsigned long long)n_signed; + } + } + + double result = 1.0; + if (n == 0ULL) { + return PyFloat_FromDouble(result * 4.0); + } + + Py_BEGIN_ALLOW_THREADS + for (unsigned long long i = 1ULL; i <= n; ++i) { + double jd1; + if (i <= ULLONG_MAX / 4ULL) { + unsigned long long j1 = i * 4ULL - 1ULL; + jd1 = (double)j1; + } else { + jd1 = (double)i * 4.0 - 1.0; + } + result -= 1.0 / jd1; + + double jd2; + if (i <= (ULLONG_MAX - 1ULL) / 4ULL) { + unsigned long long j2 = i * 4ULL + 1ULL; + jd2 = (double)j2; + } else { + jd2 = (double)i * 4.0 + 1.0; + } + result += 1.0 / jd2; + } + Py_END_ALLOW_THREADS + + return PyFloat_FromDouble(result * 4.0); +} + +static PyMethodDef CalculatePiMethods[] = { + {"leibniz_pi", leibniz_pi, METH_VARARGS, "Compute pi using the Leibniz series with the given number of iterations."}, + {NULL, NULL, 0, NULL} +}; + +static struct PyModuleDef calculate_pimodule = { + PyModuleDef_HEAD_INIT, + "calculate_pi", + "High-performance Leibniz pi calculation.", + -1, + CalculatePiMethods +}; + +PyMODINIT_FUNC PyInit_calculate_pi(void) { + return PyModule_Create(&calculate_pimodule); +} diff --git a/week4/community-contributions/c_extension_generator/gradio_dashboard.jpg b/week4/community-contributions/c_extension_generator/gradio_dashboard.jpg new file mode 100644 index 0000000..7b07689 Binary files /dev/null and b/week4/community-contributions/c_extension_generator/gradio_dashboard.jpg differ diff --git a/week4/community-contributions/c_extension_generator/python_c_ext_generator.ipynb b/week4/community-contributions/c_extension_generator/python_c_ext_generator.ipynb new file mode 100644 index 0000000..65b480a --- /dev/null +++ b/week4/community-contributions/c_extension_generator/python_c_ext_generator.ipynb @@ -0,0 +1,1616 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "4a6ab9a2-28a2-445d-8512-a0dc8d1b54e9", + "metadata": {}, + "source": [ + "# Python C extension generator\n", + "\n", + "Use a Frontier model to generate a high performance Python C extension code from Python code.\n", + "\n", + "Python C extension modules allows to integrate C coded and compiled modules into Python applications.\n", + "\n", + "* [Python C Extensions](https://docs.python.org/3.13/extending/index.html)\n", + "* [Python's C API](https://docs.python.org/3.13/c-api/index.html)" + ] + }, + { + "cell_type": "markdown", + "id": "d90e04a2-5b8a-4fd5-9db8-27c02f033313", + "metadata": {}, + "source": [ + "\n", + " \n", + " \n", + " \n", + " \n", + "
\n", + " \n", + " \n", + "

Important Note

\n", + " \n", + " In this lab, I use GPT-4o or GPT-5, which are slightly higher priced models.\n", + " \n", + "
" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "e610bf56-a46e-4aff-8de1-ab49d62b1ad3", + "metadata": {}, + "outputs": [], + "source": [ + "# Imports.\n", + "\n", + "import io\n", + "import os\n", + "import subprocess\n", + "import sys\n", + "from time import perf_counter\n", + "from timeit import timeit\n", + "\n", + "import gradio as gr\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "from pydantic import BaseModel" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "4f672e1c-87e9-4865-b760-370fa605e614", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "True" + ] + }, + "execution_count": 2, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# Load environment variables from '.env' file.\n", + "\n", + "load_dotenv(override=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "8aa149ed-9298-4d69-8fe2-8f5de0f667da", + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize client and set the default LLM model to use.\n", + "\n", + "# OPENAI_MODEL = \"gpt-4o\"\n", + "OPENAI_MODEL = \"gpt-5\"\n", + "\n", + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 4, + "id": "c6f37bf0", + "metadata": {}, + "outputs": [], + "source": [ + "# Define Pydantic model class for GPT response parsing.\n", + "\n", + "class Extension_codes(BaseModel):\n", + " \"\"\"Pydantic model of a response containing the generated C code, the 'setup.py' code and an usage example.\"\"\"\n", + " c_code: str\n", + " setup: str\n", + " usage: str" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "cb6ce77a", + "metadata": {}, + "outputs": [], + "source": [ + "# Define a function to print the optimization codes.\n", + "\n", + "def print_optimization(optimization):\n", + " \"\"\"Print the optimization codes.\"\"\"\n", + " print(f\"C CODE:\\n{optimization.c_code}\")\n", + " print(\"---------------------------\")\n", + " print(f\"setup.py:\\n{optimization.setup}\")\n", + " print(\"---------------------------\")\n", + " print(f\"USAGE:\\n{optimization.usage}\")" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "71e1ba8c-5b05-4726-a9f3-8d8c6257350b", + "metadata": {}, + "outputs": [], + "source": [ + "# Define a function to write outputs to a file with a given filename.\n", + "\n", + "def write_file(data, filename):\n", + " \"\"\"Write data to a file with the specified filename.\"\"\"\n", + " with open(filename, \"w\") as file:\n", + " file.write(data)" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "f13c9c97", + "metadata": {}, + "outputs": [], + "source": [ + "# Define a function to write the optimization codes to files.\n", + "\n", + "def write_optimization(optimization, module_name):\n", + " \"\"\"Write the optimization codes to files.\"\"\"\n", + " write_file(optimization.c_code, f\"{module_name}.c\")\n", + " write_file(optimization.setup, \"setup.py\")\n", + " write_file(optimization.usage, \"usage_example.py\")" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "6896636f-923e-4a2c-9d6c-fac07828a201", + "metadata": {}, + "outputs": [], + "source": [ + "# Define system message for the LLM with instructions for generating the C extension code.\n", + "\n", + "system_message = \"\"\"\n", + "You are an assistant that reimplements Python code in high performance C extensions for Python.\n", + "Your responses must always be a JSON with the following structure:\n", + "\n", + "{\n", + " \"c_code\": \"Optimized C extension for Python code\",\n", + " \"setup\": \"The 'setup.py' code to compile the C extension for Python\",\n", + " \"usage\": \"An example of usage of the C extension for Python code with time measurement and comparing with the original Python code\"\n", + "}\n", + "\n", + "Use comments sparingly and do not provide any explanation other than occasional comments.\n", + "The C extension for Python needs to produce an identical output in the fastest possible time.\n", + "Make sure the C extension for Python code is correct and can be compiled with 'python setup.py build' and used in Python.\n", + "The usage example must include a time measurement and a comparison with the original Python code.\n", + "Do not include any additional text or explanation outside the JSON structure.\n", + "Make sure the JSON is correctly formatted.\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "8e7b3546-57aa-4c29-bc5d-f211970d04eb", + "metadata": {}, + "outputs": [], + "source": [ + "# Define user prompt template and function to fill it.\n", + "\n", + "def user_prompt_for(python_code, module_name):\n", + " user_prompt = f\"\"\"\n", + " Reimplement this Python code as a C extension for Python with the fastest possible implementation that produces identical output in the least time.\n", + " Respond only with C extension for Python code, do not explain your work other than a few code comments.\n", + " The module name, used to import, must be \"{module_name}\", the generated C file will be named \"{module_name}.c\".\n", + " Pay attention to number types to ensure no int overflows.\n", + " Remember to #include all necessary C packages such as iomanip or \n", + "\n", + " The target architecture is {sys.platform}, take that in mind while generating the C code, specially\n", + " when choosing types to use, and use the appropriate compiler flags.\n", + " Make sure to use the Python C API correctly and manage memory properly to avoid leaks or crashes.\n", + "\n", + " Here is the Python code to reimplement:\n", + "\n", + " {python_code}\"\"\"\n", + " return user_prompt" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "c6190659-f54c-4951-bef4-4960f8e51cc4", + "metadata": {}, + "outputs": [], + "source": [ + "# Define function to create the messages for the LLM.\n", + "\n", + "def messages_for(python_code, module_name):\n", + " \"\"\"Create the messages for the LLM given the Python code and the desired module name.\"\"\"\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(python_code, module_name)}]" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "3c57bc55", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "SYSTEM: \n", + "You are an assistant that reimplements Python code in high performance C extensions for Python.\n", + "Your responses must always be a JSON with the following structure:\n", + "\n", + "{\n", + " \"c_code\": \"Optimized C extension for Python code\",\n", + " \"setup\": \"The 'setup.py' code to compile the C extension for Python\",\n", + " \"usage\": \"An example of usage of the C extension for Python code with time measurement and comparing with the original Python code\"\n", + "}\n", + "\n", + "Use comments sparingly and do not provide any explanation other than occasional comments.\n", + "The C extension for Python needs to produce an identical output in the fastest possible time.\n", + "Make sure the C extension for Python code is correct and can be compiled with 'python setup.py build' and used in Python.\n", + "The usage example must include a time measurement and a comparison with the original Python code.\n", + "Do not include any additional text or explanation outside the JSON structure.\n", + "Make sure the JSON is correctly formatted.\n", + "\n", + "--------------------------------\n", + "USER: \n", + " Reimplement this Python code as a C extension for Python with the fastest possible implementation that produces identical output in the least time.\n", + " Respond only with C extension for Python code, do not explain your work other than a few code comments.\n", + " The module name, used to import, must be \"say_hello\", the generated C file will be named \"say_hello.c\".\n", + " Pay attention to number types to ensure no int overflows.\n", + " Remember to #include all necessary C packages such as iomanip or \n", + "\n", + " The target architecture is win32, take that in mind while generating the C code, specially\n", + " when choosing types to use, and use the appropriate compiler flags.\n", + " Make sure to use the Python C API correctly and manage memory properly to avoid leaks or crashes.\n", + "\n", + " Here is the Python code to reimplement:\n", + "\n", + " print('Hello World')\n", + "--------------------------------\n" + ] + } + ], + "source": [ + "# Test the messages function and print the messages.\n", + "\n", + "for message in messages_for(\"print('Hello World')\", \"say_hello\"):\n", + " print(f\"{message['role'].upper()}: {message['content']}\")\n", + " print(\"--------------------------------\")" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "e7d2fea8-74c6-4421-8f1e-0e76d5b201b9", + "metadata": {}, + "outputs": [], + "source": [ + "# Define optimization function using OpenAI's GPT model.\n", + "\n", + "def optimize_gpt(python_code, module_name, model=OPENAI_MODEL):\n", + " \"\"\"Optimize the given Python code by generating a C extension for Python with the specified module name using the specified LLM model.\"\"\"\n", + " response = openai.chat.completions.parse(\n", + " model=model,\n", + " messages=messages_for(python_code, module_name),\n", + " response_format=Extension_codes).choices[0].message.parsed\n", + " return response" + ] + }, + { + "cell_type": "markdown", + "id": "c05b263a", + "metadata": {}, + "source": [ + "# Start with a math function that calculates ***π*** using the Leibniz formula.\n", + "\n", + "This formula implies the iterative approximation of *π* using an alternating series,\n", + "the more iterations the more the precision but with a cost of more computation.\n", + "* [Leibniz formula for π](https://en.wikipedia.org/wiki/Leibniz_formula_for_%CF%80)\n", + "\n", + "This is a good candidate to get a noticeable improvement by coding and compiling it into a Python C extension. \n", + "\n", + "> NOTE:\n", + ">\n", + "> We are creating an importable module not an executable program so the code to be optimized must contain only declarations such as DEF or CLASS." + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "a1cbb778-fa57-43de-b04b-ed523f396c38", + "metadata": {}, + "outputs": [], + "source": [ + "# Define the Python function to be converted to a C extension and its module name.\n", + "\n", + "module_name = \"calculate_pi\"\n", + "\n", + "calculate_pi_code = f\"\"\"\n", + "def leibniz_pi(iterations):\n", + " result = 1.0\n", + " for i in range(1, iterations+1):\n", + " j = i * 4 - 1\n", + " result -= (1/j)\n", + " j = i * 4 + 1\n", + " result += (1/j)\n", + " return result * 4\n", + "\"\"\"\n", + "\n", + "# Define a function to test the performance of the calculus function.\n", + "\n", + "def test_pi_calculation(calculus_function ,iterations=100_000_000):\n", + " \"\"\"Test the performance of the given calculus function.\"\"\"\n", + " start_time = perf_counter()\n", + " result = calculus_function(iterations)\n", + " end_time = perf_counter()\n", + " print(f\"Result: {result:.12f}\")\n", + " print(f\"Execution Time: {(end_time - start_time):.6f} seconds\")\n", + "\n", + "# Execute function declaration.\n", + "exec(calculate_pi_code)" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "id": "7fe1cd4b-d2c5-4303-afed-2115a3fef200", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Result: 3.141592658589\n", + "Execution Time: 20.556854 seconds\n" + ] + } + ], + "source": [ + "# Run original python code and time it.\n", + "\n", + "test_pi_calculation(leibniz_pi, 100_000_000)" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "4c0be0f2", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Timing...\n", + "Python average execution time: 21.158541 seconds\n" + ] + } + ], + "source": [ + "# Average timing the original Python code running it several times.\n", + "# (Increase 'iterations' for better timing)\n", + "\n", + "print(\"Timing...\")\n", + "iterations = 5\n", + "average = timeit(lambda: leibniz_pi(100_000_000), number=iterations) / iterations\n", + "print(f\"Python average execution time: {average:.6f} seconds\")" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "105db6f9-343c-491d-8e44-3a5328b81719", + "metadata": {}, + "outputs": [], + "source": [ + "# Request code optimization using GPT.\n", + "\n", + "optimization = optimize_gpt(calculate_pi_code, module_name)" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "378981c7", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "C CODE:\n", + "#define PY_SSIZE_T_CLEAN\n", + "#include \n", + "#include \n", + "#include \n", + "#include \n", + "#include \n", + "\n", + "static PyObject* leibniz_pi(PyObject* self, PyObject* args) {\n", + " PyObject* iterations_obj;\n", + " if (!PyArg_ParseTuple(args, \"O\", &iterations_obj)) {\n", + " return NULL;\n", + " }\n", + "\n", + " long long n_signed;\n", + " int overflow = 0;\n", + " n_signed = PyLong_AsLongLongAndOverflow(iterations_obj, &overflow);\n", + " if (n_signed == -1 && PyErr_Occurred() && overflow == 0) {\n", + " return NULL;\n", + " }\n", + "\n", + " unsigned long long n = 0ULL;\n", + " if (overflow < 0) {\n", + " n = 0ULL;\n", + " } else if (overflow > 0) {\n", + " unsigned long long tmp = PyLong_AsUnsignedLongLong(iterations_obj);\n", + " if (tmp == (unsigned long long)-1 && PyErr_Occurred()) {\n", + " return NULL;\n", + " }\n", + " n = tmp;\n", + " } else {\n", + " if (n_signed <= 0) {\n", + " n = 0ULL;\n", + " } else {\n", + " n = (unsigned long long)n_signed;\n", + " }\n", + " }\n", + "\n", + " double result = 1.0;\n", + " if (n == 0ULL) {\n", + " return PyFloat_FromDouble(result * 4.0);\n", + " }\n", + "\n", + " Py_BEGIN_ALLOW_THREADS\n", + " for (unsigned long long i = 1ULL; i <= n; ++i) {\n", + " double jd1;\n", + " if (i <= ULLONG_MAX / 4ULL) {\n", + " unsigned long long j1 = i * 4ULL - 1ULL;\n", + " jd1 = (double)j1;\n", + " } else {\n", + " jd1 = (double)i * 4.0 - 1.0;\n", + " }\n", + " result -= 1.0 / jd1;\n", + "\n", + " double jd2;\n", + " if (i <= (ULLONG_MAX - 1ULL) / 4ULL) {\n", + " unsigned long long j2 = i * 4ULL + 1ULL;\n", + " jd2 = (double)j2;\n", + " } else {\n", + " jd2 = (double)i * 4.0 + 1.0;\n", + " }\n", + " result += 1.0 / jd2;\n", + " }\n", + " Py_END_ALLOW_THREADS\n", + "\n", + " return PyFloat_FromDouble(result * 4.0);\n", + "}\n", + "\n", + "static PyMethodDef CalculatePiMethods[] = {\n", + " {\"leibniz_pi\", leibniz_pi, METH_VARARGS, \"Compute pi using the Leibniz series with the given number of iterations.\"},\n", + " {NULL, NULL, 0, NULL}\n", + "};\n", + "\n", + "static struct PyModuleDef calculate_pimodule = {\n", + " PyModuleDef_HEAD_INIT,\n", + " \"calculate_pi\",\n", + " \"High-performance Leibniz pi calculation.\",\n", + " -1,\n", + " CalculatePiMethods\n", + "};\n", + "\n", + "PyMODINIT_FUNC PyInit_calculate_pi(void) {\n", + " return PyModule_Create(&calculate_pimodule);\n", + "}\n", + "\n", + "---------------------------\n", + "setup.py:\n", + "from setuptools import setup, Extension\n", + "import sys\n", + "import os\n", + "\n", + "extra_compile_args = []\n", + "extra_link_args = []\n", + "\n", + "if os.name == 'nt':\n", + " extra_compile_args.extend(['/O2', '/fp:precise'])\n", + "else:\n", + " extra_compile_args.extend(['-O3', '-fno-strict-aliasing'])\n", + "\n", + "module = Extension(\n", + " 'calculate_pi',\n", + " sources=['calculate_pi.c'],\n", + " extra_compile_args=extra_compile_args,\n", + " extra_link_args=extra_link_args,\n", + ")\n", + "\n", + "setup(\n", + " name='calculate_pi',\n", + " version='1.0.0',\n", + " description='High-performance C extension for computing pi via the Leibniz series',\n", + " ext_modules=[module],\n", + ")\n", + "\n", + "---------------------------\n", + "USAGE:\n", + "# Build first: python setup.py build_ext --inplace\n", + "import time\n", + "import math\n", + "import calculate_pi\n", + "\n", + "# Original Python implementation\n", + "def py_leibniz_pi(iterations):\n", + " result = 1.0\n", + " for i in range(1, iterations + 1):\n", + " j = i * 4 - 1\n", + " result -= (1 / j)\n", + " j = i * 4 + 1\n", + " result += (1 / j)\n", + " return result * 4\n", + "\n", + "iters = 5_000_000\n", + "\n", + "# Warm-up\n", + "calculate_pi.leibniz_pi(10)\n", + "py_leibniz_pi(10)\n", + "\n", + "start = time.perf_counter()\n", + "res_c = calculate_pi.leibniz_pi(iters)\n", + "end = time.perf_counter()\n", + "ctime = end - start\n", + "\n", + "start = time.perf_counter()\n", + "res_py = py_leibniz_pi(iters)\n", + "end = time.perf_counter()\n", + "pytime = end - start\n", + "\n", + "print(f\"Iterations: {iters}\")\n", + "print(f\"C extension result: {res_c}\")\n", + "print(f\"Python result: {res_py}\")\n", + "print(f\"Absolute difference: {abs(res_c - res_py)}\")\n", + "print(f\"C extension time: {ctime:.6f} s\")\n", + "print(f\"Python time: {pytime:.6f} s\")\n", + "print(f\"Speedup: {pytime/ctime if ctime > 0 else float('inf'):.2f}x\")\n", + "\n" + ] + } + ], + "source": [ + "# Print generated extension code.\n", + "\n", + "print_optimization(optimization)" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "ae9a4a64", + "metadata": {}, + "outputs": [], + "source": [ + "# Write the generated code to files.\n", + "# (Will overwrite existing files)\n", + "\n", + "write_optimization(optimization, module_name)" + ] + }, + { + "cell_type": "markdown", + "id": "bf8f8018-f64d-425c-a0e1-d7862aa9592d", + "metadata": {}, + "source": [ + "# Compiling C Extension and executing\n", + "\n", + "The python setup command may fail inside Jupyter lab, if that's the case try it directly on the command line.\n", + "\n", + "There are two cells with WINDOWS ONLY, those are to manage the fact windows comes with two command lines,\n", + "the old CMD (MS-DOS style) and the new POWERSHELL (Unix style).\n", + "\n", + "It is controlled by the COMSPEC environment variable.\\\n", + "*(Using this variable is completely innocuous on UNIX systems, they will simply ignore it)*\n", + "\n", + "Most of command lines present here are Unix style but the building one requires CMD so\n", + "we switch to CMD before compiling to later restore the preset one." + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "22a9130e", + "metadata": {}, + "outputs": [], + "source": [ + "# Clean previous builds.\n", + "# (Make sure to run this cell before running the compile cell a second time only)\n", + "# (May cast errors if no previous build exists)\n", + "\n", + "!rm -r build/" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "816e7c9d", + "metadata": {}, + "outputs": [], + "source": [ + "# [WINDOWS ONLY]\n", + "# Set COMSPEC to cmd.exe to avoid issues with some C compilers on Windows.\n", + "# (Remember to restore original COMSPEC after compilation and testing)\n", + "preset_comspec = os.environ.get(\"COMSPEC\")\n", + "os.environ[\"COMSPEC\"] = \"C:\\\\Windows\\\\System32\\\\cmd.exe\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4194e40c-04ab-4940-9d64-b4ad37c5bb40", + "metadata": {}, + "outputs": [], + "source": [ + "# Compile the C extension.\n", + "# (Will fail no C compiler is installed)\n", + "# (In case of errors, try directly on the command line)\n", + "\n", + "!python setup.py build_ext --inplace" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "id": "8db12c4d", + "metadata": {}, + "outputs": [], + "source": [ + "# [WINDOWS ONLY]\n", + "# Restore original COMSPEC.\n", + "\n", + "os.environ[\"COMSPEC\"] = preset_comspec" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "id": "a8f5169f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Iterations: 5000000\n", + "C extension result: 3.1415927535897814\n", + "Python result: 3.1415927535897814\n", + "Absolute difference: 0.0\n", + "C extension time: 0.037515 s\n", + "Python time: 1.046732 s\n", + "Speedup: 27.90x\n" + ] + } + ], + "source": [ + "# Run the usage example to test the compiled C extension.\n", + "exec(optimization.usage)" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "a1972472", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Testing original Python code:\n", + "Result: 3.141592658589\n", + "Execution Time: 20.350486 seconds\n", + "Testing C extension code:\n", + "Result: 3.141592658589\n", + "Execution Time: 0.759571 seconds\n" + ] + } + ], + "source": [ + "# Import newly created C extension and compare performance with original Python code.\n", + "\n", + "from calculate_pi import leibniz_pi as c_leibniz_pi\n", + "\n", + "print(\"Testing original Python code:\")\n", + "test_pi_calculation(leibniz_pi, 100_000_000)\n", + "print(\"Testing C extension code:\")\n", + "test_pi_calculation(c_leibniz_pi, 100_000_000)\n" + ] + }, + { + "cell_type": "markdown", + "id": "77304493", + "metadata": {}, + "source": [ + "# Lets try with a more complex code\n", + "\n", + "Now we define three functions that together implements the calculation of the \"total maximum subarray sum\"\n", + "by finding the largest sum of a contiguous subarray within a given array of numbers.\n", + "\n", + "* [Maximum subarray problem](https://en.wikipedia.org/wiki/Maximum_subarray_problem)\n", + "\n", + "This algorithm requires much more computation and steps than the previous one, we may expect a heavy\n", + "improvement by coding and compiling it into a Python C extension. \n", + "\n", + "> NOTE:\n", + ">\n", + "> We are creating an importable module not an executable program so the code to be optimized must contain only declarations such as DEF or CLASS." + ] + }, + { + "cell_type": "code", + "execution_count": 22, + "id": "c3b497b3-f569-420e-b92e-fb0f49957ce0", + "metadata": {}, + "outputs": [], + "source": [ + "# Define the Python function to be converted to a C extension and its module name.\n", + "\n", + "module_name = \"python_hard\"\n", + "\n", + "python_hard_code = \"\"\"\n", + "# Be careful to support large number sizes\n", + "\n", + "def lcg(seed, a=1664525, c=1013904223, m=2**32):\n", + " value = seed\n", + " while True:\n", + " value = (a * value + c) % m\n", + " yield value\n", + "\n", + "def max_subarray_sum(n, seed, min_val, max_val):\n", + " lcg_gen = lcg(seed)\n", + " random_numbers = [next(lcg_gen) % (max_val - min_val + 1) + min_val for _ in range(n)]\n", + " max_sum = float('-inf')\n", + " for i in range(n):\n", + " current_sum = 0\n", + " for j in range(i, n):\n", + " current_sum += random_numbers[j]\n", + " if current_sum > max_sum:\n", + " max_sum = current_sum\n", + " return max_sum\n", + "\n", + "def total_max_subarray_sum(n, initial_seed, min_val, max_val):\n", + " total_sum = 0\n", + " lcg_gen = lcg(initial_seed)\n", + " for _ in range(20):\n", + " seed = next(lcg_gen)\n", + " total_sum += max_subarray_sum(n, seed, min_val, max_val)\n", + " return total_sum\n", + "\"\"\"\n", + "\n", + "# Define a function to test the performance of the calculus function.\n", + "\n", + "def test_subarray_sum(calculus_function ,n=1000, initial_seed=42, min_val=-10, max_val=10):\n", + " \"\"\"Test the performance of the given calculus function.\"\"\"\n", + " start_time = perf_counter()\n", + " result = calculus_function(n, initial_seed, min_val, max_val)\n", + " end_time = perf_counter()\n", + " print(\"Total Maximum Subarray Sum (20 runs):\", result)\n", + " print(\"Execution Time: {:.6f} seconds\".format(end_time - start_time))\n", + "\n", + "\n", + "# Execute function declarations.\n", + "exec(python_hard_code)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "dab5e4bc-276c-4555-bd4c-12c699d5e899", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Total Maximum Subarray Sum (20 runs): 10980\n", + "Execution Time: 61.362418 seconds\n" + ] + } + ], + "source": [ + "# Run original python code and time it.\n", + "\n", + "test_subarray_sum(total_max_subarray_sum, 10000, 42, -10, 10)" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "id": "e8d24ed5-2c15-4f55-80e7-13a3952b3cb8", + "metadata": {}, + "outputs": [], + "source": [ + "# Request code optimization using GPT.\n", + "\n", + "optimization = optimize_gpt(python_hard_code, module_name)" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "id": "b888d5af", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "C CODE:\n", + "#include \n", + "#include \n", + "#include \n", + "#include \n", + "#include \n", + "\n", + "// LCG step with 32-bit wrap-around\n", + "static inline uint32_t lcg_next(uint32_t *state) {\n", + " *state = (uint32_t)(1664525u * (*state) + 1013904223u);\n", + " return *state;\n", + "}\n", + "\n", + "static inline int add_overflow_int64(int64_t a, int64_t b, int64_t *res) {\n", + " if ((b > 0 && a > INT64_MAX - b) || (b < 0 && a < INT64_MIN - b)) return 1;\n", + " *res = a + b;\n", + " return 0;\n", + "}\n", + "\n", + "// Kadane for int64 array with overflow detection; returns PyLong or NULL (on overflow -> signal via *overflowed)\n", + "static PyObject* kadane_int64(const int64_t *arr, Py_ssize_t n, int *overflowed) {\n", + " if (n <= 0) {\n", + " return PyFloat_FromDouble(-INFINITY);\n", + " }\n", + " int64_t meh = arr[0];\n", + " int64_t msf = arr[0];\n", + " for (Py_ssize_t i = 1; i < n; ++i) {\n", + " int64_t x = arr[i];\n", + " if (meh > 0) {\n", + " int64_t tmp;\n", + " if (add_overflow_int64(meh, x, &tmp)) { *overflowed = 1; return NULL; }\n", + " meh = tmp;\n", + " } else {\n", + " meh = x;\n", + " }\n", + " if (meh > msf) msf = meh;\n", + " }\n", + " return PyLong_FromLongLong(msf);\n", + "}\n", + "\n", + "// Kadane for PyObject* integer array\n", + "static PyObject* kadane_big(PyObject **arr, Py_ssize_t n) {\n", + " if (n <= 0) {\n", + " return PyFloat_FromDouble(-INFINITY);\n", + " }\n", + " PyObject *meh = arr[0]; Py_INCREF(meh);\n", + " PyObject *msf = arr[0]; Py_INCREF(msf);\n", + " PyObject *zero = PyLong_FromLong(0);\n", + " if (!zero) { Py_DECREF(meh); Py_DECREF(msf); return NULL; }\n", + "\n", + " for (Py_ssize_t i = 1; i < n; ++i) {\n", + " int cmp = PyObject_RichCompareBool(meh, zero, Py_GT);\n", + " if (cmp < 0) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; }\n", + " if (cmp == 1) {\n", + " PyObject *t = PyNumber_Add(meh, arr[i]);\n", + " if (!t) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; }\n", + " Py_DECREF(meh);\n", + " meh = t;\n", + " } else {\n", + " Py_DECREF(meh);\n", + " meh = arr[i]; Py_INCREF(meh);\n", + " }\n", + " int cmp2 = PyObject_RichCompareBool(meh, msf, Py_GT);\n", + " if (cmp2 < 0) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; }\n", + " if (cmp2 == 1) {\n", + " Py_DECREF(msf);\n", + " msf = meh; Py_INCREF(msf);\n", + " }\n", + " }\n", + " Py_DECREF(meh);\n", + " Py_DECREF(zero);\n", + " return msf; // new reference\n", + "}\n", + "\n", + "// Generate int64 array fast path; returns 0 on success\n", + "static int gen_array_int64(Py_ssize_t n, uint32_t seed, int64_t min_v, int64_t max_v, int64_t *out) {\n", + " uint32_t state = seed;\n", + " uint64_t umax = (uint64_t)max_v;\n", + " uint64_t umin = (uint64_t)min_v;\n", + " uint64_t range = (umax - umin) + 1ULL; // max>=min guaranteed by caller\n", + " for (Py_ssize_t i = 0; i < n; ++i) {\n", + " state = lcg_next(&state);\n", + " uint32_t r32 = state;\n", + " uint64_t r = (range > 0x100000000ULL) ? (uint64_t)r32 : ((uint64_t)r32 % range);\n", + " int64_t val = (int64_t)(min_v + (int64_t)r);\n", + " out[i] = val;\n", + " }\n", + " return 0;\n", + "}\n", + "\n", + "// Generate PyObject* int array general path using Python arithmetic\n", + "static PyObject** gen_array_big(Py_ssize_t n, uint32_t seed, PyObject *min_val, PyObject *max_val) {\n", + " PyObject **arr = (PyObject**)PyMem_Malloc((n > 0 ? n : 1) * sizeof(PyObject*));\n", + " if (!arr) {\n", + " PyErr_NoMemory();\n", + " return NULL;\n", + " }\n", + " PyObject *one = PyLong_FromLong(1);\n", + " if (!one) { PyMem_Free(arr); return NULL; }\n", + " PyObject *diff = PyNumber_Subtract(max_val, min_val);\n", + " if (!diff) { Py_DECREF(one); PyMem_Free(arr); return NULL; }\n", + " PyObject *range_obj = PyNumber_Add(diff, one);\n", + " Py_DECREF(diff);\n", + " Py_DECREF(one);\n", + " if (!range_obj) { PyMem_Free(arr); return NULL; }\n", + "\n", + " uint32_t state = seed;\n", + " for (Py_ssize_t i = 0; i < n; ++i) {\n", + " state = lcg_next(&state);\n", + " PyObject *v = PyLong_FromUnsignedLong((unsigned long)state);\n", + " if (!v) {\n", + " Py_DECREF(range_obj);\n", + " for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]);\n", + " PyMem_Free(arr);\n", + " return NULL;\n", + " }\n", + " PyObject *r = PyNumber_Remainder(v, range_obj);\n", + " Py_DECREF(v);\n", + " if (!r) {\n", + " Py_DECREF(range_obj);\n", + " for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]);\n", + " PyMem_Free(arr);\n", + " return NULL;\n", + " }\n", + " PyObject *val = PyNumber_Add(r, min_val);\n", + " Py_DECREF(r);\n", + " if (!val) {\n", + " Py_DECREF(range_obj);\n", + " for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]);\n", + " PyMem_Free(arr);\n", + " return NULL;\n", + " }\n", + " arr[i] = val;\n", + " }\n", + " Py_DECREF(range_obj);\n", + " return arr;\n", + "}\n", + "\n", + "static PyObject* max_subarray_sum_internal(Py_ssize_t n, uint32_t seed, PyObject *min_val, PyObject *max_val) {\n", + " if (n <= 0) {\n", + " return PyFloat_FromDouble(-INFINITY);\n", + " }\n", + "\n", + " if (PyLong_Check(min_val) && PyLong_Check(max_val)) {\n", + " int overflow1 = 0, overflow2 = 0;\n", + " long long min64 = PyLong_AsLongLongAndOverflow(min_val, &overflow1);\n", + " if (overflow1) goto BIGINT_PATH;\n", + " long long max64 = PyLong_AsLongLongAndOverflow(max_val, &overflow2);\n", + " if (overflow2) goto BIGINT_PATH;\n", + " if (max64 >= min64) {\n", + " int64_t *arr = (int64_t*)PyMem_Malloc((size_t)n * sizeof(int64_t));\n", + " if (!arr) { PyErr_NoMemory(); return NULL; }\n", + " if (gen_array_int64(n, seed, (int64_t)min64, (int64_t)max64, arr) != 0) {\n", + " PyMem_Free(arr);\n", + " return NULL;\n", + " }\n", + " int overflowed = 0;\n", + " PyObject *res = kadane_int64(arr, n, &overflowed);\n", + " if (!res && overflowed) {\n", + " // fallback to big-int Kadane\n", + " PyObject **arr_obj = (PyObject**)PyMem_Malloc((size_t)n * sizeof(PyObject*));\n", + " if (!arr_obj) { PyMem_Free(arr); PyErr_NoMemory(); return NULL; }\n", + " for (Py_ssize_t i = 0; i < n; ++i) {\n", + " arr_obj[i] = PyLong_FromLongLong(arr[i]);\n", + " if (!arr_obj[i]) {\n", + " for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr_obj[k]);\n", + " PyMem_Free(arr_obj);\n", + " PyMem_Free(arr);\n", + " return NULL;\n", + " }\n", + " }\n", + " PyObject *bires = kadane_big(arr_obj, n);\n", + " for (Py_ssize_t i = 0; i < n; ++i) Py_DECREF(arr_obj[i]);\n", + " PyMem_Free(arr_obj);\n", + " PyMem_Free(arr);\n", + " return bires;\n", + " }\n", + " PyMem_Free(arr);\n", + " return res;\n", + " }\n", + " }\n", + "BIGINT_PATH: ;\n", + " PyObject **arr_obj = gen_array_big(n, seed, min_val, max_val);\n", + " if (!arr_obj) return NULL;\n", + " PyObject *res = kadane_big(arr_obj, n);\n", + " for (Py_ssize_t i = 0; i < n; ++i) Py_DECREF(arr_obj[i]);\n", + " PyMem_Free(arr_obj);\n", + " return res;\n", + "}\n", + "\n", + "static PyObject* py_max_subarray_sum(PyObject *self, PyObject *args) {\n", + " Py_ssize_t n;\n", + " PyObject *seed_obj, *min_val, *max_val;\n", + " if (!PyArg_ParseTuple(args, \"nOOO\", &n, &seed_obj, &min_val, &max_val)) return NULL;\n", + " if (n < 0) n = 0;\n", + " uint32_t seed = (uint32_t)(PyLong_AsUnsignedLongLongMask(seed_obj) & 0xFFFFFFFFULL);\n", + " if (PyErr_Occurred()) return NULL;\n", + " return max_subarray_sum_internal(n, seed, min_val, max_val);\n", + "}\n", + "\n", + "static PyObject* py_total_max_subarray_sum(PyObject *self, PyObject *args) {\n", + " Py_ssize_t n;\n", + " PyObject *init_seed_obj, *min_val, *max_val;\n", + " if (!PyArg_ParseTuple(args, \"nOOO\", &n, &init_seed_obj, &min_val, &max_val)) return NULL;\n", + " if (n < 0) n = 0;\n", + " uint32_t state = (uint32_t)(PyLong_AsUnsignedLongLongMask(init_seed_obj) & 0xFFFFFFFFULL);\n", + " if (PyErr_Occurred()) return NULL;\n", + "\n", + " PyObject *total = PyLong_FromLong(0);\n", + " if (!total) return NULL;\n", + "\n", + " for (int i = 0; i < 20; ++i) {\n", + " uint32_t seed = lcg_next(&state);\n", + " PyObject *part = max_subarray_sum_internal(n, seed, min_val, max_val);\n", + " if (!part) { Py_DECREF(total); return NULL; }\n", + " PyObject *new_total = PyNumber_Add(total, part);\n", + " Py_DECREF(part);\n", + " if (!new_total) { Py_DECREF(total); return NULL; }\n", + " Py_DECREF(total);\n", + " total = new_total;\n", + " }\n", + " return total;\n", + "}\n", + "\n", + "static PyMethodDef module_methods[] = {\n", + " {\"max_subarray_sum\", (PyCFunction)py_max_subarray_sum, METH_VARARGS, \"Compute maximum subarray sum using LCG-generated array.\"},\n", + " {\"total_max_subarray_sum\", (PyCFunction)py_total_max_subarray_sum, METH_VARARGS, \"Compute total of maximum subarray sums over 20 LCG seeds.\"},\n", + " {NULL, NULL, 0, NULL}\n", + "};\n", + "\n", + "static struct PyModuleDef moduledef = {\n", + " PyModuleDef_HEAD_INIT,\n", + " \"python_hard\",\n", + " NULL,\n", + " -1,\n", + " module_methods,\n", + " NULL,\n", + " NULL,\n", + " NULL,\n", + " NULL\n", + "};\n", + "\n", + "PyMODINIT_FUNC PyInit_python_hard(void) {\n", + " return PyModule_Create(&moduledef);\n", + "}\n", + "\n", + "---------------------------\n", + "setup.py:\n", + "from setuptools import setup, Extension\n", + "import sys\n", + "\n", + "extra_compile_args = []\n", + "extra_link_args = []\n", + "if sys.platform == 'win32':\n", + " extra_compile_args = ['/O2', '/Ot', '/GL', '/fp:fast']\n", + " extra_link_args = ['/LTCG']\n", + "else:\n", + " extra_compile_args = ['-O3', '-march=native']\n", + "\n", + "module = Extension(\n", + " name='python_hard',\n", + " sources=['python_hard.c'],\n", + " extra_compile_args=extra_compile_args,\n", + " extra_link_args=extra_link_args,\n", + " language='c'\n", + ")\n", + "\n", + "setup(\n", + " name='python_hard',\n", + " version='1.0.0',\n", + " description='High-performance C extension reimplementation',\n", + " ext_modules=[module]\n", + ")\n", + "\n", + "---------------------------\n", + "USAGE:\n", + "import time\n", + "\n", + "# Original Python code\n", + "\n", + "def lcg(seed, a=1664525, c=1013904223, m=2**32):\n", + " value = seed\n", + " while True:\n", + " value = (a * value + c) % m\n", + " yield value\n", + "\n", + "def max_subarray_sum_py(n, seed, min_val, max_val):\n", + " lcg_gen = lcg(seed)\n", + " random_numbers = [next(lcg_gen) % (max_val - min_val + 1) + min_val for _ in range(n)]\n", + " max_sum = float('-inf')\n", + " for i in range(n):\n", + " current_sum = 0\n", + " for j in range(i, n):\n", + " current_sum += random_numbers[j]\n", + " if current_sum > max_sum:\n", + " max_sum = current_sum\n", + " return max_sum\n", + "\n", + "def total_max_subarray_sum_py(n, initial_seed, min_val, max_val):\n", + " total_sum = 0\n", + " lcg_gen = lcg(initial_seed)\n", + " for _ in range(20):\n", + " seed = next(lcg_gen)\n", + " total_sum += max_subarray_sum_py(n, seed, min_val, max_val)\n", + " return total_sum\n", + "\n", + "# Build and import extension (after running: python setup.py build && install or develop)\n", + "import python_hard as ext\n", + "\n", + "# Example parameters\n", + "n = 600\n", + "initial_seed = 12345678901234567890\n", + "min_val = -1000\n", + "max_val = 1000\n", + "\n", + "# Time Python\n", + "t0 = time.perf_counter()\n", + "py_res1 = max_subarray_sum_py(n, (initial_seed * 1664525 + 1013904223) % (2**32), min_val, max_val)\n", + "t1 = time.perf_counter()\n", + "py_time1 = t1 - t0\n", + "\n", + "# Time C extension\n", + "t0 = time.perf_counter()\n", + "ext_res1 = ext.max_subarray_sum(n, (initial_seed * 1664525 + 1013904223) % (2**32), min_val, max_val)\n", + "t1 = time.perf_counter()\n", + "ext_time1 = t1 - t0\n", + "\n", + "print('max_subarray_sum equality:', py_res1 == ext_res1)\n", + "print('Python time:', py_time1)\n", + "print('C ext time:', ext_time1)\n", + "\n", + "# Total over 20 seeds\n", + "t0 = time.perf_counter()\n", + "py_res2 = total_max_subarray_sum_py(n, initial_seed, min_val, max_val)\n", + "t1 = time.perf_counter()\n", + "py_time2 = t1 - t0\n", + "\n", + "t0 = time.perf_counter()\n", + "ext_res2 = ext.total_max_subarray_sum(n, initial_seed, min_val, max_val)\n", + "t1 = time.perf_counter()\n", + "ext_time2 = t1 - t0\n", + "\n", + "print('total_max_subarray_sum equality:', py_res2 == ext_res2)\n", + "print('Python total time:', py_time2)\n", + "print('C ext total time:', ext_time2)\n", + "\n" + ] + } + ], + "source": [ + "# Print generated extension code.\n", + "\n", + "print_optimization(optimization)" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "id": "356969b8", + "metadata": {}, + "outputs": [], + "source": [ + "# Write the generated extension code to files.\n", + "# (Will overwrite existing files)\n", + "\n", + "write_optimization(optimization, module_name)" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "id": "e0b3d073-88a2-40b2-831c-6f0c345c256f", + "metadata": {}, + "outputs": [], + "source": [ + "# Clean previous builds.\n", + "# (Make sure to run this cell before running the compile cell a second time only)\n", + "# (May cast errors if no previous build exists)\n", + "\n", + "!rm -r build/" + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "id": "29ae1993", + "metadata": {}, + "outputs": [], + "source": [ + "# [WINDOWS ONLY]\n", + "# Set COMSPEC to cmd.exe to avoid issues with some C compilers on Windows.\n", + "# (Remember to restore original COMSPEC after compilation and testing)\n", + "preset_comspec = os.environ.get(\"COMSPEC\")\n", + "os.environ[\"COMSPEC\"] = \"C:\\\\Windows\\\\System32\\\\cmd.exe\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "772328a7", + "metadata": {}, + "outputs": [], + "source": [ + "# Compile the C extension.\n", + "# (Will fail no C compiler is installed)\n", + "# (In case of errors, try directly on the command line)\n", + "\n", + "!python setup.py build_ext --inplace" + ] + }, + { + "cell_type": "code", + "execution_count": 29, + "id": "4186a19e", + "metadata": {}, + "outputs": [], + "source": [ + "# [WINDOWS ONLY]\n", + "# Restore original COMSPEC.\n", + "\n", + "os.environ[\"COMSPEC\"] = preset_comspec" + ] + }, + { + "cell_type": "code", + "execution_count": 30, + "id": "64899bb1", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "max_subarray_sum equality: True\n", + "Python time: 0.010335999992094003\n", + "C ext time: 1.4399993233382702e-05\n", + "total_max_subarray_sum equality: True\n", + "Python total time: 0.21065390000876505\n", + "C ext total time: 0.00012310000602155924\n" + ] + } + ], + "source": [ + "# Run the usage example to test the compiled C extension.\n", + "exec(optimization.usage)" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "id": "ee4f8f62", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Testing original Python code:\n", + "Total Maximum Subarray Sum (20 runs): 10980\n", + "Execution Time: 57.275276 seconds\n", + "Testing C extension code:\n", + "Total Maximum Subarray Sum (20 runs): 10980\n", + "Execution Time: 0.002317 seconds\n" + ] + } + ], + "source": [ + "# Import newly created C extension and compare performance with original Python code.\n", + "\n", + "from python_hard import total_max_subarray_sum as c_total_max_subarray_sum\n", + "\n", + "print(\"Testing original Python code:\")\n", + "test_subarray_sum(total_max_subarray_sum, 10000, 42, -10, 10)\n", + "print(\"Testing C extension code:\")\n", + "test_subarray_sum(c_total_max_subarray_sum, 10000, 42, -10, 10)" + ] + }, + { + "cell_type": "markdown", + "id": "85945330", + "metadata": {}, + "source": [ + "# Let's build a Gradio service" + ] + }, + { + "cell_type": "code", + "execution_count": 32, + "id": "2f1ae8f5-16c8-40a0-aa18-63b617df078d", + "metadata": {}, + "outputs": [], + "source": [ + "# Define a function to call the optimization process and return the generated codes.\n", + "\n", + "def optimize(python_code, module_name, model):\n", + " \"\"\"Call the optimization process and return the generated codes.\"\"\"\n", + " optimization = optimize_gpt(python_code, module_name, model)\n", + " return optimization.c_code, optimization.setup, optimization.usage" + ] + }, + { + "cell_type": "code", + "execution_count": 33, + "id": "19bf2bff-a822-4009-a539-f003b1651383", + "metadata": {}, + "outputs": [], + "source": [ + "# Define a function to execute Python code and capture its output.\n", + "\n", + "def execute_python(code):\n", + " \"\"\"Execute the given Python code and capture its output.\"\"\"\n", + " try:\n", + " output = io.StringIO()\n", + " sys.stdout = output\n", + " exec(code)\n", + " finally:\n", + " sys.stdout = sys.__stdout__\n", + " return output.getvalue()" + ] + }, + { + "cell_type": "code", + "execution_count": 34, + "id": "5fd83307", + "metadata": {}, + "outputs": [], + "source": [ + "# Extension compilation function.\n", + "\n", + "def build_extension():\n", + " \"\"\"Compile the C extension using 'setup.py' and return the compilation output.\"\"\"\n", + " # Set default COMSPEC to cmd.exe on Windows to avoid issues with some C compilers.\n", + " preset_comspec = os.environ.get(\"COMSPEC\")\n", + " os.environ[\"COMSPEC\"] = \"C:\\\\Windows\\\\System32\\\\cmd.exe\"\n", + " try:\n", + " compile_cmd = [\"python\", \"setup.py\", \"build_ext\", \"--inplace\"]\n", + " compile_result = subprocess.run(compile_cmd, env=os.environ,\n", + " check=True, text=True, capture_output=True)\n", + " except subprocess.CalledProcessError as ex:\n", + " raise Exception(f\"An error occurred while building:\\n{ex.stdout}\\n{ex.stderr}\")\n", + " finally: # The 'finally' clauses executes always whether there was an exception or not.\n", + " # Restore original COMSPEC.\n", + " os.environ[\"COMSPEC\"] = preset_comspec\n", + " return compile_result.stdout" + ] + }, + { + "cell_type": "code", + "execution_count": 35, + "id": "77f3ab5d-fcfb-4d3f-8728-9cacbf833ea6", + "metadata": {}, + "outputs": [], + "source": [ + "# Extension compilation function.\n", + "\n", + "def generate_extension(c_code, setup_code, usage_code, module_name):\n", + " \"\"\"Build and install the C extension from the provided codes.\"\"\"\n", + " try: # Write the provided codes to their respective files.\n", + " write_file(c_code, f\"{module_name}.c\")\n", + " write_file(setup_code, \"setup.py\")\n", + " except Exception as ex:\n", + " return f\"An error occurred while writing files:\\n{ex}\"\n", + " # Build the extension and capture the output.\n", + " try:\n", + " build_output = build_extension()\n", + " except Exception as ex: # If build fails, return the error message.\n", + " return str(ex)\n", + " # Return the combined output of build and install processes.\n", + " return build_output" + ] + }, + { + "cell_type": "code", + "execution_count": 36, + "id": "51f78871", + "metadata": {}, + "outputs": [], + "source": [ + "# Extension testing function.\n", + "\n", + "def test_extension(usage_code):\n", + " \"\"\"Test the installed C extension by executing the provided usage code and capturing its output.\"\"\"\n", + " try: # Write the provided codes to their respective files.\n", + " write_file(usage_code, \"usage_example.py\")\n", + " except Exception as ex:\n", + " return f\"An error occurred while writing test file:\\n{ex}\"\n", + " try:\n", + " output = execute_python(usage_code)\n", + " except Exception as ex:\n", + " return f\"An error occurred while testing the extension:\\n{ex}\"\n", + " return output" + ] + }, + { + "cell_type": "code", + "execution_count": 37, + "id": "9a2274f1-d03b-42c0-8dcc-4ce159b18442", + "metadata": {}, + "outputs": [], + "source": [ + "# Define custom CSS for Gradio interface.\n", + "\n", + "css = \"\"\"\n", + ".python {background-color: #306998;}\n", + ".c_ext {background-color: #050;}\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": 38, + "id": "c7c8f5fc", + "metadata": {}, + "outputs": [], + "source": [ + "# Define default codes for the interface.\n", + "\n", + "default_p_code = \"\"\"\n", + "def hello_world():\n", + " print(\"Hello, World!\")\n", + "\"\"\"\n", + "# default_p_code = python_hard_code # Run the declaration cell before use.\n", + "# default_p_code = calculate_pi_code # Run the declaration cell before use.\n", + "\n", + "default_c_code = r\"\"\"\n", + "#include \n", + "\n", + "// Function to be called from Python\n", + "static PyObject* zz_hello_world(PyObject* self, PyObject* args) {\n", + " printf(\"Hello, World!\\n\");\n", + " Py_RETURN_NONE;\n", + "}\n", + "\n", + "// Method definition structure\n", + "static PyMethodDef zz_my_methods[] = {\n", + " {\"hello_world\", zz_hello_world, METH_VARARGS, \"Print 'Hello, World!'\"},\n", + " {NULL, NULL, 0, NULL} // Sentinel\n", + "};\n", + "\n", + "// Module definition\n", + "static struct PyModuleDef zz_my_module = {\n", + " PyModuleDef_HEAD_INIT,\n", + " \"zz_my_module\",\n", + " \"Extension module that prints Hello, World!\",\n", + " -1,\n", + " zz_my_methods\n", + "};\n", + "\n", + "// Module initialization function\n", + "PyMODINIT_FUNC PyInit_zz_my_module(void) {\n", + " return PyModule_Create(&zz_my_module);\n", + "}\n", + "\"\"\"\n", + "\n", + "default_setup = \"\"\"\n", + "from setuptools import setup, Extension\n", + "\n", + "module = Extension(\n", + " 'zz_my_module',\n", + " sources=['zz_my_module.c'],\n", + ")\n", + "\n", + "setup(\n", + " name='zz_my_module',\n", + " version='1.0',\n", + " description='This is a custom C extension module.',\n", + " ext_modules=[module]\n", + ")\n", + "\"\"\"\n", + "\n", + "default_test = \"\"\"\n", + "import time\n", + "import zz_my_module\n", + "\n", + "def python_hello_world():\n", + " print(\"Hello, World!\")\n", + "\n", + "start = time.time()\n", + "python_hello_world()\n", + "end = time.time()\n", + "print(f\"Python function execution time: {end - start:.6f} seconds\")\n", + "\n", + "start = time.time()\n", + "zz_my_module.hello_world()\n", + "end = time.time()\n", + "print(f\"C extension execution time: {end - start:.6f} seconds\")\n", + "\"\"\"" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "id": "fa64577a", + "metadata": {}, + "outputs": [], + "source": [ + "# We will use gradio auto reload feature, this way we do not need to restart the app to see changes in the code.\n", + "# * https://www.gradio.app/guides/developing-faster-with-reload-mode\n", + "\n", + "%load_ext gradio\n", + "\n", + "# This mandatory requires naming the 'gr.Blocks' interface as 'demo'.\n", + "# Now, each time we edit the code, we just need to re-run Gradio interface cell to see the changes in the app.\n", + "# The '.launch()' method is not needed anymore." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f1303932-160c-424b-97a8-d28c816721b2", + "metadata": {}, + "outputs": [], + "source": [ + "%%blocks\n", + "\n", + "with gr.Blocks(css=css) as demo:\n", + " gr.Markdown(\"## Convert code from Python to C++\")\n", + " with gr.Row():\n", + " module_name = gr.Textbox(label=\"Module name:\", lines=1, value=\"zz_my_module\")\n", + " model = gr.Dropdown([\"gpt-4o\", \"gpt-5\"], label=\"Select model\", value=\"gpt-4o\")\n", + " with gr.Row():\n", + " python = gr.Textbox(label=\"Python code:\", lines=30, value=default_p_code, elem_classes=[\"python\"])\n", + " c_code = gr.Textbox(label=\"C Extension code:\", lines=30, value=default_c_code, elem_classes=[\"c_ext\"])\n", + " with gr.Row():\n", + " get_extension = gr.Button(\"Generate extension code\")\n", + " with gr.Row():\n", + " setup_code = gr.Textbox(label=\"Compilation code:\", lines=10, value=default_setup, elem_classes=[\"python\"])\n", + " usage_code = gr.Textbox(label=\"Test compare code:\", lines=10, value=default_test, elem_classes=[\"python\"])\n", + " with gr.Row():\n", + " compile_ext = gr.Button(\"Compile extension\")\n", + " with gr.Row():\n", + " c_ext_out = gr.TextArea(label=\"C Extension result:\", elem_classes=[\"c_ext\"])\n", + " with gr.Row():\n", + " test_run = gr.Button(\"Test code\")\n", + " with gr.Row():\n", + " test_out = gr.TextArea(label=\"Test result:\", elem_classes=[\"python\"])\n", + "\n", + " get_extension.click(optimize, inputs=[python, module_name, model], outputs=[c_code, setup_code, usage_code])\n", + " compile_ext.click(generate_extension, inputs=[c_code, setup_code, usage_code, module_name ], outputs=[c_ext_out])\n", + " test_run.click(test_extension, inputs=[usage_code], outputs=[test_out])\n" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llm-engineering-pYAzjokc", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.9" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week4/community-contributions/c_extension_generator/python_hard.c b/week4/community-contributions/c_extension_generator/python_hard.c new file mode 100644 index 0000000..67968ea --- /dev/null +++ b/week4/community-contributions/c_extension_generator/python_hard.c @@ -0,0 +1,244 @@ +#include +#include +#include +#include +#include + +// LCG step with 32-bit wrap-around +static inline uint32_t lcg_next(uint32_t *state) { + *state = (uint32_t)(1664525u * (*state) + 1013904223u); + return *state; +} + +static inline int add_overflow_int64(int64_t a, int64_t b, int64_t *res) { + if ((b > 0 && a > INT64_MAX - b) || (b < 0 && a < INT64_MIN - b)) return 1; + *res = a + b; + return 0; +} + +// Kadane for int64 array with overflow detection; returns PyLong or NULL (on overflow -> signal via *overflowed) +static PyObject* kadane_int64(const int64_t *arr, Py_ssize_t n, int *overflowed) { + if (n <= 0) { + return PyFloat_FromDouble(-INFINITY); + } + int64_t meh = arr[0]; + int64_t msf = arr[0]; + for (Py_ssize_t i = 1; i < n; ++i) { + int64_t x = arr[i]; + if (meh > 0) { + int64_t tmp; + if (add_overflow_int64(meh, x, &tmp)) { *overflowed = 1; return NULL; } + meh = tmp; + } else { + meh = x; + } + if (meh > msf) msf = meh; + } + return PyLong_FromLongLong(msf); +} + +// Kadane for PyObject* integer array +static PyObject* kadane_big(PyObject **arr, Py_ssize_t n) { + if (n <= 0) { + return PyFloat_FromDouble(-INFINITY); + } + PyObject *meh = arr[0]; Py_INCREF(meh); + PyObject *msf = arr[0]; Py_INCREF(msf); + PyObject *zero = PyLong_FromLong(0); + if (!zero) { Py_DECREF(meh); Py_DECREF(msf); return NULL; } + + for (Py_ssize_t i = 1; i < n; ++i) { + int cmp = PyObject_RichCompareBool(meh, zero, Py_GT); + if (cmp < 0) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; } + if (cmp == 1) { + PyObject *t = PyNumber_Add(meh, arr[i]); + if (!t) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; } + Py_DECREF(meh); + meh = t; + } else { + Py_DECREF(meh); + meh = arr[i]; Py_INCREF(meh); + } + int cmp2 = PyObject_RichCompareBool(meh, msf, Py_GT); + if (cmp2 < 0) { Py_DECREF(meh); Py_DECREF(msf); Py_DECREF(zero); return NULL; } + if (cmp2 == 1) { + Py_DECREF(msf); + msf = meh; Py_INCREF(msf); + } + } + Py_DECREF(meh); + Py_DECREF(zero); + return msf; // new reference +} + +// Generate int64 array fast path; returns 0 on success +static int gen_array_int64(Py_ssize_t n, uint32_t seed, int64_t min_v, int64_t max_v, int64_t *out) { + uint32_t state = seed; + uint64_t umax = (uint64_t)max_v; + uint64_t umin = (uint64_t)min_v; + uint64_t range = (umax - umin) + 1ULL; // max>=min guaranteed by caller + for (Py_ssize_t i = 0; i < n; ++i) { + state = lcg_next(&state); + uint32_t r32 = state; + uint64_t r = (range > 0x100000000ULL) ? (uint64_t)r32 : ((uint64_t)r32 % range); + int64_t val = (int64_t)(min_v + (int64_t)r); + out[i] = val; + } + return 0; +} + +// Generate PyObject* int array general path using Python arithmetic +static PyObject** gen_array_big(Py_ssize_t n, uint32_t seed, PyObject *min_val, PyObject *max_val) { + PyObject **arr = (PyObject**)PyMem_Malloc((n > 0 ? n : 1) * sizeof(PyObject*)); + if (!arr) { + PyErr_NoMemory(); + return NULL; + } + PyObject *one = PyLong_FromLong(1); + if (!one) { PyMem_Free(arr); return NULL; } + PyObject *diff = PyNumber_Subtract(max_val, min_val); + if (!diff) { Py_DECREF(one); PyMem_Free(arr); return NULL; } + PyObject *range_obj = PyNumber_Add(diff, one); + Py_DECREF(diff); + Py_DECREF(one); + if (!range_obj) { PyMem_Free(arr); return NULL; } + + uint32_t state = seed; + for (Py_ssize_t i = 0; i < n; ++i) { + state = lcg_next(&state); + PyObject *v = PyLong_FromUnsignedLong((unsigned long)state); + if (!v) { + Py_DECREF(range_obj); + for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]); + PyMem_Free(arr); + return NULL; + } + PyObject *r = PyNumber_Remainder(v, range_obj); + Py_DECREF(v); + if (!r) { + Py_DECREF(range_obj); + for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]); + PyMem_Free(arr); + return NULL; + } + PyObject *val = PyNumber_Add(r, min_val); + Py_DECREF(r); + if (!val) { + Py_DECREF(range_obj); + for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr[k]); + PyMem_Free(arr); + return NULL; + } + arr[i] = val; + } + Py_DECREF(range_obj); + return arr; +} + +static PyObject* max_subarray_sum_internal(Py_ssize_t n, uint32_t seed, PyObject *min_val, PyObject *max_val) { + if (n <= 0) { + return PyFloat_FromDouble(-INFINITY); + } + + if (PyLong_Check(min_val) && PyLong_Check(max_val)) { + int overflow1 = 0, overflow2 = 0; + long long min64 = PyLong_AsLongLongAndOverflow(min_val, &overflow1); + if (overflow1) goto BIGINT_PATH; + long long max64 = PyLong_AsLongLongAndOverflow(max_val, &overflow2); + if (overflow2) goto BIGINT_PATH; + if (max64 >= min64) { + int64_t *arr = (int64_t*)PyMem_Malloc((size_t)n * sizeof(int64_t)); + if (!arr) { PyErr_NoMemory(); return NULL; } + if (gen_array_int64(n, seed, (int64_t)min64, (int64_t)max64, arr) != 0) { + PyMem_Free(arr); + return NULL; + } + int overflowed = 0; + PyObject *res = kadane_int64(arr, n, &overflowed); + if (!res && overflowed) { + // fallback to big-int Kadane + PyObject **arr_obj = (PyObject**)PyMem_Malloc((size_t)n * sizeof(PyObject*)); + if (!arr_obj) { PyMem_Free(arr); PyErr_NoMemory(); return NULL; } + for (Py_ssize_t i = 0; i < n; ++i) { + arr_obj[i] = PyLong_FromLongLong(arr[i]); + if (!arr_obj[i]) { + for (Py_ssize_t k = 0; k < i; ++k) Py_DECREF(arr_obj[k]); + PyMem_Free(arr_obj); + PyMem_Free(arr); + return NULL; + } + } + PyObject *bires = kadane_big(arr_obj, n); + for (Py_ssize_t i = 0; i < n; ++i) Py_DECREF(arr_obj[i]); + PyMem_Free(arr_obj); + PyMem_Free(arr); + return bires; + } + PyMem_Free(arr); + return res; + } + } +BIGINT_PATH: ; + PyObject **arr_obj = gen_array_big(n, seed, min_val, max_val); + if (!arr_obj) return NULL; + PyObject *res = kadane_big(arr_obj, n); + for (Py_ssize_t i = 0; i < n; ++i) Py_DECREF(arr_obj[i]); + PyMem_Free(arr_obj); + return res; +} + +static PyObject* py_max_subarray_sum(PyObject *self, PyObject *args) { + Py_ssize_t n; + PyObject *seed_obj, *min_val, *max_val; + if (!PyArg_ParseTuple(args, "nOOO", &n, &seed_obj, &min_val, &max_val)) return NULL; + if (n < 0) n = 0; + uint32_t seed = (uint32_t)(PyLong_AsUnsignedLongLongMask(seed_obj) & 0xFFFFFFFFULL); + if (PyErr_Occurred()) return NULL; + return max_subarray_sum_internal(n, seed, min_val, max_val); +} + +static PyObject* py_total_max_subarray_sum(PyObject *self, PyObject *args) { + Py_ssize_t n; + PyObject *init_seed_obj, *min_val, *max_val; + if (!PyArg_ParseTuple(args, "nOOO", &n, &init_seed_obj, &min_val, &max_val)) return NULL; + if (n < 0) n = 0; + uint32_t state = (uint32_t)(PyLong_AsUnsignedLongLongMask(init_seed_obj) & 0xFFFFFFFFULL); + if (PyErr_Occurred()) return NULL; + + PyObject *total = PyLong_FromLong(0); + if (!total) return NULL; + + for (int i = 0; i < 20; ++i) { + uint32_t seed = lcg_next(&state); + PyObject *part = max_subarray_sum_internal(n, seed, min_val, max_val); + if (!part) { Py_DECREF(total); return NULL; } + PyObject *new_total = PyNumber_Add(total, part); + Py_DECREF(part); + if (!new_total) { Py_DECREF(total); return NULL; } + Py_DECREF(total); + total = new_total; + } + return total; +} + +static PyMethodDef module_methods[] = { + {"max_subarray_sum", (PyCFunction)py_max_subarray_sum, METH_VARARGS, "Compute maximum subarray sum using LCG-generated array."}, + {"total_max_subarray_sum", (PyCFunction)py_total_max_subarray_sum, METH_VARARGS, "Compute total of maximum subarray sums over 20 LCG seeds."}, + {NULL, NULL, 0, NULL} +}; + +static struct PyModuleDef moduledef = { + PyModuleDef_HEAD_INIT, + "python_hard", + NULL, + -1, + module_methods, + NULL, + NULL, + NULL, + NULL +}; + +PyMODINIT_FUNC PyInit_python_hard(void) { + return PyModule_Create(&moduledef); +} diff --git a/week4/community-contributions/c_extension_generator/setup_calculate_pi.py b/week4/community-contributions/c_extension_generator/setup_calculate_pi.py new file mode 100644 index 0000000..ecb4067 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/setup_calculate_pi.py @@ -0,0 +1,25 @@ +from setuptools import setup, Extension +import sys +import os + +extra_compile_args = [] +extra_link_args = [] + +if os.name == 'nt': + extra_compile_args.extend(['/O2', '/fp:precise']) +else: + extra_compile_args.extend(['-O3', '-fno-strict-aliasing']) + +module = Extension( + 'calculate_pi', + sources=['calculate_pi.c'], + extra_compile_args=extra_compile_args, + extra_link_args=extra_link_args, +) + +setup( + name='calculate_pi', + version='1.0.0', + description='High-performance C extension for computing pi via the Leibniz series', + ext_modules=[module], +) diff --git a/week4/community-contributions/c_extension_generator/setup_python_hard.py b/week4/community-contributions/c_extension_generator/setup_python_hard.py new file mode 100644 index 0000000..20d3d3f --- /dev/null +++ b/week4/community-contributions/c_extension_generator/setup_python_hard.py @@ -0,0 +1,25 @@ +from setuptools import setup, Extension +import sys + +extra_compile_args = [] +extra_link_args = [] +if sys.platform == 'win32': + extra_compile_args = ['/O2', '/Ot', '/GL', '/fp:fast'] + extra_link_args = ['/LTCG'] +else: + extra_compile_args = ['-O3', '-march=native'] + +module = Extension( + name='python_hard', + sources=['python_hard.c'], + extra_compile_args=extra_compile_args, + extra_link_args=extra_link_args, + language='c' +) + +setup( + name='python_hard', + version='1.0.0', + description='High-performance C extension reimplementation', + ext_modules=[module] +) diff --git a/week4/community-contributions/c_extension_generator/setup_zz_my_module.py b/week4/community-contributions/c_extension_generator/setup_zz_my_module.py new file mode 100644 index 0000000..bb0d27b --- /dev/null +++ b/week4/community-contributions/c_extension_generator/setup_zz_my_module.py @@ -0,0 +1,14 @@ + +from setuptools import setup, Extension + +module = Extension( + 'zz_my_module', + sources=['zz_my_module.c'], +) + +setup( + name='zz_my_module', + version='1.0', + description='This is a custom C extension module.', + ext_modules=[module] +) diff --git a/week4/community-contributions/c_extension_generator/usage_example_calculate_pi.py b/week4/community-contributions/c_extension_generator/usage_example_calculate_pi.py new file mode 100644 index 0000000..c3cc418 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/usage_example_calculate_pi.py @@ -0,0 +1,38 @@ +# Build first: python setup.py build_ext --inplace +import time +import math +import calculate_pi + +# Original Python implementation +def py_leibniz_pi(iterations): + result = 1.0 + for i in range(1, iterations + 1): + j = i * 4 - 1 + result -= (1 / j) + j = i * 4 + 1 + result += (1 / j) + return result * 4 + +iters = 5_000_000 + +# Warm-up +calculate_pi.leibniz_pi(10) +py_leibniz_pi(10) + +start = time.perf_counter() +res_c = calculate_pi.leibniz_pi(iters) +end = time.perf_counter() +ctime = end - start + +start = time.perf_counter() +res_py = py_leibniz_pi(iters) +end = time.perf_counter() +pytime = end - start + +print(f"Iterations: {iters}") +print(f"C extension result: {res_c}") +print(f"Python result: {res_py}") +print(f"Absolute difference: {abs(res_c - res_py)}") +print(f"C extension time: {ctime:.6f} s") +print(f"Python time: {pytime:.6f} s") +print(f"Speedup: {pytime/ctime if ctime > 0 else float('inf'):.2f}x") diff --git a/week4/community-contributions/c_extension_generator/usage_example_python_hard.py b/week4/community-contributions/c_extension_generator/usage_example_python_hard.py new file mode 100644 index 0000000..552b0b5 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/usage_example_python_hard.py @@ -0,0 +1,69 @@ +import time + +# Original Python code + +def lcg(seed, a=1664525, c=1013904223, m=2**32): + value = seed + while True: + value = (a * value + c) % m + yield value + +def max_subarray_sum_py(n, seed, min_val, max_val): + lcg_gen = lcg(seed) + random_numbers = [next(lcg_gen) % (max_val - min_val + 1) + min_val for _ in range(n)] + max_sum = float('-inf') + for i in range(n): + current_sum = 0 + for j in range(i, n): + current_sum += random_numbers[j] + if current_sum > max_sum: + max_sum = current_sum + return max_sum + +def total_max_subarray_sum_py(n, initial_seed, min_val, max_val): + total_sum = 0 + lcg_gen = lcg(initial_seed) + for _ in range(20): + seed = next(lcg_gen) + total_sum += max_subarray_sum_py(n, seed, min_val, max_val) + return total_sum + +# Build and import extension (after running: python setup.py build && install or develop) +import python_hard as ext + +# Example parameters +n = 600 +initial_seed = 12345678901234567890 +min_val = -1000 +max_val = 1000 + +# Time Python +t0 = time.perf_counter() +py_res1 = max_subarray_sum_py(n, (initial_seed * 1664525 + 1013904223) % (2**32), min_val, max_val) +t1 = time.perf_counter() +py_time1 = t1 - t0 + +# Time C extension +t0 = time.perf_counter() +ext_res1 = ext.max_subarray_sum(n, (initial_seed * 1664525 + 1013904223) % (2**32), min_val, max_val) +t1 = time.perf_counter() +ext_time1 = t1 - t0 + +print('max_subarray_sum equality:', py_res1 == ext_res1) +print('Python time:', py_time1) +print('C ext time:', ext_time1) + +# Total over 20 seeds +t0 = time.perf_counter() +py_res2 = total_max_subarray_sum_py(n, initial_seed, min_val, max_val) +t1 = time.perf_counter() +py_time2 = t1 - t0 + +t0 = time.perf_counter() +ext_res2 = ext.total_max_subarray_sum(n, initial_seed, min_val, max_val) +t1 = time.perf_counter() +ext_time2 = t1 - t0 + +print('total_max_subarray_sum equality:', py_res2 == ext_res2) +print('Python total time:', py_time2) +print('C ext total time:', ext_time2) diff --git a/week4/community-contributions/c_extension_generator/usage_example_zz_my_module.py b/week4/community-contributions/c_extension_generator/usage_example_zz_my_module.py new file mode 100644 index 0000000..6404cd3 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/usage_example_zz_my_module.py @@ -0,0 +1,16 @@ + +import time +import zz_my_module + +def python_hello_world(): + print("Hello, World!") + +start = time.time() +python_hello_world() +end = time.time() +print(f"Python function execution time: {end - start:.6f} seconds") + +start = time.time() +zz_my_module.hello_world() +end = time.time() +print(f"C extension execution time: {end - start:.6f} seconds") diff --git a/week4/community-contributions/c_extension_generator/zz_my_module.c b/week4/community-contributions/c_extension_generator/zz_my_module.c new file mode 100644 index 0000000..2593233 --- /dev/null +++ b/week4/community-contributions/c_extension_generator/zz_my_module.c @@ -0,0 +1,28 @@ + +#include + +// Function to be called from Python +static PyObject* zz_hello_world(PyObject* self, PyObject* args) { + printf("Hello, World!\n"); + Py_RETURN_NONE; +} + +// Method definition structure +static PyMethodDef zz_my_methods[] = { + {"hello_world", zz_hello_world, METH_VARARGS, "Print 'Hello, World!'"}, + {NULL, NULL, 0, NULL} // Sentinel +}; + +// Module definition +static struct PyModuleDef zz_my_module = { + PyModuleDef_HEAD_INIT, + "zz_my_module", + "Extension module that prints Hello, World!", + -1, + zz_my_methods +}; + +// Module initialization function +PyMODINIT_FUNC PyInit_zz_my_module(void) { + return PyModule_Create(&zz_my_module); +} diff --git a/week4/community-contributions/unit-test-generator-v3.ipynb b/week4/community-contributions/unit-test-generator-v3.ipynb new file mode 100644 index 0000000..c47b6d0 --- /dev/null +++ b/week4/community-contributions/unit-test-generator-v3.ipynb @@ -0,0 +1,222 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "56957b7f-e289-4999-8a40-ce1a8378d8cd", + "metadata": {}, + "source": [ + "# Unit Test Generator\n", + "\n", + "The requirement: use a Frontier model to generate fast and repeatable unit tests for Python code.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3ef67ef0-1bda-45bb-abca-f003217602d4", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import io\n", + "import sys\n", + "import ast\n", + "import unittest, contextlib\n", + "from dotenv import load_dotenv\n", + "from openai import OpenAI\n", + "import google.generativeai\n", + "import anthropic\n", + "from IPython.display import Markdown, display, update_display\n", + "import gradio as gr\n", + "import subprocess\n", + "\n", + "# environment\n", + "\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n", + "os.environ['ANTHROPIC_API_KEY'] = os.getenv('ANTHROPIC_API_KEY', 'your-key-if-not-using-env')\n", + "\n", + "openai = OpenAI()\n", + "claude = anthropic.Anthropic()\n", + "OPENAI_MODEL = \"gpt-4o\"\n", + "CLAUDE_MODEL = \"claude-3-7-sonnet-20250219\"\n", + "\n", + "system_message = \"You are an assistant that implements unit testing for Python code. \"\n", + "system_message += \"Respond only with Python code; use comments sparingly and do not provide any explanation other than occasional comments. \"\n", + "\n", + "def remove_main_block_from_code(code):\n", + " \"\"\"\n", + " Remove top-level `if __name__ == \"__main__\":` blocks from code.\n", + " \"\"\"\n", + " try:\n", + " tree = ast.parse(code)\n", + "\n", + " class RemoveMain(ast.NodeTransformer):\n", + " def visit_If(self, node):\n", + " # check if this is: if __name__ == \"__main__\":\n", + " test = node.test\n", + " if (\n", + " isinstance(test, ast.Compare) and\n", + " isinstance(test.left, ast.Name) and\n", + " test.left.id == \"__name__\" and\n", + " len(test.ops) == 1 and isinstance(test.ops[0], ast.Eq) and\n", + " len(test.comparators) == 1 and\n", + " isinstance(test.comparators[0], ast.Constant) and\n", + " test.comparators[0].value == \"__main__\"\n", + " ):\n", + " return None # remove this node entirely\n", + " return node\n", + "\n", + " tree = RemoveMain().visit(tree)\n", + " ast.fix_missing_locations(tree)\n", + " return ast.unparse(tree) # get back code as string\n", + " except Exception as e:\n", + " print(\"Error removing __main__ block:\", e)\n", + " return code # fallback: return original code if AST fails\n", + "\n", + "def user_prompt_for(python_file):\n", + " if isinstance(python_file, dict): # from Gradio\n", + " file_path = python_file[\"name\"]\n", + " elif hasattr(python_file, \"name\"): # tempfile\n", + " file_path = python_file.name\n", + " else: # string path\n", + " file_path = python_file\n", + "\n", + " with open(file_path, \"r\", encoding=\"utf-8\") as f:\n", + " python_code = f.read()\n", + "\n", + " # strip __main__ blocks\n", + " python_code = remove_main_block_from_code(python_code)\n", + "\n", + " user_prompt = \"Write unit tests for this Python code. \"\n", + " user_prompt += \"Respond only with Python code; do not explain your work other than a few comments. \"\n", + " user_prompt += \"The unit testing is done in Jupyterlab, so you should use packages that play nicely with the Jupyter kernel. \\n\\n\"\n", + " user_prompt += \"Include the original Python code in your generated output so that I can run all in one fell swoop.\\n\\n\"\n", + " user_prompt += python_code\n", + "\n", + " return user_prompt\n", + "\n", + "def messages_for(python_file):\n", + " return [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt_for(python_file)}\n", + " ]\n", + "\t\n", + "def stream_gpt(python_file): \n", + " stream = openai.chat.completions.create(model=OPENAI_MODEL, messages=messages_for(python_file), stream=True)\n", + " reply = \"\"\n", + " for chunk in stream:\n", + " fragment = chunk.choices[0].delta.content or \"\"\n", + " reply += fragment\n", + " yield reply.replace('```python\\n','').replace('```','')\n", + "\t\t\n", + "def stream_claude(python_file):\n", + " result = claude.messages.stream(\n", + " model=CLAUDE_MODEL,\n", + " max_tokens=2000,\n", + " system=system_message,\n", + " messages=[{\"role\": \"user\", \"content\": user_prompt_for(python_file)}],\n", + " )\n", + " reply = \"\"\n", + " with result as stream:\n", + " for text in stream.text_stream:\n", + " reply += text\n", + " yield reply.replace('```python\\n','').replace('```','')\n", + "\t\t\t\n", + "def unit_test(python_file, model):\n", + " if model==\"GPT\":\n", + " result = stream_gpt(python_file)\n", + " elif model==\"Claude\":\n", + " result = stream_claude(python_file)\n", + " else:\n", + " raise ValueError(\"Unknown model\")\n", + " for stream_so_far in result:\n", + " yield stream_so_far\n", + "\n", + "def execute_python(code):\n", + " buffer = io.StringIO()\n", + " try:\n", + " with contextlib.redirect_stdout(buffer), contextlib.redirect_stderr(buffer):\n", + " # execute code in isolated namespace\n", + " ns = {}\n", + " exec(code, ns)\n", + "\n", + " # manually collect TestCase subclasses\n", + " test_cases = [\n", + " obj for obj in ns.values()\n", + " if isinstance(obj, type) and issubclass(obj, unittest.TestCase)\n", + " ]\n", + " if test_cases:\n", + " suite = unittest.TestSuite()\n", + " for case in test_cases:\n", + " suite.addTests(unittest.defaultTestLoader.loadTestsFromTestCase(case))\n", + " runner = unittest.TextTestRunner(stream=buffer, verbosity=2)\n", + " runner.run(suite)\n", + " except Exception as e:\n", + " print(f\"Error during execution: {e}\", file=buffer)\n", + "\n", + " return buffer.getvalue()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "670b8b78-0b13-488a-9533-59802b2fe101", + "metadata": {}, + "outputs": [], + "source": [ + "# --- Gradio UI ---\n", + "with gr.Blocks() as ui:\n", + " gr.Markdown(\"## Unit Test Generator\\nUpload a Python file and get structured unit testing.\")\n", + " with gr.Row(): # Row 1\n", + " orig_code = gr.File(label=\"Upload your Python file\", file_types=[\".py\"])\n", + " test_code = gr.Textbox(label=\"Unit test code:\", lines=10)\n", + " with gr.Row(): # Row 2\n", + " model = gr.Dropdown([\"GPT\", \"Claude\"], label=\"Select model\", value=\"GPT\")\n", + " with gr.Row(): # Row 3\n", + " generate = gr.Button(\"Generate unit test code\")\n", + " with gr.Row(): # Row 4\n", + " unit_run = gr.Button(\"Run Python unit test\")\n", + " with gr.Row(): # Row 5\n", + " test_out = gr.Textbox(label=\"Unit test result:\", lines=10)\n", + "\n", + " generate.click(unit_test, inputs=[orig_code, model], outputs=[test_code])\n", + "\n", + " unit_run.click(execute_python, inputs=[test_code], outputs=[test_out])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "609bbdc3-1e1c-4538-91dd-7d13134ab381", + "metadata": {}, + "outputs": [], + "source": [ + "ui.launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week4/community-contributions/week4-lchanio-code-documenter.ipynb b/week4/community-contributions/week4-lchanio-code-documenter.ipynb new file mode 100644 index 0000000..aee045d --- /dev/null +++ b/week4/community-contributions/week4-lchanio-code-documenter.ipynb @@ -0,0 +1,300 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "44d517c6", + "metadata": {}, + "source": [ + "# Code Documenter\n", + "This application documents a code module:\n", + "Generates a DocString\n", + "Inserts comments where applicable\n", + "\n", + "Output is in TextBox and can be written to a file.\n", + "\n", + "Offeres model selection" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c349e47b", + "metadata": {}, + "outputs": [], + "source": [ + "# Import Libraries\n", + "\n", + "from dotenv import load_dotenv\n", + "import os\n", + "#import requests\n", + "from IPython.display import Markdown, display, update_display\n", + "from openai import OpenAI\n", + "import anthropic\n", + "from google import genai\n", + "from google.genai import types\n", + "# from google.colab import drive\n", + "from huggingface_hub import login\n", + "#from google.colab import userdata\n", + "from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer, BitsAndBytesConfig, TextIteratorStreamer\n", + "import torch\n", + "import gradio as gr\n", + "#import threading " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "46caf8e5", + "metadata": {}, + "outputs": [], + "source": [ + "# Constants - model names\n", + "LLAMA_MODEL = \"codellama-7b-kstack\" # we will be using LM_Studio for this model\n", + "QWEN_MODEL = \"qwen2.5-coder-14b-instruct\" # we will be using LM_Studio for this model, might be too large for some systems\n", + "OPENAI_MODEL = \"gpt-4o\"\n", + "ANTHROPIC_MODEL = \"claude-3-5-haiku-latest\"\n", + "GOOGLE_MODEL = \"gemini-2.5-pro\"\n", + "model_choices = [LLAMA_MODEL, QWEN_MODEL, OPENAI_MODEL, ANTHROPIC_MODEL, GOOGLE_MODEL]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6df20bf1", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables and set up API connections\n", + "load_dotenv(override=True)\n", + "openai_api_key = os.getenv('OPENAI_API_KEY')\n", + "anthropic_api_key = os.getenv('ANTHROPIC_API_KEY')\n", + "google_api_key = os.getenv('GOOGLE_API_KEY')\n", + "hf_api_key = os.getenv('HF_API_KEY')\n", + "\n", + "\n", + "if openai_api_key:\n", + " print(f\"OpenAI API Key exists and begins {openai_api_key[:8]}\")\n", + " openai = OpenAI(api_key=openai_api_key)\n", + "else:\n", + " print(\"OpenAI API Key not set\")\n", + " \n", + "if anthropic_api_key:\n", + " print(f\"Anthropic API Key exists and begins {anthropic_api_key[:7]}\")\n", + " claude = anthropic.Anthropic(api_key=anthropic_api_key)\n", + "else:\n", + " print(\"Anthropic API Key not set\")\n", + "\n", + "if google_api_key:\n", + " print(f\"Google API Key exists and begins {google_api_key[:8]}\")\n", + " gemini = genai.Client(api_key=google_api_key) \n", + "else:\n", + " print(\"Google API Key not set\")\n", + "\n", + "if hf_api_key:\n", + " print(f\"HuggingFace API Key exists and begins {hf_api_key[:7]}\")\n", + " login(hf_api_key, add_to_git_credential=True)\n", + "else:\n", + " print(\"HuggingFace API Key not set\")\n", + "\n", + "# Set up LM Studio connection\n", + "lm_studio_via_openai = OpenAI(base_url=\"http://127.0.0.1:1234/v1\", api_key=\"lmstudio\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4c9178a6", + "metadata": {}, + "outputs": [], + "source": [ + "# Model invocation function\n", + "# This function will handle the invocation of different models based on the selected model name.\n", + "\n", + "def invoke_model(model_name, prompt, max_tokens=1000, temperature=0.4):\n", + " if model_name == OPENAI_MODEL:\n", + " stream = openai.chat.completions.create(\n", + " model=OPENAI_MODEL,\n", + " messages=prompt,\n", + " max_tokens=max_tokens,\n", + " temperature=temperature,\n", + " stream=True\n", + " )\n", + " #return response.choices[0].message.content\n", + " for chunk in stream:\n", + " yield chunk.choices[0].delta.content or ''\n", + " \n", + " elif model_name == ANTHROPIC_MODEL:\n", + " #Invoke Claude model\n", + " # Handle the prompt structure for Claude\n", + " #print(f\"Invoking model: {model_name}\")\n", + " #print(f\"System prompt: {prompt[0]['content']}\")\n", + " #print(f\"User prompt: %s\", prompt[1]['content'][:100])\n", + " try:\n", + " # Use context manager for proper stream handling\n", + " with claude.messages.stream(\n", + " model=ANTHROPIC_MODEL,\n", + " system=prompt[0]['content'],\n", + " messages=[prompt[1]],\n", + " max_tokens=max_tokens,\n", + " temperature=temperature\n", + " ) as stream:\n", + " #print(\"Stream created successfully\")\n", + " chunk_count = 0\n", + " \n", + " for chunk in stream.text_stream:\n", + " chunk_count += 1\n", + " #print(f\"Chunk {chunk_count}: {repr(chunk)}\") # Use repr to see None/empty values\n", + " \n", + " if chunk: # Only yield non-empty chunks\n", + " yield chunk\n", + " \n", + " #print(f\"Stream completed. Total chunks: {chunk_count}\")\n", + " \n", + " except Exception as e:\n", + " print(f\"Error invoking Claude model: {e}\")\n", + " yield f\"Error invoking Claude model: {e}\"\n", + " return\n", + " \n", + " elif model_name == GOOGLE_MODEL:\n", + " #Invoke Gemini model\n", + " # Handle the prompt structure for Gemini\n", + " #print(f\"Invoking model: {model_name}\")\n", + " stream = gemini.models.generate_content_stream(\n", + " model=GOOGLE_MODEL,\n", + " contents=prompt[1]['content'],\n", + " config=types.GenerateContentConfig(\n", + " temperature=temperature,\n", + " maxOutputTokens=max_tokens,\n", + " system_instruction=prompt[0]['content'],)\n", + " )\n", + " #print(\"Streaming response from Gemini...\")\n", + " for chunk in stream:\n", + " yield chunk.text or ''\n", + " \n", + " elif model_name == LLAMA_MODEL or model_name == QWEN_MODEL:\n", + " # invoke LM Studio model\n", + " #print(f\"Invoking model: {model_name}\")\n", + " stream = lm_studio_via_openai.chat.completions.create(\n", + " model=model_name,\n", + " messages=prompt,\n", + " max_tokens=max_tokens,\n", + " temperature=temperature,\n", + " stream=True\n", + " )\n", + " #print(\"Streaming response from LM Studio...\")\n", + " #return response.choices[0].message.content\n", + " #response=\"\"\n", + " for chunk in stream:\n", + " #response += chunk.choices[0].delta.content or ''\n", + " yield chunk.choices[0].delta.content or ''\n", + " else:\n", + " raise ValueError(\"Unsupported model name\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d781b697", + "metadata": {}, + "outputs": [], + "source": [ + "# Save text to a selected location\n", + "# This function will save the generated text to a specified file or the current directory if no file\n", + "def save_text_to_selected_location(text_content):\n", + " if not text_content.strip():\n", + " return \"No content to save\"\n", + "\n", + " save_path = \"output.txt\"\n", + "\n", + " # If no file is selected, save to current directory\n", + " try:\n", + " with open(save_path, 'w', encoding='utf-8') as f:\n", + " f.write(text_content)\n", + " return f\"Successfully saved to: {save_path}\"\n", + " except Exception as e:\n", + " return f\"Error saving file: {str(e)}\"\n", + " \n", + "# Set up event handlers\n", + "def generate_response(system_input, prompt, max_tokens, temperature, model_name):\n", + " if system_input == \"Documentation\":\n", + " system_prompt = \"\"\"You are an experienced coding assistant. You will identify the programming language used in a provided code snippet and generate documentation for the code. \n", + " Ensure generally acceptable documentation standards are followed. Also generate short inline comments where applicable to explain complicated code. Respond ONLY with the updated code \n", + " with documentation and comments, do not include any other preamble or explanation.\"\"\"\n", + " elif system_input == \"Test Code Generation\":\n", + " system_prompt = \"\"\"You are an experienced coding assistant. You will identify the programming language used in a provided code function and generate test code for it. \" \\\n", + " \"The code should test against normal and edge cases, and ensure proper error handling:\"\"\"\n", + " messages=[{\"role\": \"system\", \"content\": system_prompt}, {\"role\": \"user\", \"content\": f\"This is the code to process: ```\\n{prompt}\\n```\"}]\n", + " try:\n", + " acumulated_response = \"\"\n", + " for chunk in invoke_model(model_name=model_name, prompt=messages, max_tokens=max_tokens, temperature=temperature):\n", + " acumulated_response += chunk\n", + " yield acumulated_response\n", + " except Exception as e:\n", + " return f\"Error: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a06d6af3", + "metadata": {}, + "outputs": [], + "source": [ + "with gr.Blocks() as ui:\n", + " gr.Markdown(\"# Coding Assistant\\n Choose from the available models to generate responses. Choose from either Documentation generation, or Test Code generation.\")\n", + " with gr.Row():\n", + " system_input = gr.Dropdown(label=\"Task Type\", choices=[\"Documentation\", \"Test Code Generation\"], value=\"Documentation\", interactive=True, visible=True)\n", + " with gr.Row():\n", + " prompt_input= gr.Textbox(label=\"Enter your prompt\", placeholder=\"Type your prompt here...\", lines=4)\n", + " response_output = gr.Textbox(label=\"Model Response\", lines=10, interactive=False)\n", + " with gr.Row():\n", + " max_tokens_input = gr.Slider(minimum=1, maximum=4096, value=1000, step=1, label=\"Max Tokens\")\n", + " temperature_input = gr.Slider(minimum=0.0, maximum=1.0, value=0.7, step=0.05, label=\"Temperature\")\n", + " model_selector = gr.Dropdown(\n", + " label=\"Select Model\", \n", + " choices=model_choices, \n", + " value=LLAMA_MODEL, \n", + " interactive=True\n", + " )\n", + " with gr.Row():\n", + " generate_button = gr.Button(\"Generate Response\", visible=True)\n", + " download_button = gr.Button(\"Download Response\", visible=True)\n", + " \n", + " generate_button.click(\n", + " fn=generate_response,\n", + " inputs=[system_input, prompt_input, max_tokens_input, temperature_input, model_selector],\n", + " outputs=response_output\n", + " )\n", + " download_button.click(\n", + " fn=save_text_to_selected_location,\n", + " inputs=[response_output],\n", + " outputs=None\n", + " )\n", + "\n", + "# Launch the UI\n", + "ui.launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/Pinecone-RAG-using-OCR.ipynb b/week5/community-contributions/Pinecone-RAG-using-OCR.ipynb new file mode 100644 index 0000000..e5362ec --- /dev/null +++ b/week5/community-contributions/Pinecone-RAG-using-OCR.ipynb @@ -0,0 +1,233 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "metadata": {}, + "source": [ + "\n", + "\n", + "# Just before starting \n", + "\n", + "You can check completely explained working here: https://medium.com/@muhammad.bese23seecs/building-a-rag-powered-pinecone-database-using-ocr-a-practical-guide-with-pakistani-law-d83e869e1458\n", + "\n", + "## Issues and Considerations\n", + "\n", + "This notebook requires a few installation to run. First is tesseract:\n", + "\n", + "For windows: https://stackoverflow.com/questions/46140485/tesseract-installation-in-windows\n", + "\n", + "For Linux: run on the cli \"sudo apt-get install tesseract-ocr\"\n", + "\n", + "For Mac: https://www.oreilly.com/library/view/building-computer-vision/9781838644673/95de5b35-436b-4668-8ca2-44970a6e2924.xhtml\n", + "\n", + "\n", + "Next install pytesseract in your environment\n", + "\n", + "For uv: uv pip install pytesseract\n", + "\n", + "For pip install: pip install pytesseract\n", + "\n", + "\n", + "You would require an OpenAI API key and Pinecone API key in your .env file\n" + ] + }, + { + "cell_type": "code", + "execution_count": 134, + "metadata": {}, + "outputs": [], + "source": [ + "# Install LangChain Unstructured (which requires unstructured under the hood)\n", + "\n", + "from langchain.embeddings.openai import OpenAIEmbeddings\n", + "from langchain.vectorstores import Pinecone\n", + "import getpass\n", + "from pinecone import Pinecone\n", + "import os\n", + "from pinecone import ServerlessSpec\n", + "from langchain_pinecone import PineconeVectorStore\n", + "from unstructured.partition.pdf import partition_pdf\n", + "import glob" + ] + }, + { + "cell_type": "code", + "execution_count": 135, + "metadata": {}, + "outputs": [], + "source": [ + "# Initialize embeddings and Pinecone vector store\n", + "embeddings = OpenAIEmbeddings() #" + ] + }, + { + "cell_type": "code", + "execution_count": 136, + "metadata": {}, + "outputs": [], + "source": [ + "\n", + "# Pinecone client\n", + "if not os.getenv(\"PINECONE_API_KEY\"):\n", + " os.environ[\"PINECONE_API_KEY\"] = getpass.getpass(\"Enter your Pinecone API key: \")\n", + "pinecone_api_key = os.environ.get(\"PINECONE_API_KEY\")\n", + "pc = Pinecone(api_key=pinecone_api_key)" + ] + }, + { + "cell_type": "code", + "execution_count": 137, + "metadata": {}, + "outputs": [], + "source": [ + "# Index Creation and its testing \n", + "index_name = \"lahore-cases\" # Replace the name with anything you like \n", + "if not pc.has_index(index_name):\n", + " pc.create_index(\n", + " name=index_name,\n", + " dimension=1536,\n", + " metric=\"cosine\",\n", + " spec=ServerlessSpec(cloud=\"aws\", region=\"us-east-1\"),\n", + " )\n" + ] + }, + { + "cell_type": "code", + "execution_count": 138, + "metadata": {}, + "outputs": [], + "source": [ + "# vector store\n", + "index = pc.Index(index_name)\n", + "vector_store = PineconeVectorStore(index=index, embedding=embeddings)\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "files = glob.glob(pathname='./**/*.pdf',recursive=True) # I have set recursive = True so that we can check subdirectories too.\n", + "print(len(files)) # confirm that you have all of the pdfs here with the correct path" + ] + }, + { + "cell_type": "code", + "execution_count": 140, + "metadata": {}, + "outputs": [], + "source": [ + "chunks = [] # The array to store the sections in \n", + "section_content = \"\"\n", + "index = 1\n", + "for file_path in files:\n", + " print(f\"File Number {index} completed:\",file_path) # To keep track of files\n", + " index+=1\n", + " elements = partition_pdf(file_path, languages=[\"eng\"],strategy=\"fast\")\n", + " for element in elements:\n", + " if element.category == 'NarrativeText': # meaning that it is simmple text \n", + " section_content+=element.text # Then append it to the already going section content\n", + " elif element.category==\"ListItem\":\n", + " chunks.append({\"page_content\":section_content,\"metadata\":element.metadata})\n", + " section_content=\"\" # Because a new sectionn has started\n", + " section_content += element.text # The string should start with the title of the text\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "len(set([chunk['metadata'].filename for chunk in chunks])) # Check if all of the completed files are here" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "chunks # How each chunk looks like" + ] + }, + { + "cell_type": "code", + "execution_count": 143, + "metadata": {}, + "outputs": [], + "source": [ + "from langchain_core.documents import Document\n", + "# How pinecone expects each chunk to be\n", + "docs = [Document(page_content=chunk['page_content'],metadata={\"source\":chunk['metadata'].filename}) for chunk in chunks]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "for doc in docs:\n", + " print(doc.page_content)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "# Now add all of the docs in the pinceone namespace\n", + "from uuid import uuid4\n", + "uuids = [str(uuid4()) for _ in range(len(docs))]\n", + "batch_size = 200\n", + "for i in range(0, len(docs), batch_size):\n", + " print(\"Current Batch Index is:\",i)\n", + " batch = docs[i:i+batch_size]\n", + " batch_ids = uuids[i:i+batch_size]\n", + " vector_store.add_documents(batch,ids=batch_ids)" + ] + }, + { + "cell_type": "code", + "execution_count": 147, + "metadata": {}, + "outputs": [], + "source": [ + "res = vector_store.similarity_search(query=\"Which act contains the words 'nothing from this act can be removed from the railways ..\",k=10)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": {}, + "outputs": [], + "source": [ + "res" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": ".venv", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.12.7" + } + }, + "nbformat": 4, + "nbformat_minor": 2 +} diff --git a/week5/community-contributions/RAG-based-academic-assistant-v3.ipynb b/week5/community-contributions/RAG-based-academic-assistant-v3.ipynb new file mode 100644 index 0000000..7899ff8 --- /dev/null +++ b/week5/community-contributions/RAG-based-academic-assistant-v3.ipynb @@ -0,0 +1,409 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "97a93fee-6bbd-477b-aba8-577d318a9f9d", + "metadata": {}, + "source": [ + "# AI-Powered Academic Knowledge Assistant\n", + "AI-powered RAG (Retrieval-Augmented Generation) system that transforms document collections into queryable knowledge bases using OpenAI embeddings and vector search. Features configurable chunking, file size limits, and retrieval parameters with a Gradio interface for processing PDFs and generating contextually-aware responses via LangChain and ChromaDB." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3589eee0-ce34-42f4-b538-b43f3b0d9f6f", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import glob\n", + "from dotenv import load_dotenv\n", + "import gradio as gr\n", + "import shutil\n", + "import tiktoken\n", + "import time\n", + "import uuid\n", + "from typing import List, Tuple, Optional\n", + "\n", + "# imports for langchain and Chroma\n", + "from langchain.document_loaders import DirectoryLoader, TextLoader, PyPDFLoader\n", + "from langchain.text_splitter import CharacterTextSplitter\n", + "from langchain.schema import Document\n", + "from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n", + "from langchain_chroma import Chroma\n", + "from langchain.memory import ConversationBufferMemory\n", + "from langchain.chains import ConversationalRetrievalChain\n", + "from langchain.embeddings import HuggingFaceEmbeddings\n", + "\n", + "from langchain_community.document_loaders import PyPDFLoader, TextLoader\n", + "from langchain.docstore.document import Document\n", + "\n", + "# Load environment variables\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n", + "\n", + "# Global variables to store the current setup\n", + "current_vectorstore = None\n", + "current_conversation_chain = None\n", + "processing_status = \"\"\n", + "\n", + "def count_tokens(text: str, model: str = \"gpt-4o-mini\") -> int:\n", + " \"\"\"Count tokens in text using tiktoken\"\"\"\n", + " try:\n", + " encoding = tiktoken.encoding_for_model(model)\n", + " return len(encoding.encode(text))\n", + " except:\n", + " # Fallback estimation: roughly 4 characters per token\n", + " return len(text) // 4\n", + "\n", + "def filter_chunks_by_tokens(chunks: List[Document], max_total_tokens: int = 250000) -> List[Document]:\n", + " \"\"\"Filter chunks to stay within token limits\"\"\"\n", + " filtered_chunks = []\n", + " total_tokens = 0\n", + " \n", + " for chunk in chunks:\n", + " chunk_tokens = count_tokens(chunk.page_content)\n", + " \n", + " # Skip individual chunks that are too large (shouldn't happen with proper splitting)\n", + " if chunk_tokens > 8000: # Individual chunk limit\n", + " continue\n", + " \n", + " if total_tokens + chunk_tokens <= max_total_tokens:\n", + " filtered_chunks.append(chunk)\n", + " total_tokens += chunk_tokens\n", + " else:\n", + " break\n", + " \n", + " return filtered_chunks\n", + "\n", + "def add_metadata(doc, doc_type, file_path):\n", + " \"\"\"Add metadata including document type and file information\"\"\"\n", + " doc.metadata[\"doc_type\"] = doc_type\n", + " doc.metadata[\"file_path\"] = file_path\n", + " doc.metadata[\"file_name\"] = os.path.basename(file_path)\n", + " return doc\n", + "\n", + "def check_file_size(file_path, max_size_bytes):\n", + " \"\"\"Check if file size is within the limit\"\"\"\n", + " try:\n", + " file_size = os.path.getsize(file_path)\n", + " return file_size <= max_size_bytes, file_size\n", + " except OSError:\n", + " return False, 0\n", + "\n", + "def load_pdfs_with_size_limit(folder_path, doc_type, max_size_bytes):\n", + " \"\"\"Load PDF files from a folder with size restrictions\"\"\"\n", + " pdf_files = glob.glob(os.path.join(folder_path, \"**/*.pdf\"), recursive=True)\n", + " loaded_docs = []\n", + " skipped_files = []\n", + " \n", + " for pdf_file in pdf_files:\n", + " is_valid_size, file_size = check_file_size(pdf_file, max_size_bytes)\n", + " \n", + " if is_valid_size:\n", + " try:\n", + " loader = PyPDFLoader(pdf_file)\n", + " docs = loader.load()\n", + " docs_with_metadata = [add_metadata(doc, doc_type, pdf_file) for doc in docs]\n", + " loaded_docs.extend(docs_with_metadata)\n", + " except Exception as e:\n", + " skipped_files.append((pdf_file, f\"Loading error: {str(e)}\"))\n", + " else:\n", + " file_size_mb = file_size / 1024 / 1024\n", + " skipped_files.append((pdf_file, f\"File too large: {file_size_mb:.2f} MB\"))\n", + " \n", + " return loaded_docs, skipped_files\n", + "\n", + "def process_documents(knowledge_base_dir: str, max_file_size_mb: float, chunk_size: int, chunk_overlap: int) -> Tuple[str, str]:\n", + " \"\"\"Process documents and create vector store\"\"\"\n", + " global current_vectorstore, current_conversation_chain\n", + " \n", + " try:\n", + " # Validate directory\n", + " if not knowledge_base_dir or not knowledge_base_dir.strip():\n", + " return \"❌ Error: Please enter a directory path!\", \"\"\n", + " \n", + " directory_path = knowledge_base_dir.strip()\n", + " \n", + " if not os.path.exists(directory_path):\n", + " return \"❌ Error: Directory does not exist! Please check the path.\", \"\"\n", + " \n", + " # Configuration\n", + " MAX_FILE_SIZE_BYTES = int(max_file_size_mb * 1024 * 1024)\n", + " \n", + " # Find folders\n", + " if directory_path.endswith('*'):\n", + " folders = glob.glob(directory_path)\n", + " else:\n", + " folders = glob.glob(os.path.join(directory_path, \"*\"))\n", + " \n", + " if not folders:\n", + " return \"❌ Error: No folders found in the specified directory!\", \"\"\n", + " \n", + " # Process documents\n", + " documents = []\n", + " all_skipped_files = []\n", + " status_lines = []\n", + " \n", + " status_lines.append(f\"🔍 Processing folders with {max_file_size_mb} MB file size limit...\")\n", + " status_lines.append(\"-\" * 60)\n", + " \n", + " for folder in folders:\n", + " if os.path.isdir(folder):\n", + " doc_type = os.path.basename(folder)\n", + " status_lines.append(f\"📁 Processing folder: {doc_type}\")\n", + " \n", + " folder_docs, skipped_files = load_pdfs_with_size_limit(folder, doc_type, MAX_FILE_SIZE_BYTES)\n", + " documents.extend(folder_docs)\n", + " all_skipped_files.extend(skipped_files)\n", + " \n", + " if folder_docs:\n", + " status_lines.append(f\" ✅ Loaded {len(folder_docs)} document pages\")\n", + " if skipped_files:\n", + " status_lines.append(f\" ⚠️ Skipped {len(skipped_files)} files\")\n", + " \n", + " if not documents:\n", + " error_msg = \"❌ No PDF documents were loaded successfully.\"\n", + " if all_skipped_files:\n", + " error_msg += f\"\\n\\nAll {len(all_skipped_files)} files were skipped:\"\n", + " for file_path, reason in all_skipped_files[:10]: # Show first 10\n", + " error_msg += f\"\\n • {os.path.basename(file_path)}: {reason}\"\n", + " if len(all_skipped_files) > 10:\n", + " error_msg += f\"\\n ... and {len(all_skipped_files) - 10} more\"\n", + " return error_msg, \"\"\n", + " \n", + " # Text splitting\n", + " status_lines.append(\"\\n\" + \"=\"*40)\n", + " status_lines.append(\"✂️ TEXT SPLITTING\")\n", + " status_lines.append(\"=\"*40)\n", + " \n", + " text_splitter = CharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)\n", + " chunks = text_splitter.split_documents(documents)\n", + " \n", + " # Filter chunks by token count to prevent API errors\n", + " status_lines.append(\"🔢 Checking token limits...\")\n", + " original_chunk_count = len(chunks)\n", + " chunks = filter_chunks_by_tokens(chunks, max_total_tokens=250000)\n", + " \n", + " if len(chunks) < original_chunk_count:\n", + " status_lines.append(f\"⚠️ Filtered from {original_chunk_count} to {len(chunks)} chunks to stay within token limits\")\n", + " \n", + " # Create vectorstore\n", + " status_lines.append(\"🧮 Creating vector embeddings...\")\n", + " embeddings = OpenAIEmbeddings()\n", + " \n", + " # Use a temporary database name\n", + " db_name = \"temp_vector_db\"\n", + " \n", + " # Delete if already exists\n", + " if os.path.exists(db_name):\n", + " shutil.rmtree(db_name)\n", + " \n", + " # Create vectorstore\n", + " vectorstore = Chroma.from_documents(\n", + " documents=chunks, \n", + " embedding=embeddings, \n", + " persist_directory=db_name\n", + " )\n", + " \n", + " # Update global variables\n", + " current_vectorstore = vectorstore\n", + " \n", + " # Create conversation chain\n", + " llm = ChatOpenAI(temperature=0.7, model_name=\"gpt-4o-mini\")\n", + " memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)\n", + " retriever = vectorstore.as_retriever(search_kwargs={\"k\": 25})\n", + " current_conversation_chain = ConversationalRetrievalChain.from_llm(\n", + " llm=llm, \n", + " retriever=retriever, \n", + " memory=memory\n", + " )\n", + " \n", + " # Summary statistics\n", + " status_lines.append(\"\\n\" + \"=\"*40)\n", + " status_lines.append(\"📊 SUMMARY\")\n", + " status_lines.append(\"=\"*40)\n", + " status_lines.append(f\"✅ Total PDFs processed: {len(set(doc.metadata['file_path'] for doc in documents))}\")\n", + " status_lines.append(f\"📄 Total document pages: {len(documents)}\")\n", + " status_lines.append(f\"🧩 Total text chunks: {len(chunks)}\")\n", + " status_lines.append(f\"📁 Document types: {', '.join(set(doc.metadata['doc_type'] for doc in documents))}\")\n", + " status_lines.append(f\"🗃️ Vector store size: {vectorstore._collection.count()} embeddings\")\n", + " \n", + " if all_skipped_files:\n", + " status_lines.append(f\"\\n⚠️ Skipped files: {len(all_skipped_files)}\")\n", + " for file_path, reason in all_skipped_files[:5]: # Show first 5\n", + " status_lines.append(f\" • {os.path.basename(file_path)}: {reason}\")\n", + " if len(all_skipped_files) > 5:\n", + " status_lines.append(f\" ... and {len(all_skipped_files) - 5} more\")\n", + " \n", + " success_msg = \"✅ Knowledge base successfully created and ready for questions!\"\n", + " detailed_status = \"\\n\".join(status_lines)\n", + " \n", + " return success_msg, detailed_status\n", + " \n", + " except Exception as e:\n", + " error_msg = f\"❌ Error processing documents: {str(e)}\"\n", + " return error_msg, \"\"\n", + "\n", + "def chat_with_documents(message, history, num_chunks):\n", + " \"\"\"Chat with the processed documents\"\"\"\n", + " global current_conversation_chain, current_vectorstore\n", + " \n", + " if current_conversation_chain is None:\n", + " return \"❌ Please process documents first before asking questions!\"\n", + " \n", + " try:\n", + " # Update retriever with new chunk count\n", + " if current_vectorstore is not None:\n", + " retriever = current_vectorstore.as_retriever(search_kwargs={\"k\": num_chunks})\n", + " current_conversation_chain.retriever = retriever\n", + " \n", + " result = current_conversation_chain.invoke({\"question\": message})\n", + " return result[\"answer\"]\n", + " \n", + " except Exception as e:\n", + " return f\"❌ Error generating response: {str(e)}\"\n", + "\n", + "def reset_conversation():\n", + " \"\"\"Reset the conversation memory\"\"\"\n", + " global current_conversation_chain\n", + " if current_conversation_chain is not None:\n", + " current_conversation_chain.memory.clear()\n", + " return \"✅ Conversation history cleared!\"\n", + " return \"No active conversation to reset.\"\n", + "\n", + "# Create Gradio Interface\n", + "with gr.Blocks(title=\"AI-Powered Academic Knowledge Assistant\", theme=gr.themes.Soft()) as app:\n", + " gr.Markdown(\"# 🎓 AI-Powered Academic Knowledge Assistant\")\n", + " gr.Markdown(\"Transform your entire document library into an intelligent, searchable AI tutor that answers questions instantly.\")\n", + " \n", + " with gr.Tabs():\n", + " # Configuration Tab\n", + " with gr.Tab(\"⚙️ Configuration\"):\n", + " gr.Markdown(\"### 📁 Document Processing Settings\")\n", + " \n", + " gr.Markdown(\"💡 **Tip:** Copy and paste your folder path here. On mobile, you can use file manager apps to copy folder paths.\")\n", + " \n", + " with gr.Row():\n", + " with gr.Column():\n", + " knowledge_dir = gr.Textbox(\n", + " label=\"Knowledge Base Directory\",\n", + " value=r\"C:\\Users\\Documents\\Syllabi\\Georgia Tech\\Spring 22\\Microwave Design\",\n", + " placeholder=\"Enter or paste your document directory path\",\n", + " lines=1\n", + " )\n", + " \n", + " max_file_size = gr.Slider(\n", + " label=\"Max File Size (MB)\",\n", + " minimum=0.5,\n", + " maximum=50,\n", + " value=4,\n", + " step=0.5\n", + " )\n", + " \n", + " with gr.Column():\n", + " chunk_size = gr.Slider(\n", + " label=\"Chunk Size (characters)\",\n", + " minimum=200,\n", + " maximum=1500,\n", + " value=800,\n", + " step=100,\n", + " info=\"Smaller chunks = better token management\"\n", + " )\n", + " \n", + " chunk_overlap = gr.Slider(\n", + " label=\"Chunk Overlap (characters)\",\n", + " minimum=0,\n", + " maximum=300,\n", + " value=150,\n", + " step=25,\n", + " info=\"Overlap preserves context between chunks\"\n", + " )\n", + " \n", + " process_btn = gr.Button(\"🚀 Process Documents\", variant=\"primary\", size=\"lg\")\n", + " \n", + " with gr.Row():\n", + " status_output = gr.Textbox(\n", + " label=\"Status\",\n", + " lines=2,\n", + " max_lines=2\n", + " )\n", + " \n", + " detailed_output = gr.Textbox(\n", + " label=\"Detailed Processing Log\",\n", + " lines=15,\n", + " max_lines=20\n", + " )\n", + " \n", + " # Chat Tab\n", + " with gr.Tab(\"💬 Chat\"):\n", + " gr.Markdown(\"### 🤖 Ask Questions About Your Documents\")\n", + " \n", + " with gr.Row():\n", + " with gr.Column(scale=1):\n", + " num_chunks = gr.Slider(\n", + " label=\"Number of chunks to retrieve\",\n", + " minimum=1,\n", + " maximum=50,\n", + " value=25,\n", + " step=1\n", + " )\n", + " \n", + " reset_btn = gr.Button(\"🗑️ Clear Chat History\", variant=\"secondary\")\n", + " reset_output = gr.Textbox(label=\"Reset Status\", lines=1)\n", + " \n", + " with gr.Column(scale=3):\n", + " chatbot = gr.ChatInterface(\n", + " fn=lambda msg, history: chat_with_documents(msg, history, num_chunks.value),\n", + " type=\"messages\",\n", + " title=\"Academic Assistant Chat\",\n", + " description=\"Ask questions about your processed documents\"\n", + " )\n", + " \n", + " # Event handlers\n", + " process_btn.click(\n", + " fn=process_documents,\n", + " inputs=[knowledge_dir, max_file_size, chunk_size, chunk_overlap],\n", + " outputs=[status_output, detailed_output]\n", + " )\n", + " \n", + " reset_btn.click(\n", + " fn=reset_conversation,\n", + " outputs=reset_output\n", + " )\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9eb807e0-194b-48dd-a1e9-b1b9b8a99620", + "metadata": {}, + "outputs": [], + "source": [ + "app.launch(share=True, inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/RAG_Using_Gemini b/week5/community-contributions/RAG_Using_Gemini new file mode 100644 index 0000000..73709ba --- /dev/null +++ b/week5/community-contributions/RAG_Using_Gemini @@ -0,0 +1 @@ +Testing diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Gmail_API_Credential_Guide.ipynb b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Gmail_API_Credential_Guide.ipynb new file mode 100644 index 0000000..1f5e1c6 --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Gmail_API_Credential_Guide.ipynb @@ -0,0 +1,154 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "35177005-196a-48b3-bf92-fa37d84693f4", + "metadata": {}, + "source": [ + "# Gmail API Credential Guide" + ] + }, + { + "cell_type": "markdown", + "id": "7bcad9ee-cd11-4b12-834d-9f1ddcefb190", + "metadata": {}, + "source": [ + "Use Gmail API to Read Your Emails\n", + "1. Set up a Google Cloud Project\n", + "\n", + " Go to Google Cloud Platform(GCP) Console\n", + "\n", + " Create a new project\n", + "\n", + "2. Enable the Gmail API for that project\n", + "\n", + " Select the created project and go to \"APIs & services\" page\n", + "\n", + " Click \"+ Enable APIs and services\" button, search \"Gmail API\" and enable it\n", + "\n", + "3. Go to \"OAuth Consent Screen\" and configure:\n", + "\n", + " Choose External and Fill in app name, dedveloper email, etc.\n", + "\n", + "4. Create OAuth Credentials\n", + "\n", + " Go to APIs & Services > Credentials\n", + "\n", + " Click \"+ Create Credentials\" > \"OAuth client ID\"\n", + "\n", + " Choose Desktop App\n", + "\n", + " Download the generated credentials.json\n", + "\n", + " Sometimes, GCP will navigate you to \"Google Auth Platform\" > \"Clients\", and you can click \"+ Create client\" here to create the OAuth Credentials\n", + "\n", + " \n", + "5. Add Test Users for Gmail API OAuth Access\n", + " \n", + " Go to \"APIs & Services\" > \"OAuth consent screen\" > \"Audience\" > \"Test Users\"\n", + "\n", + " Add the email account from which you want to extract email content.\n", + "\n", + "\n", + "6. Create 'credentials' folders to store gmail credential and user tokens" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bc86bec0-bda8-4e9e-9c85-423179a99981", + "metadata": {}, + "outputs": [], + "source": [ + "# !pip install --upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4270e52e-378c-4127-bd52-1d082e9834e0", + "metadata": {}, + "outputs": [], + "source": [ + "from __future__ import print_function\n", + "import os.path\n", + "import base64\n", + "import re\n", + "from email import message_from_bytes\n", + "from google.oauth2.credentials import Credentials\n", + "from google_auth_oauthlib.flow import InstalledAppFlow\n", + "from googleapiclient.discovery import build\n", + "\n", + "# If modifying these SCOPES, delete the token.json\n", + "SCOPES = ['https://www.googleapis.com/auth/gmail.readonly']\n", + "PORT = 18000\n", + "\n", + "def main():\n", + " creds = None\n", + " # token.json stores the user's access and refresh tokens\n", + " if os.path.exists('token.json'):\n", + " creds = Credentials.from_authorized_user_file('token.json', SCOPES)\n", + " else:\n", + " flow = InstalledAppFlow.from_client_secrets_file('credentials/gmail_credentials.json', SCOPES)\n", + " creds = flow.run_local_server(port=PORT)\n", + " with open('token.json', 'w') as token:\n", + " token.write(creds.to_json())\n", + "\n", + " service = build('gmail', 'v1', credentials=creds)\n", + "\n", + " # Get the latest message\n", + " results = service.users().messages().list(userId='me', maxResults=1).execute()\n", + " messages = results.get('messages', [])\n", + "\n", + " if not messages:\n", + " print(\"No messages found.\")\n", + " return\n", + "\n", + " msg = service.users().messages().get(userId='me', id=messages[0]['id'], format='raw').execute()\n", + " raw_msg = base64.urlsafe_b64decode(msg['raw'].encode('ASCII'))\n", + " email_message = message_from_bytes(raw_msg)\n", + "\n", + " subject = email_message['Subject']\n", + " print(\"Subject:\", subject)\n", + "\n", + " # Extract text/plain body\n", + " for part in email_message.walk():\n", + " if part.get_content_type() == 'text/plain':\n", + " print(\"Body:\")\n", + " print(part.get_payload(decode=True).decode('utf-8'))\n", + "\n", + "if __name__ == '__main__':\n", + " main()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5ff68e06-3cfb-48ae-9dad-fa431d0d548a", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Google_Workspace_API_Credential_Guide.ipynb b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Google_Workspace_API_Credential_Guide.ipynb new file mode 100644 index 0000000..c300ec4 --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Google_Workspace_API_Credential_Guide.ipynb @@ -0,0 +1,294 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "35177005-196a-48b3-bf92-fa37d84693f4", + "metadata": {}, + "source": [ + "# Google Workspace API Credential Guide" + ] + }, + { + "cell_type": "markdown", + "id": "7bcad9ee-cd11-4b12-834d-9f1ddcefb190", + "metadata": {}, + "source": [ + "Use Google Drive API to Read files in Google Workspace \n", + "1. Set up a Google Cloud Project\n", + "\n", + " Go to Google Cloud Platform(GCP) Console\n", + "\n", + " Create a new project\n", + "\n", + "2. Enable the Gmail API for that project\n", + "\n", + " Select the created project and go to \"APIs & services\" page\n", + "\n", + " Click \"+ Enable APIs and services\" button, enable these APIs: Google Drive API, Google Docs API, Google Sheets API, and Google Slides API \n", + "\n", + "3. Go to \"OAuth Consent Screen\" and configure:\n", + "\n", + " Choose External and Fill in app name, dedveloper email, etc.\n", + "\n", + "4. Create OAuth Credentials\n", + "\n", + " Go to APIs & Services > Credentials\n", + "\n", + " Click \"+ Create Credentials\" > \"OAuth client ID\"\n", + "\n", + " Choose Desktop App\n", + "\n", + " Download the generated credentials.json\n", + "\n", + " Sometimes, GCP will navigate you to \"Google Auth Platform\" > \"Clients\", and you can click \"+ Create client\" here to create the OAuth Credentials\n", + "\n", + " \n", + "5. Add Test Users for Gmail API OAuth Access\n", + " \n", + " Go to \"APIs & Services\" > \"OAuth consent screen\" > \"Audience\" > \"Test Users\"\n", + "\n", + " Add the email account from which you want to extract email content.\n", + "\n", + "\n", + "6. Create 'credentials' folders to store google workspace credential and user tokens" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bc86bec0-bda8-4e9e-9c85-423179a99981", + "metadata": {}, + "outputs": [], + "source": [ + "# !pip install PyPDF2\n", + "# !pip install --upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4270e52e-378c-4127-bd52-1d082e9834e0", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5ff68e06-3cfb-48ae-9dad-fa431d0d548a", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "69c20f2d-2f49-408c-8700-f12d6745efd3", + "metadata": {}, + "outputs": [], + "source": [ + "from google_auth_oauthlib.flow import InstalledAppFlow\n", + "from googleapiclient.discovery import build\n", + "from google.oauth2.credentials import Credentials\n", + "from googleapiclient.http import MediaIoBaseDownload\n", + "import os\n", + "\n", + "import io\n", + "from PyPDF2 import PdfReader\n", + "from langchain.vectorstores import Chroma\n", + "from langchain.embeddings import OpenAIEmbeddings\n", + "from langchain.schema import Document\n", + "\n", + "GOOGLE_WORKSPACE_SCOPES = [\"https://www.googleapis.com/auth/drive.readonly\",\n", + " 'https://www.googleapis.com/auth/documents.readonly',\n", + " 'https://www.googleapis.com/auth/spreadsheets.readonly',\n", + " 'https://www.googleapis.com/auth/presentations.readonly'\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7164903b-be81-46b2-8c04-886397599c27", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_google_doc(docs_service, file_id):\n", + " doc = docs_service.documents().get(documentId=file_id).execute()\n", + " content = \"\"\n", + " for elem in doc.get(\"body\", {}).get(\"content\", []):\n", + " if \"paragraph\" in elem:\n", + " for run in elem[\"paragraph\"][\"elements\"]:\n", + " content += run.get(\"textRun\", {}).get(\"content\", \"\")\n", + " return content.strip()\n", + "\n", + "def extract_google_sheet(service, file_id):\n", + " # Get spreadsheet metadata\n", + " spreadsheet = service.spreadsheets().get(spreadsheetId=file_id).execute()\n", + " all_text = \"\"\n", + "\n", + " # Loop through each sheet\n", + " for sheet in spreadsheet.get(\"sheets\", []):\n", + " title = sheet[\"properties\"][\"title\"]\n", + " result = service.spreadsheets().values().get(\n", + " spreadsheetId=file_id,\n", + " range=title\n", + " ).execute()\n", + "\n", + " values = result.get(\"values\", [])\n", + " sheet_text = f\"### Sheet: {title} ###\\n\"\n", + " sheet_text += \"\\n\".join([\", \".join(row) for row in values])\n", + " all_text += sheet_text + \"\\n\\n\"\n", + "\n", + " return all_text.strip()\n", + "\n", + "\n", + "def extract_google_slide(slides_service, file_id):\n", + " pres = slides_service.presentations().get(presentationId=file_id).execute()\n", + " text = \"\"\n", + " for slide in pres.get(\"slides\", []):\n", + " for element in slide.get(\"pageElements\", []):\n", + " shape = element.get(\"shape\")\n", + " if shape:\n", + " for p in shape.get(\"text\", {}).get(\"textElements\", []):\n", + " if \"textRun\" in p:\n", + " text += p[\"textRun\"][\"content\"]\n", + " return text.strip()\n", + "\n", + "def extract_pdf_from_drive(drive_service, file_id, filename='downloaded.pdf'):\n", + " request = drive_service.files().get_media(fileId=file_id)\n", + " fh = io.BytesIO()\n", + " downloader = MediaIoBaseDownload(fh, request)\n", + " done = False\n", + " while not done:\n", + " _, done = downloader.next_chunk()\n", + " fh.seek(0)\n", + " reader = PdfReader(fh)\n", + " return \"\\n\".join([page.extract_text() for page in reader.pages if page.extract_text()])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5f2edc68-f9f8-4cba-810e-159bea4fe4ac", + "metadata": {}, + "outputs": [], + "source": [ + "def get_creds():\n", + " if os.path.exists(\"token.json\"):\n", + " creds = Credentials.from_authorized_user_file(\"token.json\", SCOPES)\n", + " else:\n", + " flow = InstalledAppFlow.from_client_secrets_file(\"credentials/google_drive_workspace_credentials.json\", SCOPES)\n", + " creds = flow.run_local_server(port=0)\n", + " with open(\"token.json\", \"w\") as token:\n", + " token.write(creds.to_json())\n", + " return creds\n", + " \n", + "\n", + "def get_folder_id_by_name(drive_service, folder_name):\n", + " query = f\"mimeType='application/vnd.google-apps.folder' and name='{folder_name}' and trashed=false\"\n", + " results = drive_service.files().list(\n", + " q=query,\n", + " fields=\"files(id, name)\",\n", + " pageSize=1\n", + " ).execute()\n", + "\n", + " folders = results.get(\"files\", [])\n", + " if not folders:\n", + " raise ValueError(f\"❌ Folder named '{folder_name}' not found.\")\n", + " return folders[0]['id']\n", + "\n", + "\n", + "def extract_docs_from_google_workspace(folder_name):\n", + " info = \"\"\n", + " \n", + " creds = get_creds()\n", + "\n", + " file_types = {\n", + " 'application/vnd.google-apps.document': lambda fid: extract_google_doc(docs_service, fid),\n", + " 'application/vnd.google-apps.spreadsheet': lambda fid: extract_google_sheet(sheets_service, fid),\n", + " 'application/vnd.google-apps.presentation': lambda fid: extract_google_slide(slides_service, fid),\n", + " 'application/pdf': lambda fid: extract_pdf_from_drive(drive_service, fid),\n", + " }\n", + " \n", + " drive_service = build(\"drive\", \"v3\", credentials=creds)\n", + " docs_service = build('docs', 'v1', credentials=creds)\n", + " sheets_service = build('sheets', 'v4', credentials=creds)\n", + " slides_service = build('slides', 'v1', credentials=creds)\n", + "\n", + " folder_id = get_folder_id_by_name(drive_service, folder_name)\n", + " info += f\"Collection files from folder: {folder_name}\\n\"\n", + " \n", + " query = (\n", + " f\"'{folder_id}' in parents and (\"\n", + " 'mimeType=\"application/vnd.google-apps.document\" or '\n", + " 'mimeType=\"application/vnd.google-apps.spreadsheet\" or '\n", + " 'mimeType=\"application/vnd.google-apps.presentation\" or '\n", + " 'mimeType=\"application/pdf\")'\n", + " )\n", + " \n", + " results = drive_service.files().list(\n", + " q=query,\n", + " fields=\"files(id, name, mimeType)\",\n", + " pageSize=20\n", + " ).execute()\n", + "\n", + " docs = []\n", + " summary_info = {\n", + " 'application/vnd.google-apps.document': {'file_type': 'Google Doc', 'count': 0},\n", + " 'application/vnd.google-apps.spreadsheet': {'file_type': 'Google Sheet', 'count': 0},\n", + " 'application/vnd.google-apps.presentation': {'file_type': 'Google Silde', 'count': 0},\n", + " 'application/pdf': {'file_type': 'PDF', 'count': 0}\n", + " }\n", + " for file in results.get(\"files\", []):\n", + " extractor = file_types.get(file['mimeType'])\n", + " if extractor:\n", + " try:\n", + " content = extractor(file[\"id\"])\n", + " if content:\n", + " docs.append(Document(page_content=content, metadata={\"source\": file[\"name\"]}))\n", + " summary_info[file['mimeType']]['count'] += 1\n", + " except Exception as e:\n", + " print(f\"❌ Error processing {file['name']}: {e}\")\n", + " \n", + " total = 0;\n", + " for file_type, element in summary_info.items():\n", + " total += element['count']\n", + " info += f\"Found {element['count']} {element['file_type']} files\\n\"\n", + " info += f\"Total documents loaded: {total}\"\n", + " return docs, info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8a9da5c9-415c-4856-973a-627a1790f38d", + "metadata": {}, + "outputs": [], + "source": [ + "docs, info = extract_docs_from_google_workspace(\"google_workspace_knowledge_base\")" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Outlook_API_Credential_Guide.ipynb b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Outlook_API_Credential_Guide.ipynb new file mode 100644 index 0000000..785d5dd --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Outlook_API_Credential_Guide.ipynb @@ -0,0 +1,178 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "35177005-196a-48b3-bf92-fa37d84693f4", + "metadata": {}, + "source": [ + "# Outlook API Credential Guide" + ] + }, + { + "cell_type": "markdown", + "id": "7bcad9ee-cd11-4b12-834d-9f1ddcefb190", + "metadata": {}, + "source": [ + "Extract Outlook Emails via Microsoft Graph API\n", + "\n", + "1. Register Your App on Azure Portal\n", + "\n", + " Go to Azure Portal > Azure Active Directory > App registrations\n", + "\n", + " Click “New registration”\n", + "\n", + " Choose Mobole/Desktop app\n", + " \n", + " After creation, note the Application (client) ID\n", + "\n", + "2. API Permissions\n", + "\n", + " Go to API permissions tab\n", + "\n", + " Click Add permission > Microsoft Graph > Delegated\n", + "\n", + " Choose: Mail.Read\n", + "\n", + " Click Grant admin consent\n", + "\n", + "3. Allow public client flows\n", + "\n", + " Navigate to: Azure Active Directory > App registrations > Your App\n", + "\n", + " Go to Authentication tab\n", + "\n", + " Under \"Advanced settings\" → \"Allow public client flows\", set to \"Yes\"\n", + "\n", + " Save changes" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bc86bec0-bda8-4e9e-9c85-423179a99981", + "metadata": {}, + "outputs": [], + "source": [ + "!pip install msal requests" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4270e52e-378c-4127-bd52-1d082e9834e0", + "metadata": {}, + "outputs": [], + "source": [ + "from msal import PublicClientApplication\n", + "import os\n", + "from dotenv import load_dotenv\n", + "import requests" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5ff68e06-3cfb-48ae-9dad-fa431d0d548a", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv()\n", + "\n", + "CLIENT_ID = os.getenv(\"AZURE_CLIENT_ID\")\n", + "AUTHORITY = \"https://login.microsoftonline.com/common\" \n", + "SCOPES = [\"Mail.Read\"]\n", + "\n", + "app = PublicClientApplication(CLIENT_ID, authority=AUTHORITY)\n", + "\n", + "flow = app.initiate_device_flow(scopes=SCOPES)\n", + "print(\"Go to:\", flow[\"verification_uri\"])\n", + "print(\"Enter code:\", flow[\"user_code\"])\n", + "\n", + "result = app.acquire_token_by_device_flow(flow)\n", + "\n", + "if \"access_token\" not in result:\n", + " raise Exception(\"Failed to authenticate:\", result)\n", + "\n", + "access_token = result[\"access_token\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9c7f97da-68cc-4923-b280-1ddf7e5b7aa3", + "metadata": {}, + "outputs": [], + "source": [ + "print(\"Granted scopes:\", result.get(\"scope\"))\n", + "\n", + "headers = {\n", + " \"Authorization\": f\"Bearer {access_token}\",\n", + " \"Prefer\": \"outlook.body-content-type='text'\"\n", + "}\n", + "\n", + "query = (\n", + " \"https://graph.microsoft.com/v1.0/me/messages\"\n", + " \"?$top=1\"\n", + " \"&$select=id,subject,receivedDateTime,body\"\n", + ")\n", + "\n", + "all_emails = []\n", + "\n", + "while query:\n", + " response = requests.get(query, headers=headers)\n", + "\n", + " if not response.ok:\n", + " print(response.text)\n", + " print(f\"❌ HTTP {response.status_code}: {response.text}\")\n", + " break\n", + "\n", + " try:\n", + " res = response.json()\n", + " except ValueError:\n", + " print(\"❌ Invalid JSON:\", response.text)\n", + " break\n", + "\n", + " for msg in res.get(\"value\", []):\n", + " all_emails.append({\n", + " \"id\": msg.get(\"id\"),\n", + " \"subject\": msg.get(\"subject\", \"\"),\n", + " \"body\": msg.get(\"body\", {}).get(\"content\", \"\"),\n", + " \"date\": msg.get(\"receivedDateTime\", \"\")\n", + " })\n", + "\n", + " query = res.get(\"@odata.nextLink\")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "e29493b6-0a9e-4106-93c9-e58ff6aa0f97", + "metadata": {}, + "outputs": [], + "source": [ + "all_emails" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Week5_Exercise_Personal_Knowledge_Assistant.ipynb b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Week5_Exercise_Personal_Knowledge_Assistant.ipynb new file mode 100644 index 0000000..9bab26f --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/Week5_Exercise_Personal_Knowledge_Assistant.ipynb @@ -0,0 +1,1862 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "e096ce5d-71a1-4fde-b171-8b9fed16cd7b", + "metadata": {}, + "source": [ + "# Personal Knowledge Assistant" + ] + }, + { + "cell_type": "markdown", + "id": "7bcad9ee-cd11-4b12-834d-9f1ddcefb190", + "metadata": {}, + "source": [ + "## Week 5 exercise\n", + "\n", + "\n", + "### Features:\n", + "1. Chat powered of uploaded knowlege\n", + "\n", + " The system prompt is designed to make the chatbot simulate a person based on the provided documents.\n", + "\n", + "2. Load files from local system\n", + "\n", + " Reuse code from bluebells1 [Wk5-final-multi-doc-type-KB.ipynb](../Wk5-final-multi-doc-type-KB.ipynb). Really appreciate it!\n", + "\n", + " Choose a folder located in the same directory as this script to extract content from. You can also specify subfolders to exclude from the extraction.\n", + "\n", + "3. Load emails from Gmail\n", + "\n", + " Enter an alias first, and a Google popup will guide you to grant permissions and log in, then extract emails for your specified time range\n", + "\n", + "4. Load emails from Outlook\n", + "\n", + " First, enter an alias. After clicking the 'Get Verification Code' button, a URI and code will appear in the 'Verification Instructions' textbox. Visit the Outlook website using the code, and follow the guide to grant permissions and complete the login.\n", + " Then, extract emails for your specified time range\n", + " \n", + "5. Load files from Google Workspace\n", + "\n", + " Enter with an alias first, and Google popup will guide you to grant permissions and log in, then extract emails for your specified folder in your Google Drive\n", + "\n", + "\n", + "### TO-DO Features:\n", + "1. Load messages from Slack\n", + "2. Use local inference/embedding models (llama) instead of relying on OpenAI-hosted models \n", + "3. Optimize Gmail/Outlook/Google Workspace login logic\n", + "4. Label different files. For example, extract prrivate and work emails respectively and store them into different vector stores\n", + "5. Add vector visualization\n", + "\n", + "### Requirements:\n", + "1. Store gmail credential json file under the 'credentials' folder\n", + "\n", + " The setup and configuration steps for Gmail API are in this guide: [Gmail_API_Credential_Guide](./Gmail_API_Credential_Guide.ipynb)\n", + "\n", + "2. Set AZURE_CLIENT_ID in .env file\n", + "\n", + " The setup and configuration steps for Outlook API are in this guide: [Outlook_API_Credential_Guide](./Outlook_API_Credential_Guide.ipynb)\n", + "\n", + "\n", + "3. Store google workspace credential json file under the 'credentials' folder\n", + "\n", + " The setup and configuration steps for Gmail API are in this guide: [Google_Workspace_API_Credential_Guide](./Google_Workspace_API_Credential_Guide.ipynb)\n", + "\n", + "The directories should be structured before launch as follows:\n", + "\n", + " ```text\n", + " The project/\n", + " │\n", + " ├── credentials/ <-- Need to create and store manually before launch; download from Google Cloud Plafotm(GCP)\n", + " │ ├── gmail_credentials.json\n", + " │ └── google_workspace_credentials.json\n", + " ├── tokens/ <-- Automatically created and saved\n", + " │ ├── gmail_tokens \n", + " │ │ └── gmail_token_{alias}.json\n", + " │ ├── google_workspace_tokens\n", + " │ └── outlook_tokens\n", + " ├── vector_index/ <-- Need to create manually before launch\n", + " │ ├── local_vector_index\n", + " │ ├── google_workspace_vector_index\n", + " │ ├── gmail_vector_index\n", + " │ └── output_vector_index\n", + " └── ***.ipynb <-- Script" + ] + }, + { + "cell_type": "markdown", + "id": "99c271af-9054-4066-9583-65a9253cb70a", + "metadata": {}, + "source": [ + "Feel free to contact me via zhufqiu@gmail.com or via [Linkedin](https://www.linkedin.com/in/zhufeng-zephyr-qiu/) if you have any questions about this project. If you have better idea about system prompt, chunk config or search_kwargs, I will be happy to talk with you!" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "bc86bec0-bda8-4e9e-9c85-423179a99981", + "metadata": {}, + "outputs": [], + "source": [ + "# !pip install pymupdf\n", + "# !pip install openpyxl\n", + "# !pip install --upgrade google-api-python-client google-auth-httplib2 google-auth-oauthlib" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "4270e52e-378c-4127-bd52-1d082e9834e0", + "metadata": {}, + "outputs": [], + "source": [ + "import os\n", + "import base64\n", + "from datetime import datetime\n", + "from email import message_from_bytes\n", + "from email.utils import parsedate_to_datetime\n", + "\n", + "from google.auth.transport.requests import Request\n", + "from google.oauth2.credentials import Credentials\n", + "from google_auth_oauthlib.flow import InstalledAppFlow\n", + "from googleapiclient.discovery import build\n", + "\n", + "from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n", + "from langchain.vectorstores import FAISS\n", + "from langchain.schema import Document\n", + "from langchain.text_splitter import CharacterTextSplitter\n", + "from langchain_chroma import Chroma\n", + "from langchain.memory import ConversationBufferMemory\n", + "from langchain.chains import ConversationalRetrievalChain\n", + "from langchain.chains import ConversationChain\n", + "from langchain.retrievers import MergerRetriever\n", + "from collections import defaultdict\n", + "from langchain.document_loaders import (\n", + " DirectoryLoader, TextLoader, \n", + " Docx2txtLoader,\n", + " TextLoader,\n", + " PyPDFLoader,\n", + " UnstructuredExcelLoader,\n", + " BSHTMLLoader\n", + ")\n", + "import glob\n", + "from dotenv import load_dotenv\n", + "import gradio as gr\n", + "import tiktoken\n", + "\n", + "from msal import PublicClientApplication\n", + "import requests\n", + "from datetime import datetime, timezone\n", + "import json\n", + "import shutil\n", + "\n", + "from PIL import Image\n", + "import pytesseract\n", + "import fitz\n", + "import ebooklib\n", + "from ebooklib import epub\n", + "import io\n", + "\n", + "from langchain.prompts.chat import (\n", + " ChatPromptTemplate,\n", + " SystemMessagePromptTemplate,\n", + " HumanMessagePromptTemplate\n", + ")\n", + "from langchain.prompts import PromptTemplate" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c3478cbe-2854-4011-b1b4-70be3f1623fd", + "metadata": {}, + "outputs": [], + "source": [ + "MODEL = \"gpt-4o-mini\"\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')" + ] + }, + { + "cell_type": "markdown", + "id": "a5195792-f6e1-43a1-9c5f-d6f8c84a253f", + "metadata": {}, + "source": [ + "### If it is your first time to create VECTOR_DIR and its sub-folder, you should create them, close this script and re-open it" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6ca9623f-fb8c-45d1-a968-370c92762924", + "metadata": {}, + "outputs": [], + "source": [ + "LOCAL_VECTOR_DIR = 'vector_index/local_vector_index'\n", + "GMAIL_VECTOR_DIR = 'vector_index/gmail_vector_index'\n", + "OUTLOOK_VECTOR_DIR = \"vector_index/outlook_vector_index\"\n", + "GOOGLE_WORKSPACE_VECTOR_DIR = 'vector_index/google_workspace_vector_index'\n", + "SLACK_VECTOR_DIR = 'vector_index/slack_vector_index'\n", + "\n", + "os.makedirs(LOCAL_VECTOR_DIR, exist_ok=True)\n", + "os.makedirs(GMAIL_VECTOR_DIR, exist_ok=True)\n", + "os.makedirs(OUTLOOK_VECTOR_DIR, exist_ok=True)\n", + "os.makedirs(GOOGLE_WORKSPACE_VECTOR_DIR, exist_ok=True)" + ] + }, + { + "cell_type": "markdown", + "id": "b0f2a2ee-c9fb-49ad-8e09-919a7a7130ea", + "metadata": {}, + "source": [ + "#### Utilize functions" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f185451f-2e2a-4ebb-a570-8b7349f3df48", + "metadata": {}, + "outputs": [], + "source": [ + "def get_num_tokens(text, model=\"text-embedding-3-large\"):\n", + " enc = tiktoken.encoding_for_model(model)\n", + " return len(enc.encode(text))\n", + "\n", + "def batch_chunks(chunks, max_tokens=250000, model=\"text-embedding-3-large\"):\n", + " batches = []\n", + " current_batch = []\n", + " current_tokens = 0\n", + "\n", + " for doc in chunks:\n", + " doc_tokens = get_num_tokens(doc.page_content, model)\n", + " if current_tokens + doc_tokens > max_tokens:\n", + " batches.append(current_batch)\n", + " current_batch = [doc]\n", + " current_tokens = doc_tokens\n", + " else:\n", + " current_batch.append(doc)\n", + " current_tokens += doc_tokens\n", + "\n", + " if current_batch:\n", + " batches.append(current_batch)\n", + " \n", + " return batches" + ] + }, + { + "cell_type": "markdown", + "id": "a5546fd7-46bf-4a36-8eef-7b4192f247e9", + "metadata": {}, + "source": [ + "### 1. Local" + ] + }, + { + "cell_type": "markdown", + "id": "937c4f19-5e5b-46b8-b15d-f7ceddd81384", + "metadata": {}, + "source": [ + "Reuse code from bluebells1 [Wk5-final-multi-doc-type-KB.ipynb](../Wk5-final-multi-doc-type-KB.ipynb). Really appreciate it!\n", + "\n", + "Advanced features:\n", + "1. ImgLoader added to load image file (png, jpg, jpeg)\n", + "2. Add logic to use DocumentLoader, extract files and show summary in Gradio textbox" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "74b85882-c2d6-42af-9079-9f2a61d9eb72", + "metadata": {}, + "outputs": [], + "source": [ + "from ebooklib import epub\n", + "from bs4 import BeautifulSoup\n", + "from langchain.document_loaders.base import BaseLoader\n", + "\n", + "class EpubLoader(BaseLoader):\n", + " def __init__(self, file_path: str):\n", + " self.file_path = file_path\n", + "\n", + " def load(self) -> list[Document]:\n", + " book = epub.read_epub(self.file_path)\n", + " text = ''\n", + " for item in book.get_items():\n", + " if item.get_type() == ebooklib.ITEM_DOCUMENT:\n", + " soup = BeautifulSoup(item.get_content().decode(\"utf-8\"), 'html.parser')\n", + " extracted = soup.get_text().strip()\n", + " if extracted:\n", + " text += extracted + '\\n\\n'\n", + "\n", + " return [Document(page_content=text.strip(), metadata={\"source\": self.file_path})]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "85f94e96-83e1-4b5a-ad63-373a37474d25", + "metadata": {}, + "outputs": [], + "source": [ + "from pptx import Presentation\n", + "\n", + "class PptxLoader(BaseLoader):\n", + " def __init__(self, file_path: str):\n", + " self.file_path = file_path\n", + "\n", + " def load(self) -> list[Document]:\n", + " prs = Presentation(self.file_path)\n", + " text = ''\n", + " for slide in prs.slides:\n", + " for shape in slide.shapes:\n", + " if hasattr(shape, \"text\") and shape.text:\n", + " text += shape.text + '\\n'\n", + "\n", + " return [Document(page_content=text, metadata={\"source\": self.file_path})]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "cd3932ce-5179-4e83-9a2c-bdefc37028aa", + "metadata": {}, + "outputs": [], + "source": [ + "from PIL import Image\n", + "import pytesseract\n", + "\n", + "class ImgLoader(BaseLoader):\n", + " def __init__(self, file_path: str):\n", + " self.file_path = file_path\n", + "\n", + " def load(self) -> list[Document]:\n", + " text = ''\n", + " try:\n", + " text = pytesseract.image_to_string(Image.open(self.file_path))\n", + " except Exception as e:\n", + " print(f\"OCR failed for {path}: {e}\")\n", + " return [Document(page_content=text, metadata={\"source\": self.file_path})]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "427e758a-77ab-4de1-ae14-8f2f233ea6db", + "metadata": {}, + "outputs": [], + "source": [ + "# Class based version of document loader which can be expanded more easily for other document types. (Currently includes file types: docx, txt (windows encoding), xlsx, pdfs, epubs, pptx)\n", + "\n", + "class DocumentLoader:\n", + " \"\"\"A clean, extensible document loader for multiple file types.\"\"\"\n", + " \n", + " def __init__(self, base_path, exclude_folders=None):\n", + " self.base_path = base_path\n", + " self.documents = []\n", + " self.exclude_folders = exclude_folders or []\n", + " self.print_info = \"\"\n", + " \n", + " # Configuration for different file types\n", + " self.loader_config = {\n", + " 'docx': {\n", + " 'loader_cls': Docx2txtLoader,\n", + " 'glob_pattern': \"**/*.docx\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'txt': {\n", + " 'loader_cls': TextLoader,\n", + " 'glob_pattern': \"**/*.txt\",\n", + " 'loader_kwargs': {\"encoding\": 'utf-8'},\n", + " 'post_process': None\n", + " },\n", + " 'md': {\n", + " 'loader_cls': TextLoader,\n", + " 'glob_pattern': \"**/*.md\",\n", + " 'loader_kwargs': {\"encoding\": 'utf-8'},\n", + " 'post_process': None\n", + " },\n", + " 'pdf': {\n", + " 'loader_cls': PyPDFLoader,\n", + " 'glob_pattern': \"**/*.pdf\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'xlsx': {\n", + " 'loader_cls': UnstructuredExcelLoader,\n", + " 'glob_pattern': \"**/*.xlsx\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'html': {\n", + " 'loader_cls': BSHTMLLoader,\n", + " 'glob_pattern': \"**/*.html\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'epub': {\n", + " 'loader_cls': EpubLoader,\n", + " 'glob_pattern': \"**/*.epub\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': self._process_epub_metadata\n", + " },\n", + " 'pptx': {\n", + " 'loader_cls': PptxLoader,\n", + " 'glob_pattern': \"**/*.pptx\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'png': {\n", + " 'loader_cls': ImgLoader,\n", + " 'glob_pattern': \"**/*.png\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'jpeg': {\n", + " 'loader_cls': ImgLoader,\n", + " 'glob_pattern': \"**/*.jpeg\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " },\n", + " 'jpg': {\n", + " 'loader_cls': ImgLoader,\n", + " 'glob_pattern': \"**/*.jpg\",\n", + " 'loader_kwargs': {},\n", + " 'post_process': None\n", + " }\n", + " }\n", + " \n", + " def _get_epub_metadata(self, file_path):\n", + " \"\"\"Extract metadata from EPUB files.\"\"\"\n", + " try:\n", + " book = epub.read_epub(file_path)\n", + " title = book.get_metadata('DC', 'title')[0][0] if book.get_metadata('DC', 'title') else None\n", + " author = book.get_metadata('DC', 'creator')[0][0] if book.get_metadata('DC', 'creator') else None\n", + " return title, author\n", + " except Exception as e:\n", + " self.print_info += f\"Error extracting EPUB metadata: {e}\\n\"\n", + " return None, None\n", + " \n", + " def _process_epub_metadata(self, doc) -> None:\n", + " \"\"\"Post-process EPUB documents to add metadata.\"\"\"\n", + " title, author = self._get_epub_metadata(doc.metadata['source'])\n", + " doc.metadata[\"author\"] = author\n", + " doc.metadata[\"title\"] = title\n", + " \n", + " def _load_file_type(self, folder, file_type, config):\n", + " \"\"\"Load documents of a specific file type from a folder.\"\"\"\n", + " try:\n", + " loader = DirectoryLoader(\n", + " folder, \n", + " glob=config['glob_pattern'], \n", + " loader_cls=config['loader_cls'],\n", + " loader_kwargs=config['loader_kwargs']\n", + " )\n", + " docs = loader.load()\n", + " self.print_info += f\"Found {len(docs)} .{file_type} files\\n\"\n", + " \n", + " # Apply post-processing if defined\n", + " if config['post_process']:\n", + " for doc in docs:\n", + " config['post_process'](doc)\n", + " \n", + " return docs\n", + " \n", + " except Exception as e:\n", + " self.print_info += f\"Error loading .{file_type} files: {e}\\n\"\n", + " return []\n", + " \n", + " def load_all(self):\n", + " \"\"\"Load all documents from configured folders.\"\"\"\n", + " all_folders = [f for f in glob.glob(self.base_path) if os.path.isdir(f)]\n", + "\n", + " #filter out excluded folders\n", + " folders = []\n", + " for folder in all_folders:\n", + " folder_name = os.path.basename(folder)\n", + " if folder_name not in self.exclude_folders:\n", + " folders.append(folder)\n", + " else:\n", + " self.print_info += f\"Excluded folder: {folder_name}\\n\"\n", + " \n", + " self.print_info += f\"Scanning folders (directories only):{folders}\\n\" \n", + " \n", + " self.documents = []\n", + " \n", + " for folder in folders:\n", + " doc_type = os.path.basename(folder)\n", + " self.print_info += f\"\\nProcessing folder: {doc_type}\\n\"\n", + " \n", + " for file_type, config in self.loader_config.items():\n", + " docs = self._load_file_type(folder, file_type, config)\n", + " \n", + " # Add doc_type metadata to all documents\n", + " for doc in docs:\n", + " doc.metadata[\"doc_type\"] = doc_type\n", + " self.documents.append(doc)\n", + " \n", + " self.print_info += f\"\\nTotal documents loaded: {len(self.documents)}\\n\"\n", + " return self.documents\n", + " \n", + " def add_file_type(self, extension, loader_cls, glob_pattern=None, \n", + " loader_kwargs=None, post_process=None):\n", + " \"\"\"Add support for a new file type.\"\"\"\n", + " self.loader_config[extension] = {\n", + " 'loader_cls': loader_cls,\n", + " 'glob_pattern': glob_pattern or f\"**/*.{extension}\",\n", + " 'loader_kwargs': loader_kwargs or {},\n", + " 'post_process': post_process\n", + " }\n", + "\n", + "# load\n", + "# loader = DocumentLoader(\"local-knowledge-base/**\", exclude_folders=[\"Music\", \"Online Courses\", \"Fitness\"])\n", + "# documents = loader.load_all()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "53e65a63-29fd-4db3-91f0-246cc2b61941", + "metadata": {}, + "outputs": [], + "source": [ + "def local_embed_and_store(docs):\n", + " text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + " chunks = [doc for doc in text_splitter.split_documents(docs) if doc.page_content.strip()]\n", + "\n", + " if not chunks:\n", + " return \"⚠️ No non-empty chunks to embed. Skipping vectorstore update.\"\n", + "\n", + " embeddings = OpenAIEmbeddings()\n", + "\n", + " vectorstore = None\n", + " if os.path.exists(LOCAL_VECTOR_DIR):\n", + " vectorstore = Chroma(persist_directory=LOCAL_VECTOR_DIR, embedding_function=embeddings)\n", + " else:\n", + " if chunks:\n", + " vectorstore = Chroma.from_documents(documents=chunks[:1], embedding=embeddings, persist_directory=LOCAL_VECTOR_DIR)\n", + " chunks = chunks[1:]\n", + " else:\n", + " return \"⚠️ No chunks to create new vectorstore.\"\n", + " \n", + " batches = batch_chunks(chunks)\n", + " total = 1 if not os.path.exists(LOCAL_VECTOR_DIR) else 0\n", + " \n", + " for batch in batches:\n", + " vectorstore.add_documents(batch)\n", + " total += len(batch)\n", + "\n", + " info = \"\"\n", + " info += f\"Vectorstore updated with {total} new chunks.\\n\"\n", + " num_docs = vectorstore._collection.count()\n", + " info += f\"Vectorstore contains {num_docs} chunks.\\n\"\n", + " return info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d0a70a0e-08cd-4827-b42b-9a5394ff6dec", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_local_folder(folder_path=\"local-knowledge-base\", exclude=\"\"):\n", + "\n", + " # try:\n", + " info = f\"Process files under: {folder_path}\\n\"\n", + " loader = DocumentLoader(os.path.join(folder_path, \"**\"), exclude_folders=[folder.strip() for folder in exclude.split(',')])\n", + " docs = loader.load_all()\n", + " info += loader.print_info\n", + " if not docs:\n", + " return info + \"No valid files found in the given range.\"\n", + " info += f\"Fetched {len(docs)} files.\\n\"\n", + " info += local_embed_and_store(docs)\n", + " return info\n", + "\n", + " # except Exception as e:\n", + " # return f\"❌ Extraction failed: {str(e)}\"" + ] + }, + { + "cell_type": "markdown", + "id": "0e47d670-8c50-4744-8fbd-78112fa941dd", + "metadata": {}, + "source": [ + "### 2. Gmail" + ] + }, + { + "cell_type": "markdown", + "id": "4d52fe40-65e3-4d82-9999-1ed3e4cbae0a", + "metadata": {}, + "source": [ + "#### Store gmail credential json file under the credentials folder\n", + "\n", + "To avoid complicated steps and focus on LLMs stuff, I chose to utilize the Gmail API in test mode.\n", + "\n", + "I have included the setup and configuration steps in this guide:\n", + "[Gmail_API_Credential_Guide](./Gmail_API_Credential_Guide.ipynb)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3f32c4e4-fa7a-42a1-9ef8-b981af02f585", + "metadata": {}, + "outputs": [], + "source": [ + "GMAIL_SCOPES = ['https://www.googleapis.com/auth/gmail.readonly']\n", + "GMAIL_CREDENTIALS_FILE = 'credentials/gmail_credentials.json'\n", + "GMAIL_TOKEN_DIR = 'tokens/gmail_tokens'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "db344254-8c92-4e82-8414-40b3bef56db5", + "metadata": {}, + "outputs": [], + "source": [ + "def gmail_get_credentials(account_alias):\n", + " token_path = os.path.join(GMAIL_TOKEN_DIR, f'gmail_token_{account_alias}.json')\n", + " creds = None\n", + " if os.path.exists(token_path):\n", + " creds = Credentials.from_authorized_user_file(token_path, GMAIL_SCOPES)\n", + " if not creds or not creds.valid:\n", + " if creds and creds.expired and creds.refresh_token:\n", + " creds.refresh(Request())\n", + " else:\n", + " flow = InstalledAppFlow.from_client_secrets_file(GMAIL_CREDENTIALS_FILE, GMAIL_SCOPES)\n", + " creds = flow.run_local_server(port=0)\n", + " with open(token_path, 'w') as token_file:\n", + " token_file.write(creds.to_json())\n", + " return creds" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "119558f0-4d35-4737-ad8a-eef516b540d2", + "metadata": {}, + "outputs": [], + "source": [ + "def parse_message(service, msg_id):\n", + " msg = service.users().messages().get(userId='me', id=msg_id, format='raw').execute()\n", + " raw_msg = base64.urlsafe_b64decode(msg['raw'].encode('ASCII'))\n", + " email_message = message_from_bytes(raw_msg)\n", + " subject = email_message['Subject'] or \"(No Subject)\"\n", + " date = parsedate_to_datetime(email_message['Date'])\n", + " sender = email_message['From'] or \"\"\n", + " to = email_message['To'] or \"\"\n", + " cc = email_message['Cc'] or \"\"\n", + " body = \"\"\n", + " \n", + " for part in email_message.walk():\n", + " if part.get_content_type() == 'text/plain' and not part.get('Content-Disposition'):\n", + " body = part.get_payload(decode=True).decode('utf-8', errors='ignore')\n", + " break\n", + "\n", + " content = f\"\"\"Subject: {subject}\n", + " From: {sender}\n", + " To: {to}\n", + " Cc: {cc}\n", + " {body}\n", + " \"\"\"\n", + " return {\n", + " \"id\": msg_id,\n", + " \"subject\": subject,\n", + " \"date\": date,\n", + " \"body\": content\n", + " }" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "481d0500-6270-47ec-bc30-44400c86dff2", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_emails(service, start_date, end_date):\n", + " query = (\n", + " f\"(category:primary OR is:important OR is:starred OR is:snoozed OR is:sent OR in:chats OR label:SCHEDULED) \"\n", + " f\"after:{start_date} before:{end_date} -in:spam -in:trash -category:promotions -category:forums\"\n", + " ) \n", + " \n", + " all_messages = []\n", + " page_token = None\n", + "\n", + " while True:\n", + " response = service.users().messages().list(userId='me', q=query, pageToken=page_token).execute()\n", + " messages = response.get('messages', [])\n", + " print(f\"Found {len(messages)} sub-messages.\")\n", + " all_messages.extend(messages)\n", + " page_token = response.get('nextPageToken')\n", + " if not page_token:\n", + " break\n", + " print(f\"Total messages fetched: {len(all_messages)}\")\n", + " parsed_emails = []\n", + " for msg in all_messages:\n", + " parsed = parse_message(service, msg['id'])\n", + " if parsed:\n", + " parsed_emails.append(parsed)\n", + " \n", + " return parsed_emails\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aebb1598-e95d-4a7b-9d40-44afb62f587d", + "metadata": {}, + "outputs": [], + "source": [ + "def gmail_embed_and_store(emails, account):\n", + " docs = []\n", + " for email in emails:\n", + " content = f\"Subject: {email['subject']}\\n\\n{email['body']}\"\n", + " doc = Document(\n", + " page_content=content.strip(),\n", + " metadata={\n", + " \"date\": str(email['date']),\n", + " \"gmail_id\": email['id'],\n", + " \"account\": account\n", + " }\n", + " )\n", + " docs.append(doc)\n", + "\n", + " text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + " chunks = [doc for doc in text_splitter.split_documents(docs) if doc.page_content.strip()]\n", + "\n", + " if not chunks:\n", + " return \"⚠️ No non-empty chunks to embed. Skipping vectorstore update.\"\n", + "\n", + " embeddings = OpenAIEmbeddings()\n", + "\n", + " vectorstore = None\n", + " if os.path.exists(GMAIL_VECTOR_DIR):\n", + " vectorstore = Chroma(persist_directory=GMAIL_VECTOR_DIR, embedding_function=embeddings)\n", + " else:\n", + " if chunks:\n", + " vectorstore = Chroma.from_documents(documents=chunks[:1], embedding=embeddings, persist_directory=GMAIL_VECTOR_DIR)\n", + " chunks = chunks[1:]\n", + " else:\n", + " return \"⚠️ No chunks to create new vectorstore.\"\n", + " \n", + " batches = batch_chunks(chunks)\n", + " total = 1 if not os.path.exists(GMAIL_VECTOR_DIR) else 0\n", + " \n", + " for batch in batches:\n", + " vectorstore.add_documents(batch)\n", + " total += len(batch)\n", + "\n", + " info = \"\"\n", + " info += f\"Vectorstore updated with {total} new chunks from {account}.\\n\"\n", + " num_docs = vectorstore._collection.count()\n", + " info += f\"Vectorstore contains {num_docs} chunks.\\n\"\n", + " return info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "a3d67265-ef62-4104-ae79-783e6d20d31c", + "metadata": {}, + "outputs": [], + "source": [ + "def login_gmail(alias):\n", + " try:\n", + " creds = gmail_get_credentials(alias)\n", + " service = build('gmail', 'v1', credentials=creds)\n", + " profile = service.users().getProfile(userId='me').execute()\n", + " email = profile.get(\"emailAddress\")\n", + "\n", + " # Store in session\n", + " SESSION_STATE[\"gmail_service\"] = service\n", + " SESSION_STATE[\"gmail_email\"] = email\n", + " SESSION_STATE[\"gmail_alias\"] = alias\n", + "\n", + " return f\"✅ Logged in as: {email}\"\n", + " except Exception as e:\n", + " return f\"❌ Login failed: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "69cb6320-7ef0-49bb-8893-d51d6d2cd87c", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_gmail(start_date, end_date):\n", + " service = SESSION_STATE.get(\"gmail_service\")\n", + " email_address = SESSION_STATE.get(\"gmail_email\")\n", + "\n", + " if not service:\n", + " return \"❌ Please login first.\"\n", + "\n", + " # try:\n", + " info = f\"Connected to: {email_address}\\n\"\n", + " emails = fetch_emails(service, start_date, end_date)\n", + "\n", + " if not emails:\n", + " return info + \"No emails found in the given range.\"\n", + " info += f\"Fetched {len(emails)} emails.\\n\"\n", + " info += gmail_embed_and_store(emails, account=email_address)\n", + " return info\n", + "\n", + " # except Exception as e:\n", + " # return f\"❌ Extraction failed: {str(e)}\"" + ] + }, + { + "cell_type": "markdown", + "id": "b049fee6-5b51-4458-b089-6a11c6050492", + "metadata": {}, + "source": [ + "### 3. Outlook" + ] + }, + { + "cell_type": "markdown", + "id": "7660ec50-23ca-476f-97f7-42b764de46fa", + "metadata": {}, + "source": [ + "#### Set AZURE_CLIENT_ID in .env file\n", + "\n", + "I have included the setup and configuration steps in this guide:\n", + "[Outlook_API_Credential_Guide](./Outlook_API_Credential_Guide.ipynb)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f1f2b0d2-d2c0-414f-be53-c3bc74ceb6a6", + "metadata": {}, + "outputs": [], + "source": [ + "load_dotenv()\n", + "\n", + "OUTLOOK_TOKEN_DIR = \"tokens/outlook_tokens\"\n", + "OUTLOOK_CLIENT_ID = os.getenv(\"AZURE_CLIENT_ID\")\n", + "OUTLOOK_AUTHORITY = \"https://login.microsoftonline.com/common\" \n", + "OUTLOOK_SCOPES = [\"Mail.Read\", \"User.Read\"]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2197700b-1103-4ba0-b929-28fea4af6881", + "metadata": {}, + "outputs": [], + "source": [ + "def fetch_outlook_emails(access_token, start_date, end_date):\n", + " headers = {\n", + " \"Authorization\": f\"Bearer {access_token}\",\n", + " \"Prefer\": \"outlook.body-content-type='text'\"\n", + " }\n", + "\n", + " # Filter format: yyyy-mm-ddTHH:MM:SSZ\n", + " query = (\n", + " \"https://graph.microsoft.com/v1.0/me/messages\"\n", + " f\"?$top=100\"\n", + " \"&$select=id,subject,receivedDateTime,body,sender,toRecipients,ccRecipients\"\n", + " )\n", + "\n", + " all_emails = []\n", + "\n", + " while query:\n", + " response = requests.get(query, headers=headers)\n", + " if not response.ok:\n", + " print(f\"❌ HTTP {response.status_code}: {response.text}\")\n", + " break\n", + "\n", + " res = response.json()\n", + " for msg in res.get(\"value\", []):\n", + " received = msg.get(\"receivedDateTime\", \"\")\n", + " try:\n", + " received_dt = datetime.fromisoformat(received.replace(\"Z\", \"+00:00\"))\n", + " except Exception:\n", + " continue\n", + "\n", + " if not (start_date <= received_dt <= end_date):\n", + " continue\n", + "\n", + " email_data = {\n", + " \"id\": msg.get(\"id\"),\n", + " \"subject\": msg.get(\"subject\", \"\"),\n", + " \"body\": msg.get(\"body\", {}).get(\"content\", \"\"),\n", + " \"sender\": msg.get(\"sender\", {}).get(\"emailAddress\", {}).get(\"address\", \"\"),\n", + " \"to\": [r[\"emailAddress\"][\"address\"] for r in msg.get(\"toRecipients\", [])],\n", + " \"cc\": [r[\"emailAddress\"][\"address\"] for r in msg.get(\"ccRecipients\", [])],\n", + " \"date\": received_dt.isoformat()\n", + " }\n", + "\n", + " all_emails.append(email_data)\n", + "\n", + " query = res.get(\"@odata.nextLink\")\n", + "\n", + " print(f\"✅ Total emails extracted: {len(all_emails)}\")\n", + " return all_emails" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5d9759ad-47fa-4f3a-8e67-59b59ccccfd9", + "metadata": {}, + "outputs": [], + "source": [ + "def outlook_embed_and_store(emails):\n", + " if not emails:\n", + " return \"No emails to embed.\\n\"\n", + "\n", + " docs = []\n", + " for email in emails:\n", + " content = (\n", + " f\"Subject: {email['subject']}\\n\"\n", + " f\"From: {email['sender']}\\n\"\n", + " f\"To: {', '.join(email['to'])}\\n\"\n", + " f\"CC: {', '.join(email['cc'])}\\n\\n\"\n", + " f\"{email['body']}\"\n", + " )\n", + " doc = Document(\n", + " page_content=content,\n", + " metadata={\"date\": email[\"date\"], \"outlook_id\": email[\"id\"]}\n", + " )\n", + " docs.append(doc)\n", + "\n", + " text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + " chunks = [doc for doc in text_splitter.split_documents(docs) if doc.page_content.strip()]\n", + "\n", + " if not chunks:\n", + " return \"⚠️ No non-empty chunks to embed. Skipping vectorstore update.\"\n", + "\n", + " embeddings = OpenAIEmbeddings()\n", + "\n", + " vectorstore = None\n", + " if os.path.exists(OUTLOOK_VECTOR_DIR):\n", + " vectorstore = Chroma(persist_directory=OUTLOOK_VECTOR_DIR, embedding_function=embeddings)\n", + " else:\n", + " if chunks:\n", + " vectorstore = Chroma.from_documents(documents=chunks[:1], embedding=embeddings, persist_directory=OUTLOOK_VECTOR_DIR)\n", + " chunks = chunks[1:]\n", + " else:\n", + " return \"⚠️ No chunks to create new vectorstore.\\n\"\n", + " \n", + " batches = batch_chunks(chunks)\n", + " total = 1 if not os.path.exists(OUTLOOK_VECTOR_DIR) else 0\n", + " \n", + " for batch in batches:\n", + " vectorstore.add_documents(batch)\n", + " total += len(batch)\n", + "\n", + " info = \"\"\n", + " info += f\"✅ Vectorstore updated with {total} chunks.\\n\"\n", + " num_docs = vectorstore._collection.count()\n", + " info += f\"Vectorstore contains {num_docs} chunks.\\n\"\n", + " return info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1feb49b5-9df0-4232-a233-c6ceb97361a7", + "metadata": {}, + "outputs": [], + "source": [ + "def login_outlook(alias):\n", + " # try:\n", + " token_path = os.path.join(OUTLOOK_TOKEN_DIR, f\"outlook_token_{alias}.json\")\n", + " SESSION_STATE[\"outlook_alias\"] = alias\n", + " access_token = None\n", + "\n", + " # Load existing token\n", + " if os.path.exists(token_path):\n", + " with open(token_path, \"r\") as f:\n", + " result = json.load(f)\n", + " access_token = result.get(\"access_token\")\n", + "\n", + " # If no token, run device flow\n", + " if not access_token:\n", + " app = PublicClientApplication(OUTLOOK_CLIENT_ID, authority=OUTLOOK_AUTHORITY)\n", + " flow = app.initiate_device_flow(scopes=OUTLOOK_SCOPES)\n", + "\n", + " if \"user_code\" not in flow:\n", + " return \"❌ Failed to initiate device login.\"\n", + "\n", + " print(\"🔗 Visit:\", flow[\"verification_uri\"])\n", + " print(\"🔐 Enter code:\", flow[\"user_code\"])\n", + "\n", + " result = app.acquire_token_by_device_flow(flow)\n", + "\n", + " if \"access_token\" not in result:\n", + " return f\"❌ Login failed: {result.get('error_description', 'Unknown error')}\"\n", + "\n", + " access_token = result[\"access_token\"]\n", + "\n", + " with open(token_path, \"w\") as f:\n", + " json.dump(result, f)\n", + "\n", + " # Get user's email via Microsoft Graph\n", + " headers = {\"Authorization\": f\"Bearer {access_token}\"}\n", + " user_info = requests.get(\"https://graph.microsoft.com/v1.0/me\", headers=headers).json()\n", + " email = user_info.get(\"mail\") or user_info.get(\"userPrincipalName\")\n", + "\n", + " # Store in session\n", + " SESSION_STATE[\"outlook_token\"] = access_token\n", + " SESSION_STATE[\"outlook_email\"] = email\n", + "\n", + " return f\"✅ Logged in to Outlook as: {email}\"\n", + "\n", + " # except Exception as e:\n", + " # return f\"❌ Login failed: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9e11523b-a757-459c-8c4c-1ceef586439f", + "metadata": {}, + "outputs": [], + "source": [ + "def start_outlook_login(alias):\n", + " token_path = os.path.join(OUTLOOK_TOKEN_DIR, f\"outlook_token_{alias}.json\")\n", + " access_token = None\n", + " SESSION_STATE[\"outlook_token_path\"] = token_path\n", + " \n", + " # Load existing token\n", + " if os.path.exists(token_path):\n", + " return True, \"This alias already verified\"\n", + "\n", + " # If no token, run device flow\n", + " if not access_token:\n", + " app = PublicClientApplication(OUTLOOK_CLIENT_ID, authority=OUTLOOK_AUTHORITY)\n", + " flow = app.initiate_device_flow(scopes=OUTLOOK_SCOPES)\n", + "\n", + " if \"user_code\" not in flow:\n", + " return False, \"❌ Failed to initiate device login.\"\n", + "\n", + " # Store the flow for next step\n", + " SESSION_STATE[\"outlook_alias\"] = alias\n", + " SESSION_STATE[\"outlook_app\"] = app\n", + " SESSION_STATE[\"outlook_flow\"] = flow\n", + " \n", + " msg = f\"🔗 Visit: {flow['verification_uri']}\\n🔐 Enter code: {flow['user_code']}\"\n", + " return False, \"🔄 Waiting for verification...\\n\" + msg\n", + "\n", + "def finish_outlook_login():\n", + " flag = SESSION_STATE.get(\"outlook_login_flag\")\n", + " token_path = SESSION_STATE.get(\"outlook_token_path\")\n", + " if flag:\n", + " with open(token_path, \"r\") as f:\n", + " result = json.load(f)\n", + " access_token = result.get(\"access_token\")\n", + " else: \n", + " app = SESSION_STATE.get(\"outlook_app\")\n", + " flow = SESSION_STATE.get(\"outlook_flow\")\n", + " \n", + " result = app.acquire_token_by_device_flow(flow)\n", + " \n", + " if \"access_token\" not in result:\n", + " return f\"❌ Login failed: {result.get('error_description', 'Unknown error')}\"\n", + " \n", + " access_token = result[\"access_token\"]\n", + " \n", + " with open(token_path, \"w\") as f:\n", + " json.dump(result, f)\n", + " \n", + "\n", + " # Get user's email via Microsoft Graph\n", + " headers = {\"Authorization\": f\"Bearer {access_token}\"}\n", + " user_info = requests.get(\"https://graph.microsoft.com/v1.0/me\", headers=headers).json()\n", + " email = user_info.get(\"mail\") or user_info.get(\"userPrincipalName\")\n", + "\n", + " # Store in session\n", + " SESSION_STATE[\"outlook_token\"] = access_token\n", + " SESSION_STATE[\"outlook_email\"] = email\n", + "\n", + " return f\"✅ Logged in to Outlook as: {email}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d3faea0d-723d-41e3-9683-db92dd918aba", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_outlook_emails(start, end, alias):\n", + " try:\n", + " start_date = datetime.strptime(start.strip(), \"%Y/%m/%d\").replace(tzinfo=timezone.utc)\n", + " end_date = datetime.strptime(end.strip(), \"%Y/%m/%d\").replace(tzinfo=timezone.utc)\n", + " except ValueError:\n", + " return \"❌ Invalid date format. Use YYYY/MM/DD.\"\n", + "\n", + " access_token = SESSION_STATE[\"outlook_token\"]\n", + "\n", + " if not access_token:\n", + " return f\"❌ No access token found for '{alias}'. Please login first.\"\n", + "\n", + " info = \"\"\n", + " try:\n", + " emails = fetch_outlook_emails(access_token, start_date, end_date)\n", + " if not emails:\n", + " return f\"❌ No email found.\"\n", + " info += f\"✅ Extracted and embedded {len(emails)} Outlook emails.\\n\"\n", + " info += outlook_embed_and_store(emails)\n", + " return info\n", + " except Exception as e:\n", + " return f\"❌ Error: {str(e)}\"\n" + ] + }, + { + "cell_type": "markdown", + "id": "0c030701-8f16-4101-a501-f310ce61871c", + "metadata": {}, + "source": [ + "### 4. Google Workspace" + ] + }, + { + "cell_type": "markdown", + "id": "4b04baa3-0dfe-491a-974e-c1b97c978031", + "metadata": {}, + "source": [ + "#### Store google workspace credential json file under the credentials folder\n", + "\n", + "To avoid complicated steps and focus on LLMs stuff, I chose to utilize the Google Drive/Workspace API in test mode.\n", + "\n", + "I have included the setup and configuration steps in this guide:\n", + "[Google_Workspace_API_Credential_Guide](./Google_Workspace_API_Credential_Guide.ipynb)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1aeb8a99-d039-4550-8cac-f9370e7d7401", + "metadata": {}, + "outputs": [], + "source": [ + "GOOGLE_WORKSPACE_SCOPES = [\n", + " 'https://www.googleapis.com/auth/gmail.readonly',\n", + " 'https://www.googleapis.com/auth/drive.readonly',\n", + " 'https://www.googleapis.com/auth/documents.readonly',\n", + " 'https://www.googleapis.com/auth/spreadsheets.readonly',\n", + " 'https://www.googleapis.com/auth/presentations.readonly'\n", + "]\n", + "GOOGLE_WORKSPACE_CREDENTIALS_FILE = 'credentials/google_drive_workspace_credentials.json'\n", + "GOOGLE_WORKSPACE_TOKEN_DIR = 'tokens/google_workspace_tokens'" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0d7c1ad3-d288-42a7-bc7b-4ddae0f3aaa3", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_google_doc(docs_service, file_id):\n", + " doc = docs_service.documents().get(documentId=file_id).execute()\n", + " content = \"\"\n", + " for elem in doc.get(\"body\", {}).get(\"content\", []):\n", + " if \"paragraph\" in elem:\n", + " for run in elem[\"paragraph\"][\"elements\"]:\n", + " content += run.get(\"textRun\", {}).get(\"content\", \"\")\n", + " return content.strip()\n", + "\n", + "def extract_google_sheet(service, file_id):\n", + " # Get spreadsheet metadata\n", + " spreadsheet = service.spreadsheets().get(spreadsheetId=file_id).execute()\n", + " all_text = \"\"\n", + "\n", + " # Loop through each sheet\n", + " for sheet in spreadsheet.get(\"sheets\", []):\n", + " title = sheet[\"properties\"][\"title\"]\n", + " result = service.spreadsheets().values().get(\n", + " spreadsheetId=file_id,\n", + " range=title\n", + " ).execute()\n", + "\n", + " values = result.get(\"values\", [])\n", + " sheet_text = f\"### Sheet: {title} ###\\n\"\n", + " sheet_text += \"\\n\".join([\", \".join(row) for row in values])\n", + " all_text += sheet_text + \"\\n\\n\"\n", + "\n", + " return all_text.strip()\n", + "\n", + "\n", + "def extract_google_slide(slides_service, file_id):\n", + " pres = slides_service.presentations().get(presentationId=file_id).execute()\n", + " text = \"\"\n", + " for slide in pres.get(\"slides\", []):\n", + " for element in slide.get(\"pageElements\", []):\n", + " shape = element.get(\"shape\")\n", + " if shape:\n", + " for p in shape.get(\"text\", {}).get(\"textElements\", []):\n", + " if \"textRun\" in p:\n", + " text += p[\"textRun\"][\"content\"]\n", + " return text.strip()\n", + "\n", + "def extract_pdf_from_drive(drive_service, file_id):\n", + " request = drive_service.files().get_media(fileId=file_id)\n", + " fh = io.BytesIO()\n", + " downloader = MediaIoBaseDownload(fh, request)\n", + " done = False\n", + " while not done:\n", + " _, done = downloader.next_chunk()\n", + " fh.seek(0)\n", + " reader = PdfReader(fh)\n", + " return \"\\n\".join([page.extract_text() for page in reader.pages if page.extract_text()])" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "005640a5-b7b3-4397-be1c-d86dccafdc61", + "metadata": {}, + "outputs": [], + "source": [ + "def login_google_workspace(alias):\n", + " try:\n", + " creds = google_workspace_get_creds(alias)\n", + " service = build('gmail', 'v1', credentials=creds)\n", + " profile = service.users().getProfile(userId='me').execute()\n", + " email = profile.get(\"emailAddress\")\n", + "\n", + " drive_service = build(\"drive\", \"v3\", credentials=creds)\n", + " docs_service = build('docs', 'v1', credentials=creds)\n", + " sheets_service = build('sheets', 'v4', credentials=creds)\n", + " slides_service = build('slides', 'v1', credentials=creds)\n", + "\n", + " # Store in session\n", + " SESSION_STATE[\"google_workspace_drive_service\"] = drive_service\n", + " SESSION_STATE[\"google_workspace_docs_service\"] = docs_service\n", + " SESSION_STATE[\"google_workspace_sheets_service\"] = sheets_service\n", + " SESSION_STATE[\"google_workspace_slides_service\"] = slides_service\n", + " SESSION_STATE[\"google_workspace_email\"] = email\n", + " SESSION_STATE[\"google_workspace_alias\"] = alias\n", + "\n", + " return f\"✅ Logged in as: {email}\"\n", + " except Exception as e:\n", + " return f\"❌ Login failed: {str(e)}\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2677d0aa-d61d-45e2-994a-b4707d839b48", + "metadata": {}, + "outputs": [], + "source": [ + "def google_workspace_get_creds(account_alias):\n", + " token_path = os.path.join(GOOGLE_WORKSPACE_TOKEN_DIR, f'google_workspace_token_{account_alias}.json')\n", + " \n", + " if os.path.exists(token_path):\n", + " creds = Credentials.from_authorized_user_file(token_path, GOOGLE_WORKSPACE_SCOPES)\n", + " else:\n", + " flow = InstalledAppFlow.from_client_secrets_file(GOOGLE_WORKSPACE_CREDENTIALS_FILE, GOOGLE_WORKSPACE_SCOPES)\n", + " creds = flow.run_local_server(port=0)\n", + " with open(\"token.json\", \"w\") as token:\n", + " token.write(creds.to_json())\n", + " return creds\n", + " \n", + "\n", + "def get_folder_id_by_name(drive_service, folder_name):\n", + " query = f\"mimeType='application/vnd.google-apps.folder' and name='{folder_name}' and trashed=false\"\n", + " results = drive_service.files().list(\n", + " q=query,\n", + " fields=\"files(id, name)\",\n", + " pageSize=1\n", + " ).execute()\n", + "\n", + " folders = results.get(\"files\", [])\n", + " if not folders:\n", + " raise ValueError(f\"❌ Folder named '{folder_name}' not found.\")\n", + " return folders[0]['id']\n", + "\n", + "\n", + "def extract_docs_from_google_workspace(folder_name):\n", + " info = \"\"\n", + "\n", + " file_types = {\n", + " 'application/vnd.google-apps.document': lambda fid: extract_google_doc(docs_service, fid),\n", + " 'application/vnd.google-apps.spreadsheet': lambda fid: extract_google_sheet(sheets_service, fid),\n", + " 'application/vnd.google-apps.presentation': lambda fid: extract_google_slide(slides_service, fid),\n", + " 'application/pdf': lambda fid: extract_pdf_from_drive(drive_service, fid),\n", + " }\n", + "\n", + " drive_service = SESSION_STATE.get(\"google_workspace_drive_service\")\n", + " docs_service = SESSION_STATE.get(\"google_workspace_docs_service\")\n", + " sheets_service = SESSION_STATE.get(\"google_workspace_sheets_service\")\n", + " slides_service = SESSION_STATE.get(\"google_workspace_slides_service\")\n", + " \n", + " if not drive_service or not docs_service or not sheets_service or not slides_service: \n", + " return None, \"Please login first.\\n\"\n", + " \n", + "\n", + " folder_id = get_folder_id_by_name(drive_service, folder_name)\n", + " print(\"folder_id\")\n", + " print(folder_id)\n", + " info += f\"Collection files from folder: {folder_name}\\n\"\n", + " \n", + " query = (\n", + " f\"'{folder_id}' in parents and (\"\n", + " 'mimeType=\"application/vnd.google-apps.document\" or '\n", + " 'mimeType=\"application/vnd.google-apps.spreadsheet\" or '\n", + " 'mimeType=\"application/vnd.google-apps.presentation\" or '\n", + " 'mimeType=\"application/pdf\")'\n", + " )\n", + " \n", + " results = drive_service.files().list(\n", + " q=query,\n", + " fields=\"files(id, name, mimeType)\",\n", + " pageSize=20\n", + " ).execute()\n", + "\n", + " docs = []\n", + " summary_info = {\n", + " 'application/vnd.google-apps.document': {'file_type': 'Google Doc', 'count': 0},\n", + " 'application/vnd.google-apps.spreadsheet': {'file_type': 'Google Sheet', 'count': 0},\n", + " 'application/vnd.google-apps.presentation': {'file_type': 'Google Silde', 'count': 0},\n", + " 'application/pdf': {'file_type': 'PDF', 'count': 0}\n", + " }\n", + " for file in results.get(\"files\", []):\n", + " print(file['mimeType'])\n", + " extractor = file_types.get(file['mimeType'])\n", + " if extractor:\n", + " try:\n", + " content = extractor(file[\"id\"])\n", + " if content:\n", + " docs.append(Document(page_content=content, metadata={\"source\": file[\"name\"]}))\n", + " summary_info[file['mimeType']]['count'] += 1\n", + " print(file['mimeType'])\n", + " print(summary_info[file['mimeType']]['count'])\n", + " except Exception as e:\n", + " print(f\"❌ Error processing {file['name']}: {e}\")\n", + " \n", + " total = 0;\n", + " for file_type, element in summary_info.items():\n", + " total += element['count']\n", + " info += f\"Found {element['count']} {element['file_type']} files\\n\"\n", + " info += f\"Total documents loaded: {total}\\n\"\n", + " return docs, info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "f5e7aee3-a7fe-4dd1-ada4-e7290cb1d1c4", + "metadata": {}, + "outputs": [], + "source": [ + "def google_workspace_embed_and_store(docs):\n", + " text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + " chunks = [doc for doc in text_splitter.split_documents(docs) if doc.page_content.strip()]\n", + "\n", + " if not chunks:\n", + " return \"⚠️ No non-empty chunks to embed. Skipping vectorstore update.\"\n", + "\n", + " embeddings = OpenAIEmbeddings()\n", + "\n", + " vectorstore = None\n", + " if os.path.exists(GOOGLE_WORKSPACE_VECTOR_DIR):\n", + " vectorstore = Chroma(persist_directory=GOOGLE_WORKSPACE_VECTOR_DIR, embedding_function=embeddings)\n", + " else:\n", + " if chunks:\n", + " vectorstore = Chroma.from_documents(documents=chunks[:1], embedding=embeddings, persist_directory=GOOGLE_WORKSPACE_VECTOR_DIR)\n", + " chunks = chunks[1:]\n", + " else:\n", + " return \"⚠️ No chunks to create new vectorstore.\"\n", + " \n", + " batches = batch_chunks(chunks)\n", + " total = 1 if not os.path.exists(GOOGLE_WORKSPACE_VECTOR_DIR) else 0\n", + " \n", + " for batch in batches:\n", + " vectorstore.add_documents(batch)\n", + " total += len(batch)\n", + "\n", + " info = \"\"\n", + " info += f\"Vectorstore updated with {total} new chunks.\\n\"\n", + " num_docs = vectorstore._collection.count()\n", + " info += f\"Vectorstore contains {num_docs} chunks.\\n\"\n", + " return info" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "9fd47dcc-03be-4ff2-8e13-406067242c0d", + "metadata": {}, + "outputs": [], + "source": [ + "def extract_google_workspace_folder(folder_path):\n", + "\n", + " # try:\n", + " info = f\"Process files under: {folder_path}\\n\"\n", + " docs, embed_store_info = extract_docs_from_google_workspace(folder_path)\n", + " info += embed_store_info\n", + " if not docs:\n", + " return info + \"No valid files found in the given range.\"\n", + " info += f\"Fetched {len(docs)} files.\\n\"\n", + " info += google_workspace_embed_and_store(docs)\n", + " return info\n", + "\n", + " # except Exception as e:\n", + " # return f\"❌ Extraction failed: {str(e)}\"" + ] + }, + { + "cell_type": "markdown", + "id": "59794946-dfdd-40b7-909d-f8290d628242", + "metadata": {}, + "source": [ + "### 5. Slack" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "33c6bf19-f685-4654-9fda-06ec32afd2e5", + "metadata": {}, + "outputs": [], + "source": [] + }, + { + "cell_type": "markdown", + "id": "9de15f01-7749-46df-9526-306c51310797", + "metadata": {}, + "source": [ + "### 6. Gradio UI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d613a92f-e16b-4cc0-a454-7fbae162f27b", + "metadata": {}, + "outputs": [], + "source": [ + "VECTOR_DIR = [LOCAL_VECTOR_DIR, GMAIL_VECTOR_DIR, OUTLOOK_VECTOR_DIR, GOOGLE_WORKSPACE_VECTOR_DIR, SLACK_VECTOR_DIR]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6d6f2df3-8471-443c-94da-dde496a9a02d", + "metadata": {}, + "outputs": [], + "source": [ + "# system prompt\n", + "prompt_template = PromptTemplate(\n", + " input_variables=[\"question\", \"context\", \"chat_history\"],\n", + " template=\"\"\"\n", + "You are a personal assistant trained on the user's private documents, emails, and notes.\n", + "Your role is to answer questions as if you are the user themself — based on their experiences, thoughts, habits, personality, and preferences reflected in the uploaded materials.\n", + "Also, you are having a conversation with the user. Use the chat history to understand the context of the conversation.\n", + "At the beginning of each conversation, ask the user what name they would like to assign to you. If the user later requests a name change, update your name accordingly without delay.\n", + "\n", + "Use the retrieved documents to:\n", + "- Summarize the user's background, actions, and communication patterns\n", + "- Simulate how the user would respond to questions\n", + "- Infer personality traits, professional history, and personal interests\n", + "\n", + "Always cite the type of source (e.g., email, resume, journal) when appropriate. If no relevant information is available, say so honestly.\n", + "\n", + "You must never make assumptions beyond what the user's data reveals.\n", + "\n", + "Chat History:\n", + "{chat_history}\n", + "\n", + "Retrieved Context:\n", + "{context}\n", + "\n", + "User Question:\n", + "{question}\n", + "\"\"\"\n", + ")\n", + "\n", + "llm = ChatOpenAI(temperature=0.7, model_name=MODEL)\n", + "memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)\n", + "embeddings = OpenAIEmbeddings()\n", + "retrievers = []\n", + "for vec_dir in VECTOR_DIR:\n", + " if os.path.exists(vec_dir):\n", + " vectorstore = Chroma(persist_directory=vec_dir, embedding_function=embeddings)\n", + " retriever = vectorstore.as_retriever(search_kwargs={\"k\": 10})\n", + " retrievers.append(retriever)\n", + "\n", + "merged_retriever = MergerRetriever(retrievers=retrievers)\n", + "conversation_chain = ConversationalRetrievalChain.from_llm(\n", + " llm=llm, \n", + " retriever=merged_retriever, \n", + " memory=memory,\n", + " combine_docs_chain_kwargs={\"prompt\": prompt_template}\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "30861ea3-1005-4fe0-b06b-060571b382bc", + "metadata": {}, + "outputs": [], + "source": [ + "def chat_with_rag(user_input, chat_history):\n", + " result = conversation_chain.invoke({\"question\": user_input})\n", + " answer = result[\"answer\"]\n", + " chat_history.append({\"role\": \"user\", \"content\": user_input})\n", + " chat_history.append({\"role\": \"assistant\", \"content\": answer})\n", + " return \"\", chat_history" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "55be9260-4a05-4f7b-990a-3979ea0a49c3", + "metadata": {}, + "outputs": [], + "source": [ + "def delete_knowledge(delete_type):\n", + " global conversation_chain, retrievers\n", + " \n", + " if delete_type == \"Local Folder\":\n", + " vector_dir = LOCAL_VECTOR_DIR\n", + " elif delete_type == \"Gmail\":\n", + " vector_dir = GMAIL_VECTOR_DIR\n", + " elif delete_type == \"Outlook\":\n", + " vector_dir = OUTLOOK_VECTOR_DIR\n", + " elif delete_type == \"Google Workspace\":\n", + " vector_dir = GOOGLE_WORKSPACE_VECTOR_DIR\n", + " elif delete_type == \"Slack\":\n", + " vector_dir = SLACK_VECTOR_DIR\n", + " \n", + " if os.path.exists(vector_dir):\n", + " Chroma(persist_directory=vector_dir, embedding_function=embeddings).delete_collection()\n", + " retrievers = []\n", + " for vec_dir in VECTOR_DIR:\n", + " if os.path.exists(vec_dir):\n", + " vectorstore = Chroma(persist_directory=vec_dir, embedding_function=embeddings)\n", + " retriever = vectorstore.as_retriever(search_kwargs={\"k\": 10})\n", + " retrievers.append(retriever)\n", + " \n", + " merged_retriever = MergerRetriever(retrievers=retrievers)\n", + " conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=merged_retriever, memory=memory)\n", + " return \"Deleted successfully.\"\n", + " else:\n", + " return \"Vector store does not exist.\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "02f80830-8dd2-4a6a-aca3-0c79f28e703a", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "with gr.Blocks(title=\"Personla Knowledge Assistant\", theme=gr.themes.Citrus(), css=\"\"\"\n", + ".selected {\n", + " background-color: orange !important;\n", + " box-shadow: 0 4px 12px rgba(255, 140, 0, 0.5) !important;\n", + " color: black;\n", + "}\n", + ".unselected {\n", + " background-color: gray !important;\n", + " box-shadow: 0 4px 12px rgba(128, 128, 128, 0.4);\n", + " color: white;\n", + "}\n", + ".gr-button-stop {\n", + " background-color: #cf142b !important;\n", + " color: white !important;\n", + " box-shadow: 0 4px 12px rgba(128, 128, 128, 0.4);\n", + "}\n", + "\"\"\") as ui:\n", + " SESSION_STATE = {\n", + " \"gmail_service\": None, \"gmail_email\": None, \"gmail_alias\": None,\n", + " \"outlook_email\": None, \"outlook_alias\": None,\n", + " \"outlook_login_app\": None, \"outlook_login_flow\": None,\n", + " \"outlook_token_path\": None,\n", + " \"google_workspace_email\": None, \"google_workspace_alias\": None, \n", + " \"google_workspace_drive_service\": None, \"google_workspace_docs_service\": None,\n", + " \"google_workspace_sheets_service\": None, \"google_workspace_slides_service\": None\n", + " }\n", + " outlook_login_flag = gr.State(False)\n", + " current_selected = gr.State(\"\")\n", + " section_names = [\"Local Folder\", \"Gmail\", \"Outlook\", \"Google Workspace\", \"Slack\"]\n", + "\n", + " def show_section(current_selected, current_section):\n", + " updates = []\n", + " if current_selected == current_section:\n", + "\n", + " for sec in section_names:\n", + " updates.append(gr.update(visible=False))\n", + " for sec in section_names:\n", + " updates.append(gr.update(elem_classes=[\"unselected\"]))\n", + " updates.append(\"\")\n", + " else:\n", + " updates = []\n", + " for sec in section_names:\n", + " if sec == current_selected:\n", + " updates.append(gr.update(visible=True))\n", + " else:\n", + " updates.append(gr.update(visible=False))\n", + " for sec in section_names:\n", + " if sec == current_selected:\n", + " updates.append(gr.update(elem_classes=[\"selected\"]))\n", + " else:\n", + " updates.append(gr.update(elem_classes=[\"unselected\"]))\n", + " updates.append(current_selected)\n", + " return tuple(updates)\n", + "\n", + " \n", + " \n", + " gr.Markdown(\"## Personal Knowledge Assistant\")\n", + "\n", + " chatbot = gr.Chatbot(label=\"Chat\", show_copy_button=True, type=\"messages\")\n", + " user_input = gr.Textbox(\n", + " placeholder=\"Talk with your personal knowledge assistant...\",\n", + " label=\"Enter Message\",\n", + " lines=1\n", + " )\n", + " user_input.submit(\n", + " fn=chat_with_rag,\n", + " inputs=[user_input, chatbot],\n", + " outputs=[user_input, chatbot]\n", + " )\n", + " \n", + " gr.HTML(\"
\")\n", + "\n", + " with gr.Row():\n", + " local_folder_show_up = gr.Button(\"Local folder\", elem_id=\"local-folder-btn\", elem_classes=[\"unselected\"])\n", + " gmail_show_up = gr.Button(\"Gmail\", elem_id=\"gmail-btn\", elem_classes=[\"unselected\"])\n", + " outlook_show_up = gr.Button(\"Outlook\", elem_id=\"outlook-btn\", elem_classes=[\"unselected\"])\n", + " google_workspace_show_up = gr.Button(\"Google Workspace\", elem_id=\"google_workspace-btn\", elem_classes=[\"unselected\"])\n", + " slack_show_up = gr.Button(\"Slack\", elem_id=\"Slack-btn\", elem_classes=[\"unselected\"])\n", + " \n", + " local_input = gr.Textbox(value=\"Local Folder\", visible=False)\n", + " gmail_input = gr.Textbox(value=\"Gmail\", visible=False)\n", + " outlook_input = gr.Textbox(value=\"Outlook\", visible=False)\n", + " workspace_input = gr.Textbox(value=\"Google Workspace\", visible=False)\n", + " slack_input = gr.Textbox(value=\"Slack\", visible=False)\n", + " \n", + " local_folder_section = gr.Column(visible=False)\n", + " gmail_section = gr.Column(visible=False)\n", + " outlook_section = gr.Column(visible=False)\n", + " google_workspace_section = gr.Column(visible=False)\n", + " slack_section = gr.Column(visible=False)\n", + "\n", + "\n", + " with local_folder_section:\n", + " gr.Markdown(\"### Local Documents Extractor\")\n", + "\n", + " with gr.Row():\n", + " local_folder_input = gr.Textbox(label=\"Folder Path\", info=\"All subfolders under the selected folder will be extracted.\", value=\"local-knowledge-base\")\n", + " with gr.Row():\n", + " local_exclude_folder_input = gr.Textbox(label=\"Folders to Exclude\", info=\"\\u00A0\", placeholder=\"Join by comma. e.g. dir1, dir2\")\n", + " with gr.Row(): \n", + " local_extract_button = gr.Button(\"Extract Local Documents\")\n", + " with gr.Row(): \n", + " local_extract_log = gr.Textbox(label=\"Extraction Log\", lines=15)\n", + "\n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row(): \n", + " local_delete_button = gr.Button(\"Delete Local Knowledge\", elem_classes=[\"gr-button-stop\"])\n", + " with gr.Row(): \n", + " local_delete_log = gr.Textbox(label=\"Delete Log\", lines=1)\n", + " \n", + " local_delete_button.click(fn=delete_knowledge, inputs=local_input, outputs=local_delete_log)\n", + " local_extract_button.click(fn=extract_local_folder, inputs=[local_folder_input, local_exclude_folder_input], outputs=local_extract_log)\n", + " \n", + " with gmail_section:\n", + " gr.Markdown(\"### Local Documents Extractor\")\n", + " \n", + " with gr.Row():\n", + " gmail_alias_input = gr.Textbox(label=\"Gmail Alias (e.g., zhufqiu)\", placeholder=\"Gmail alias\") \n", + " with gr.Row():\n", + " gmail_login_log = gr.Textbox(label=\"Login Status\", lines=1)\n", + " with gr.Row():\n", + " gmail_login_btn = gr.Button(\"Login\")\n", + " \n", + " gr.HTML(\"
\")\n", + "\n", + " with gr.Row():\n", + " gmail_start_date = gr.Textbox(label=\"Start Date (YYYY/MM/DD)\")\n", + " gmail_end_date = gr.Textbox(label=\"End Date (YYYY/MM/DD)\")\n", + " with gr.Row(): \n", + " gmail_extract_btn = gr.Button(\"Extract Gmail Emails\")\n", + " with gr.Row(): \n", + " gmail_extract_log = gr.Textbox(label=\"Extraction Log\", lines=15)\n", + "\n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row(): \n", + " gmail_delete_button = gr.Button(\"Delete Gmail Knowledge\", elem_classes=[\"gr-button-stop\"])\n", + " with gr.Row(): \n", + " gmail_delete_log = gr.Textbox(label=\"Delete Log\", lines=1)\n", + " \n", + " gmail_delete_button.click(fn=delete_knowledge, inputs=gmail_input, outputs=gmail_delete_log)\n", + " gmail_login_btn.click(fn=login_gmail, inputs=gmail_alias_input, outputs=gmail_login_log)\n", + " gmail_extract_btn.click(fn=extract_gmail, inputs=[gmail_start_date, gmail_end_date], outputs=gmail_extract_log)\n", + " \n", + " with outlook_section:\n", + " gr.Markdown(\"### Outlook Email Extractor\")\n", + "\n", + " with gr.Row():\n", + " outlook_alias = gr.Textbox(label=\"Outlook Alias(e.g., zhufqiu)\", placeholder=\"Outlook alias\")\n", + "\n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row():\n", + " outlook_verify_info = gr.Textbox(label=\"Verification Instructions\", lines=3)\n", + " with gr.Row():\n", + " outlook_start_login_btn = gr.Button(\"Get Verification Code\")\n", + "\n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row():\n", + " outlook_login_log = gr.Textbox(label=\"Login Status\", info=\"\", lines=1)\n", + " with gr.Row():\n", + " outlook_finish_login_btn = gr.Button(\"Login\")\n", + " \n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row():\n", + " outlook_start_date = gr.Textbox(label=\"Start Date (YYYY/MM/DD)\")\n", + " outlook_end_date = gr.Textbox(label=\"End Date (YYYY/MM/DD)\")\n", + " \n", + " with gr.Row():\n", + " outlook_extract_btn = gr.Button(\"Extract Outlook Emails\")\n", + " \n", + " with gr.Row():\n", + " outlook_log = gr.Textbox(label=\"Extraction Log\", lines=15)\n", + "\n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row(): \n", + " outlook_delete_button = gr.Button(\"Delete Outlook Knowledge\", elem_classes=[\"gr-button-stop\"])\n", + " with gr.Row(): \n", + " outlook_delete_log = gr.Textbox(label=\"Delete Log\", lines=1)\n", + " \n", + " outlook_delete_button.click(fn=delete_knowledge, inputs=outlook_input, outputs=outlook_delete_log)\n", + " outlook_start_login_btn.click(fn=start_outlook_login, inputs=outlook_alias, outputs=[outlook_login_flag, outlook_verify_info])\n", + " outlook_finish_login_btn.click(fn=finish_outlook_login, outputs=outlook_login_log)\n", + " outlook_extract_btn.click(fn=extract_outlook_emails, inputs=[outlook_start_date, outlook_end_date], outputs=outlook_log)\n", + "\n", + " with google_workspace_section:\n", + " gr.Markdown(\"### Google Workspace Extractor\")\n", + "\n", + " with gr.Row():\n", + " google_workspace_alias_input = gr.Textbox(label=\"Google Account Alias (e.g., zhufqiu)\", placeholder=\"Google Account alias\") \n", + " with gr.Row():\n", + " google_workspace_login_log = gr.Textbox(label=\"Login Status\", lines=1)\n", + " with gr.Row():\n", + " google_workspace_login_btn = gr.Button(\"Login\")\n", + " \n", + " gr.HTML(\"
\")\n", + "\n", + " with gr.Row():\n", + " google_workspace_folder_input = gr.Textbox(label=\"Folder Path\", info=\"All files under the selected folder will be extracted.\", value=\"google_workspace_knowledge_base\")\n", + " with gr.Row(): \n", + " google_workspace_extract_button = gr.Button(\"Extract Google Workspace Documents\")\n", + " \n", + " with gr.Row(): \n", + " google_workspace_extract_log = gr.Textbox(label=\"Extraction Log\", lines=15)\n", + " \n", + " gr.HTML(\"
\")\n", + " \n", + " with gr.Row(): \n", + " google_workspace_delete_button = gr.Button(\"Delete Google Workspace Knowledge\", elem_classes=[\"gr-button-stop\"])\n", + " with gr.Row(): \n", + " google_workspace_delete_log = gr.Textbox(label=\"Delete Log\", lines=1)\n", + " \n", + " google_workspace_delete_button.click(fn=delete_knowledge, inputs=workspace_input, outputs=google_workspace_delete_log)\n", + " google_workspace_login_btn.click(fn=login_google_workspace, inputs=google_workspace_alias_input, outputs=google_workspace_login_log)\n", + " google_workspace_extract_button.click(fn=extract_google_workspace_folder, inputs=google_workspace_folder_input, outputs=google_workspace_extract_log)\n", + " \n", + " with slack_section:\n", + " gr.Markdown(\"Slack part\")\n", + " gr.Markdown(\"To be developed\")\n", + " \n", + " switch_outputs = [\n", + " local_folder_section, gmail_section, outlook_section, google_workspace_section, slack_section,\n", + " local_folder_show_up, gmail_show_up, outlook_show_up, google_workspace_show_up, slack_show_up,\n", + " current_selected\n", + " ]\n", + "\n", + " gmail_show_up.click(fn=show_section, inputs=[gmail_input, current_selected], outputs=switch_outputs)\n", + " local_folder_show_up.click(fn=show_section, inputs=[local_input, current_selected], outputs=switch_outputs)\n", + " outlook_show_up.click(fn=show_section, inputs=[outlook_input, current_selected], outputs=switch_outputs)\n", + " google_workspace_show_up.click(fn=show_section, inputs=[workspace_input, current_selected], outputs=switch_outputs)\n", + " slack_show_up.click(fn=show_section, inputs=[slack_input, current_selected], outputs=switch_outputs)" + ] + }, + { + "cell_type": "markdown", + "id": "d98536e1-9be1-4b52-8535-dfa4778bb7d8", + "metadata": {}, + "source": [ + "### 7. Launch" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "5ff68e06-3cfb-48ae-9dad-fa431d0d548a", + "metadata": { + "scrolled": true + }, + "outputs": [], + "source": [ + "# Logout all the gmail accounts before launch\n", + "if os.path.exists(GMAIL_TOKEN_DIR):\n", + " shutil.rmtree(GMAIL_TOKEN_DIR)\n", + "os.makedirs(GMAIL_TOKEN_DIR, exist_ok=True)\n", + "\n", + "# Logout all the outlook accounts before launch\n", + "if os.path.exists(OUTLOOK_TOKEN_DIR):\n", + " shutil.rmtree(OUTLOOK_TOKEN_DIR)\n", + "os.makedirs(OUTLOOK_TOKEN_DIR, exist_ok=True)\n", + "\n", + "# Logout all the google accounts before launch\n", + "if os.path.exists(GOOGLE_WORKSPACE_TOKEN_DIR):\n", + " shutil.rmtree(GOOGLE_WORKSPACE_TOKEN_DIR)\n", + "os.makedirs(GOOGLE_WORKSPACE_TOKEN_DIR, exist_ok=True)\n", + "\n", + "ui.launch()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d31bf212-896e-492c-9e3f-88ea5001ab9e", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/gmail_credentials.json b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/gmail_credentials.json new file mode 100644 index 0000000..43b09df --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/gmail_credentials.json @@ -0,0 +1,3 @@ +// delete key + +{"installed":{"client_id":"196620306719-vr5i30l44mqmkmnp7j96iavjfqsfl41f.apps.googleusercontent.com","project_id":"llms-personal-knowledge","auth_uri":"https://accounts.google.com/o/oauth2/auth","token_uri":"https://oauth2.googleapis.com/token","auth_provider_x509_cert_url":"https://www.googleapis.com/oauth2/v1/certs","redirect_uris":["http://localhost"]}} \ No newline at end of file diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/google_drive_workspace_credentials.json b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/google_drive_workspace_credentials.json new file mode 100644 index 0000000..b5af177 --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/credentials/google_drive_workspace_credentials.json @@ -0,0 +1,3 @@ +// delete key + +{"installed":{"client_id":"196620306719-7qvdhd86sau3ngmrrlcb1314us9nuli4.apps.googleusercontent.com","project_id":"llms-personal-knowledge","auth_uri":"https://accounts.google.com/o/oauth2/auth","token_uri":"https://oauth2.googleapis.com/token","auth_provider_x509_cert_url":"https://www.googleapis.com/oauth2/v1/certs","redirect_uris":["http://localhost"]}} \ No newline at end of file diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/image/JPEG.jpg b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/image/JPEG.jpg new file mode 100644 index 0000000..3ac004a Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/image/JPEG.jpg differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/LLMGooglePDF.pdf b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/LLMGooglePDF.pdf new file mode 100644 index 0000000..025a28c Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/LLMGooglePDF.pdf differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/Presentation.pptx b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/Presentation.pptx new file mode 100644 index 0000000..876b4e4 Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/Presentation.pptx differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/excel.xlsx b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/excel.xlsx new file mode 100644 index 0000000..8e57747 Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/excel.xlsx differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/word.docx b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/word.docx new file mode 100644 index 0000000..ac58076 Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/ms_office/word.docx differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/Epub.epub b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/Epub.epub new file mode 100644 index 0000000..45fe318 Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/Epub.epub differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/HTML.html b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/HTML.html new file mode 100644 index 0000000..94fca83 --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/HTML.html @@ -0,0 +1,9 @@ + + + + My First Web Page + + +

Zephyr won ZHTML award

+ + \ No newline at end of file diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/MD.md b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/MD.md new file mode 100644 index 0000000..ac8709f --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/MD.md @@ -0,0 +1 @@ +Zephyr won ZMD award diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/PDF.pdf b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/PDF.pdf new file mode 100644 index 0000000..0de975e Binary files /dev/null and b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/PDF.pdf differ diff --git a/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/text.txt b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/text.txt new file mode 100644 index 0000000..a458ce1 --- /dev/null +++ b/week5/community-contributions/Week5_Exercise_Personal_Knowledge/local-knowledge-base/text/text.txt @@ -0,0 +1 @@ +Zephyr won ZTXT award \ No newline at end of file diff --git a/week5/community-contributions/elchanio_rag_bot/IR_Scraper.ipynb b/week5/community-contributions/elchanio_rag_bot/IR_Scraper.ipynb new file mode 100644 index 0000000..b03e8e3 --- /dev/null +++ b/week5/community-contributions/elchanio_rag_bot/IR_Scraper.ipynb @@ -0,0 +1,101 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "18d85036", + "metadata": {}, + "source": [ + "\n", + "![image](img/spider_bot.png)\n", + "\n", + "## Investor Relations Web Scraping bot\n", + "This code will pop up a Gradio interface to start scraping a website. This is a utility notebook, created to quickly gather documents from IR sites to create a KB. \n", + "I've tuned the scraper to go through the Investor Relations tree of a company website and save all documents with extensions (xls, pdf, word, etc), but not the HTML content.\n", + "\n", + "Due to the way scrapy works with async loops, I had to make a separate script and run it as a subprocess, in order for it to work in a Jupyter notebook.\n", + "\n", + "Can be used to scrape multiple websites (one at a time). Saves scraped files in a kb/{domain} subdirectory (it does **not** preserve website tree structure)\n", + "\n", + "Uses **spider_runner.py**, which needs to be in the same directory as the notebook (will check and abort if not present).\n", + "\n", + "\n", + "### Scraping logic\n", + "scrapy does a pretty decent job of getting the necessary files, although some dynamic sites will not yield the best results. For a more robust scraper I probably need to move to Selenium in a future upgrade. Still, the tool is quite practical for many occasions, as many companies keep their IR websites static. You may need to tweak the follow-on link scraping patterns, I have kept it very simple (it will follow whatever link has 'investor-relations/' in it and limit the links to follow per page to avoid infinite scraping)\n", + "\n", + "In a real application environment we would be running the spider class inside the application - this would enable simpler real-time updates in the output. For an interactive notebook I find this approach sufficient enough." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "69f99b6a", + "metadata": {}, + "outputs": [], + "source": [ + "import subprocess, os, sys\n", + "import gradio as gr\n", + "from urllib.parse import urlparse, urljoin\n", + "\n", + "\n", + "# from urllib.parse import urljoin, urlparse\n", + "# from scrapy.crawler import CrawlerRunner\n", + "# from scrapy.utils.log import configure_logging\n", + "# from twisted.internet import reactor, defer\n", + "# import asyncio\n", + "\n", + "is_scraper_completed = False # global variable to check if the scraper has completed\n", + "status_value= \"Ready\"\n", + "\n", + "with gr.Blocks() as scraper_ui:\n", + " gr.Markdown(\"## Web Scraper\")\n", + " gr.Markdown(\"This is a simple web scraper that can be used to scrape investor relations pages.\")\n", + " \n", + " url = gr.Textbox(label=\"Enter URL\", placeholder=\"https://example.com\")\n", + " \n", + " status = gr.Textbox(label=\"Status\", interactive=False, value=\"Ready to scrape. Enter a URL and press Enter.\", lines=5)\n", + "\n", + " def run_scraper(url):\n", + " # Run the spider as a subprocess\n", + " if not url.startswith(\"http\"):\n", + " url = \"http://\" + url\n", + " # Extract the domain from the URL\n", + " parsed_url = urlparse(url)\n", + " domain = parsed_url.netloc.replace(\"www.\", \"\")\n", + " if not domain:\n", + " return \"Invalid URL. Please enter a valid URL.\"\n", + " # Check if the spider_runner.py file exists\n", + " if not os.path.exists('spider_runner.py'):\n", + " return \"Error: spider_runner.py not found. Please ensure it is in the current directory.\"\n", + " # Run the spider using subprocess\n", + " try:\n", + " result = subprocess.run([sys.executable, 'spider_runner.py', url, domain], check=True, text=True, capture_output=True)\n", + " status_value = f\"Scraping completed for {url}.\"\n", + " is_scraper_completed = True # Set the global variable to True\n", + " return result.stderr, status_value\n", + " except subprocess.CalledProcessError as e:\n", + " is_scraper_completed = True\n", + " status_value = \"Error during scraping. Check the logs for details.\"\n", + " return f\"Error: {e}\", status_value\n", + " \n", + " output = gr.Textbox(label=\"Output\", interactive=False)\n", + " \n", + " url.submit(run_scraper, inputs=url, outputs=[output,status]) \n", + "\n", + "scraper_ui.launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "name": "python", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/elchanio_rag_bot/rag_bot_v01_local.ipynb b/week5/community-contributions/elchanio_rag_bot/rag_bot_v01_local.ipynb new file mode 100644 index 0000000..8bd7449 --- /dev/null +++ b/week5/community-contributions/elchanio_rag_bot/rag_bot_v01_local.ipynb @@ -0,0 +1,303 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2f01b288", + "metadata": {}, + "source": [ + "# RAG personal bot\n", + "\n", + "Exercise for week 5 of LLM Engineering course.\n", + "\n", + "This notebook will create a personal RAG bot. It will use a the ./kb directory to store the files that we want to include in the RAG. Subdirectories will be used to denote categories for the files.\n", + "**Important: only one level of subdirectories will be used for the categories**\n", + "\n", + "It uses LangChain to create and process the RAG pipeline and chat.\n", + "The voector database persistent sotre is in the ./vdb folder. \n", + "\n", + "In this version we use chromadb for the vector store.\n", + "The store is recreated each run. This is not efficient for large datasets. \n", + "\n", + "Future upgrades - To Do (in no particular order): \n", + "- [X] Create a fully local version for security and privacy\n", + "- [ ] Create persistent data store - only load, chunk and embed changed documents. \n", + "- [ ] Provide selection of vector db engines (Chroma DB as default, or connect to external vector db e.g. ElasticSearch or AWS Opensearch)\n", + "- [ ] Add an interface to upload documents in data store - including user-defined metadata tags\n", + "- [ ] Add more document data types\n", + "- [ ] Add online search capability - use web crawler tool to crawl a website and create website-specific RAG bot\n", + "- [ ] Read e-mails/calendars/online docs (Amazon S3 bucket, Google Drive)\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6dfe8e48", + "metadata": {}, + "outputs": [], + "source": [ + "# These were necessary as langchain does not install them by default\n", + "!pip install pypdf\n", + "!pip install pdfminer.six\n", + "!pip install python-docx\n", + "!pip install docx2txt" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "193171c0", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import glob\n", + "from dotenv import load_dotenv\n", + "import gradio as gr\n", + "\n", + "# imports for langchain, plotly and Chroma\n", + "# plotly is commented out, as it is not used in the current code\n", + "\n", + "from langchain.document_loaders import DirectoryLoader, TextLoader, PDFMinerLoader, Docx2txtLoader\n", + "from langchain.text_splitter import RecursiveCharacterTextSplitter\n", + "# from langchain.schema import Document\n", + "from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n", + "from langchain_chroma import Chroma\n", + "#import matplotlib.pyplot as plt\n", + "#from sklearn.manifold import TSNE\n", + "#import numpy as np\n", + "#import plotly.graph_objects as go\n", + "from langchain.memory import ConversationBufferMemory\n", + "from langchain.chains import ConversationalRetrievalChain\n", + "# from langchain.embeddings import HuggingFaceEmbeddings" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d22d2e48", + "metadata": {}, + "outputs": [], + "source": [ + "MODEL = \"gpt-4o-mini\"\n", + "db_name = \"vdb\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc23bf8c", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n" + ] + }, + { + "cell_type": "markdown", + "id": "0103ef35", + "metadata": {}, + "source": [ + "## Loading the documents\n", + "In the code below we read in the KB documents and create the vector store. \n", + "We will be adding PDF documents, Word documents and text/markdown documents.\n", + "Each document has its own loader, which we are calling separately through DirectoryLoader.\n", + "At the end, we are combining the results, and then start splitting the documents using the Recursive Character Text Splitter." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2f20fd20", + "metadata": {}, + "outputs": [], + "source": [ + "# Read in documents using LangChain's loaders\n", + "# Take everything in all the sub-folders of our knowledgebase\n", + "\n", + "folders = glob.glob(\"kb/*\")\n", + "print(f\"Found {len(folders)} folders in the knowledge base.\")\n", + "\n", + "def add_metadata(doc, doc_type):\n", + " doc.metadata[\"doc_type\"] = doc_type\n", + " return doc\n", + "\n", + "# For text files\n", + "text_loader_kwargs = {'encoding': 'utf-8'}\n", + "\n", + "documents = []\n", + "for folder in folders:\n", + " print(f\"Loading documents from folder: {folder}\")\n", + " doc_type = os.path.basename(folder)\n", + " # PDF Loader\n", + " pdf_loader = DirectoryLoader(folder, glob=\"**/*.pdf\", loader_cls=PDFMinerLoader)\n", + " # Text loaders\n", + " txt_loader = DirectoryLoader(folder, glob=\"**/*.txt\", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)\n", + " md_loader = DirectoryLoader(folder, glob=\"**/*.md\", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)\n", + " # Load MS Word documents - UnstructuredWordDocumentLoader does not play well with numpy > 1.24.0, and we use Docx2txtLoader instead. \n", + " # doc_loader = DirectoryLoader(folder, glob=\"**/*.doc\", loader_cls=UnstructuredWordDocumentLoader)\n", + " docx_loader = DirectoryLoader(folder, glob=\"**/*.docx\", loader_cls=Docx2txtLoader)\n", + " # document doc_type is used to identify the type of document\n", + " # Load documents from PDF, text and word files and combine the results\n", + " pdf_docs = pdf_loader.load()\n", + " print(f\"Loaded {len(pdf_docs)} PDF documents from {folder}\")\n", + " text_docs = txt_loader.load() + md_loader.load()\n", + " print(f\"Loaded {len(text_docs)} text documents from {folder}\")\n", + " word_docs = docx_loader.load()\n", + " print(f\"Loaded {len(word_docs)} Word documents from {folder}\")\n", + " folder_docs = pdf_docs + text_docs + word_docs\n", + " # Add metadata to each document\n", + " if not folder_docs:\n", + " print(f\"No documents found in folder: {folder}\")\n", + " continue\n", + " documents.extend([add_metadata(doc, doc_type) for doc in folder_docs])\n", + "\n", + "# Split the documents into chunks\n", + "text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + "chunks = text_splitter.split_documents(documents)\n", + "\n", + "# Print out some basic info for the loaded documents and chunks\n", + "print(f\"Total number of documents: {len(documents)}\")\n", + "print(f\"Total number of chunks: {len(chunks)}\")\n", + "print(f\"Document types found: {set(doc.metadata['doc_type'] for doc in documents)}\")\n" + ] + }, + { + "cell_type": "markdown", + "id": "021cadc7", + "metadata": {}, + "source": [ + "## Vector Store\n", + "\n", + "We use Chromadb for vector store\n", + "Same code as the one in the lesson notebook, minus the visualization part" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "efc70e3a", + "metadata": {}, + "outputs": [], + "source": [ + "# embeddings = OpenAIEmbeddings()\n", + "\n", + "# If you would rather use the free Vector Embeddings from HuggingFace sentence-transformers\n", + "# Then replace embeddings = OpenAIEmbeddings()\n", + "# with:\n", + "from langchain.embeddings import HuggingFaceEmbeddings\n", + "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-MiniLM-L6-v2\")\n", + "\n", + "# Delete if already exists\n", + "\n", + "if os.path.exists(db_name):\n", + " Chroma(persist_directory=db_name, embedding_function=embeddings).delete_collection()\n", + "\n", + "# Create vectorstore\n", + "\n", + "vectorstore = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=db_name)\n", + "print(f\"Vectorstore created with {vectorstore._collection.count()} documents\")\n", + "\n", + "# Let's investigate the vectors\n", + "\n", + "collection = vectorstore._collection\n", + "count = collection.count()\n", + "\n", + "sample_embedding = collection.get(limit=1, include=[\"embeddings\"])[\"embeddings\"][0]\n", + "dimensions = len(sample_embedding)\n", + "print(f\"There are {count:,} vectors with {dimensions:,} dimensions in the vector store\")" + ] + }, + { + "cell_type": "markdown", + "id": "c9af1d32", + "metadata": {}, + "source": [ + "## LangChain\n", + "Create Langchain chat, memory and retrievers.\n", + "\n", + "Note: for this localized version, Gemma3 4B worked much better than Llama 3.2, with my documents. " + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2360006e", + "metadata": {}, + "outputs": [], + "source": [ + "# create a new Chat with OpenAI\n", + "#llm = ChatOpenAI(temperature=0.7, model_name=MODEL)\n", + "\n", + "# Alternative - if you'd like to use Ollama locally, uncomment this line instead\n", + "llm = ChatOpenAI(temperature=0.7, model_name='gemma3:4b', base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "# set up the conversation memory for the chat\n", + "memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)\n", + "\n", + "# the retriever is an abstraction over the VectorStore that will be used during RAG\n", + "retriever = vectorstore.as_retriever(search_kwargs={\"k\": 20}) # k is the number of documents to retrieve\n", + "\n", + "# putting it together: set up the conversation chain with the GPT 3.5 LLM, the vector store and memory\n", + "conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=retriever, memory=memory)" + ] + }, + { + "cell_type": "markdown", + "id": "88a21bb3", + "metadata": {}, + "source": [ + "## UI part\n", + "Create Gradio interface\n", + "\n", + "Simple built-in chat interface\n", + "\n", + "To Do: Add upload interface to include additional documents in data store." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0dfe7d75", + "metadata": {}, + "outputs": [], + "source": [ + "# Wrapping that in a function\n", + "\n", + "def chat(question, history):\n", + " result = conversation_chain.invoke({\"question\": question})\n", + " return result[\"answer\"]\n", + "\n", + "# And in Gradio:\n", + "\n", + "view = gr.ChatInterface(chat, type=\"messages\").launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/elchanio_rag_bot/rag_bot_v02_IR.ipynb b/week5/community-contributions/elchanio_rag_bot/rag_bot_v02_IR.ipynb new file mode 100644 index 0000000..1e26116 --- /dev/null +++ b/week5/community-contributions/elchanio_rag_bot/rag_bot_v02_IR.ipynb @@ -0,0 +1,547 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "2f01b288", + "metadata": {}, + "source": [ + "![image](img/librarian_bot.png)\n", + "\n", + "# RAG bot for investor information\n", + "\n", + "Exercise for week 5 of LLM Engineering course\n", + "\n", + "Specialized bot focusing on analysing financial documents from Investor Relations webpages. \n", + "Comes together with a web crawler spider to gather documents quickly.\n", + "\n", + "This notebook will create a personal RAG bot. It will use a the ./kb directory to store the files that we want to include in the RAG. Subdirectories will be used to denote categories for the files.\n", + "**Important: only one level of subdirectories will be used for the categories**\n", + "\n", + "It uses LangChain to create and process the RAG pipeline and chat.\n", + "The vector database persistent sotre is in the ./vdb folder. \n", + "\n", + "In this version we use chromadb for the vector store.\n", + "The store is recreated each run. This is not efficient for large datasets. \n", + "\n", + "Future upgrades - To Do (in no particular order): \n", + "- [x] Create a fully local version for security and privacy (*see v01_local*) \n", + " NOTE: will require a fairly advanced LLM to answer questions without losing context. 2-4bn parameters LLM's struggle and tend to hallucinate. Best options are gpt-4o-mini and claude-3.5-haiku.\n", + "- [x] Fine tune the pdf scraper to handle financial reports better\n", + "- [x] Create custom retriever for financial information\n", + "- [ ] Create persistent data store between runs - only load, chunk and embed changed documents. \n", + "- [ ] Provide selection of vector db engines (Chroma DB as default, or connect to external vector db e.g. ElasticSearch or AWS Opensearch)\n", + "- [ ] Add an interface to upload documents in data store - including user-defined metadata tags\n", + "- [ ] Multimodality: Process more document data types (e.g. ppt) \n", + "- [x] Add online search capability - use web crawler tool to crawl a website and create website-specific RAG bot\n", + "- [ ] Read e-mails/calendars/online docs (Amazon S3 bucket, Google Drive)\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6dfe8e48", + "metadata": {}, + "outputs": [], + "source": [ + "# These were necessary as langchain does not install them by default\n", + "# !pip install pypdf\n", + "# !pip install pdfminer.six\n", + "# !pip install python-docx\n", + "!pip install docx2txt\n", + "!pip install pymupdf4llm" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "193171c0", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import glob\n", + "from dotenv import load_dotenv\n", + "import gradio as gr\n", + "\n", + "# imports for langchain, plotly and Chroma\n", + "# plotly is commented out, as it is not used in the current code\n", + "\n", + "from langchain.document_loaders import DirectoryLoader, TextLoader, PDFMinerLoader, Docx2txtLoader\n", + "from langchain.text_splitter import RecursiveCharacterTextSplitter\n", + "# from langchain.schema import Document\n", + "from langchain_openai import OpenAIEmbeddings, ChatOpenAI\n", + "from langchain_chroma import Chroma\n", + "#import matplotlib.pyplot as plt\n", + "#from sklearn.manifold import TSNE\n", + "#import numpy as np\n", + "#import plotly.graph_objects as go\n", + "from langchain.memory import ConversationBufferMemory\n", + "from langchain.chains import ConversationalRetrievalChain\n", + "from langchain.embeddings import HuggingFaceEmbeddings" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d22d2e48", + "metadata": {}, + "outputs": [], + "source": [ + "MODEL = \"gpt-4o-mini\"\n", + "db_name = \"vdb\"" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fc23bf8c", + "metadata": {}, + "outputs": [], + "source": [ + "# Load environment variables in a file called .env\n", + "\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n" + ] + }, + { + "cell_type": "markdown", + "id": "0103ef35", + "metadata": {}, + "source": [ + "## Loading the documents\n", + "\n", + "In the code below we read in the KB documents and create the vector store. \n", + "We will be adding PDF documents, Word documents and text/markdown documents. \n", + "Each document has its own loader, which we are calling separately through DirectoryLoader.\n", + "For PDF we implement custom loader to manage financial data. \n", + "\n", + "At the end, we are combining the results, and then start splitting the documents using the Recursive Character Text Splitter.\n", + "\n", + "This approach is not optimal for financial tables.\n", + "TO DO:\n", + " - [x] Replace splitter with better technique that preserves tables.\n", + " - [x] Replace PDF Reader with pymupdf4llm" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "918cbbf0", + "metadata": {}, + "outputs": [], + "source": [ + "# Utility functions for EU financial reporting (read from PDF)\n", + "# We're using pymupdf4llm for better handling of financial reports\n", + "# This function does not utilize a loader class, but directly processes the PDF file\n", + "# It extracts financial sections and returns them as Document objects\\\n", + "\n", + "import pymupdf4llm\n", + "from langchain.schema import Document\n", + "import re\n", + "import string\n", + "from pathlib import Path\n", + "\n", + "def extract_eu_financial_reports(pdf_path):\n", + " \"\"\"\n", + " Extracts financial sections from an EU financial report PDF using pymupdf4llm.\n", + "\n", + " Args:\n", + " pdf_path (str): Path to the PDF file.\n", + "\n", + " Returns:\n", + " list[Document]: A list of LangChain Document objects, each representing a detected financial section\n", + " (e.g., income statement, balance sheet, cash flow statement, etc.) with associated metadata.\n", + "\n", + " The function processes the PDF, detects section headers based on common financial report section names,\n", + " and splits the content accordingly. Each Document contains the section text and metadata including section name,\n", + " content type, source file, and page range.\n", + " \"\"\"\n", + " md_text = pymupdf4llm.to_markdown(\n", + " pdf_path,\n", + " page_chunks=True, # Preserve page boundaries\n", + " write_images=False,\n", + " embed_images=False\n", + " )\n", + " \n", + " # EU financial reports have predictable structures\n", + " financial_sections = [\n", + " \"consolidated income statement\", \"profit and loss\", \"p&l\", \"remuneration report\",\n", + " \"balance sheet\", \"cash flow statement\", \"statement of financial position\",\n", + " \"notes to the consolidated financial statements\", \"segment reporting\",\n", + " \"risk management\", \"capital adequacy\", \"basel\", \"ifrs\", \"regulatory capital\"\n", + " ]\n", + " \n", + " documents = []\n", + " current_section = None\n", + " current_content = \"\"\n", + " start_page = 1\n", + " \n", + " for page_dict in md_text:\n", + " # Extract the actual text content from the dictionary\n", + " page_content = page_dict.get(\"text\", \"\")\n", + " page_num = page_dict.get(\"page\", start_page)\n", + "\n", + " # Detect financial section headers\n", + " content_lower = page_content.lower()\n", + " detected_section = None\n", + " \n", + " for section in financial_sections:\n", + " if section in content_lower:\n", + " detected_section = section\n", + " break\n", + " \n", + " # Process section changes\n", + " if detected_section and detected_section != current_section:\n", + " if current_content:\n", + " # Save previous section\n", + " documents.append(Document(\n", + " page_content=current_content.strip(),\n", + " metadata={\n", + " \"content_type\": \"financial_statement\",\n", + " \"section\": current_section or \"general\",\n", + " \"source\": pdf_path,\n", + " \"pages\": f\"{start_page}-{page_num-1}\"\n", + " }\n", + " ))\n", + " current_section = detected_section\n", + " current_content = page_content\n", + " else:\n", + " current_content += \"\\n---\\n\" + page_content\n", + " \n", + " # Handle final section\n", + " if current_content:\n", + " documents.append(Document(\n", + " page_content=current_content.strip(),\n", + " metadata={\n", + " \"content_type\": \"financial_statement\",\n", + " \"section\": current_section or \"general\",\n", + " \"source\": pdf_path,\n", + " \"pages\": f\"{start_page}-{page_num}\"\n", + " }\n", + " ))\n", + " \n", + " return documents\n", + "\n", + "# Utility functions for loading documents from a folder\n", + "def load_eu_financial_reports_from_directory(directory_path: str, glob_pattern: str = \"*.pdf\"):\n", + " \"\"\"\n", + " Load and process all EU financial reports from a directory.\n", + "\n", + " Args:\n", + " directory_path (str): Path to the directory containing PDF files\n", + " glob_pattern (str, optional): Pattern to match PDF files. Defaults to \"*.pdf\"\n", + "\n", + " Returns:\n", + " list[Document]: A list of LangChain Document objects containing the extracted financial sections\n", + " from all successfully processed PDFs in the directory.\n", + "\n", + " The function iterates through PDF files in the specified directory that match the glob pattern,\n", + " processes each file using extract_eu_financial_reports(), and combines the results into a single list.\n", + " Files that cannot be processed are skipped with an error message printed to stdout.\n", + " \"\"\"\n", + " all_documents = []\n", + " directory = Path(directory_path)\n", + " \n", + " for pdf_file in directory.glob(glob_pattern):\n", + " try:\n", + " documents = extract_eu_financial_reports(str(pdf_file))\n", + " all_documents.extend(documents)\n", + " except Exception as e:\n", + " print(f\"Error processing {pdf_file}: {e}\")\n", + " \n", + " return all_documents\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2f20fd20", + "metadata": {}, + "outputs": [], + "source": [ + "# Read in documents using LangChain's loaders\n", + "# Take everything in all the sub-folders of our knowledgebase\n", + "\n", + "folders = glob.glob(\"kb/*\")\n", + "print(f\"Found {len(folders)} folders in the knowledge base.\")\n", + "\n", + "def add_metadata(doc, doc_type):\n", + " doc.metadata[\"doc_type\"] = doc_type\n", + " return doc\n", + "\n", + "# For text files\n", + "text_loader_kwargs = {'encoding': 'utf-8'}\n", + "\n", + "documents = []\n", + "for folder in folders:\n", + " print(f\"Loading documents from folder: {folder}\")\n", + " doc_type = os.path.basename(folder)\n", + " # PDF Loader\n", + " # We're not using the PDFMinerLoader as it does not handle EU financial reports well.\n", + " # Instead, we use our custom extract_eu_financial_reports function.\n", + " # Uncomment the next line if you want to use the standard loader for PDF files\n", + " # pdf_loader = DirectoryLoader(folder, glob=\"**/*.pdf\", loader_cls=extract_eu_financial_reports)\n", + " # Text loaders\n", + " txt_loader = DirectoryLoader(folder, glob=\"**/*.txt\", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)\n", + " md_loader = DirectoryLoader(folder, glob=\"**/*.md\", loader_cls=TextLoader, loader_kwargs=text_loader_kwargs)\n", + " # Load MS Word documents - UnstructuredWordDocumentLoader does not play well with numpy > 1.24.0, and we use Docx2txtLoader instead. \n", + " # doc_loader = DirectoryLoader(folder, glob=\"**/*.doc\", loader_cls=UnstructuredWordDocumentLoader)\n", + " docx_loader = DirectoryLoader(folder, glob=\"**/*.docx\", loader_cls=Docx2txtLoader)\n", + " # document doc_type is used to identify the type of document\n", + " # Load documents from PDF, text and word files and combine the results\n", + " pdf_docs = load_eu_financial_reports_from_directory(folder)\n", + " print(f\"Loaded {len(pdf_docs)} PDF documents from {folder}\")\n", + " text_docs = txt_loader.load() + md_loader.load()\n", + " print(f\"Loaded {len(text_docs)} text documents from {folder}\")\n", + " word_docs = docx_loader.load()\n", + " print(f\"Loaded {len(word_docs)} Word documents from {folder}\")\n", + " folder_docs = pdf_docs + text_docs + word_docs\n", + " # Add metadata to each document\n", + " if not folder_docs:\n", + " print(f\"No documents found in folder: {folder}\")\n", + " continue\n", + " documents.extend([add_metadata(doc, doc_type) for doc in folder_docs])\n", + "\n", + "# Split the documents into chunks\n", + "text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)\n", + "chunks = text_splitter.split_documents(documents)\n", + "\n", + "# Print out some basic info for the loaded documents and chunks\n", + "print(f\"Total number of documents: {len(documents)}\")\n", + "print(f\"Total number of chunks: {len(chunks)}\")\n", + "print(f\"Document types found: {set(doc.metadata['doc_type'] for doc in documents)}\")\n" + ] + }, + { + "cell_type": "markdown", + "id": "749ad5d8", + "metadata": {}, + "source": [ + "## Vector Store\n", + "\n", + "We use Chromadb for vector store.\n", + "\n", + "Same code as the one in the lesson notebook, minus the visualization part\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "efc70e3a", + "metadata": {}, + "outputs": [], + "source": [ + "#embeddings = OpenAIEmbeddings()\n", + "\n", + "# If you would rather use the free Vector Embeddings from HuggingFace sentence-transformers\n", + "# Then replace embeddings = OpenAIEmbeddings()\n", + "# with:\n", + "from langchain.embeddings import HuggingFaceEmbeddings\n", + "embeddings = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-mpnet-base-v2\") # A bit slower, but better than all-MiniLM-L6-v2 for financial documents\n", + "\n", + "# Delete if already exists\n", + "\n", + "if os.path.exists(db_name):\n", + " Chroma(persist_directory=db_name, embedding_function=embeddings).delete_collection()\n", + "\n", + "# Create vectorstore\n", + "\n", + "vectorstore = Chroma.from_documents(documents=chunks, embedding=embeddings, persist_directory=db_name)\n", + "print(f\"Vectorstore created with {vectorstore._collection.count()} documents\")\n", + "\n", + "# Let's investigate the vectors\n", + "\n", + "collection = vectorstore._collection\n", + "count = collection.count()\n", + "\n", + "sample_embedding = collection.get(limit=1, include=[\"embeddings\"])[\"embeddings\"][0]\n", + "dimensions = len(sample_embedding)\n", + "print(f\"There are {count:,} vectors with {dimensions:,} dimensions in the vector store\")" + ] + }, + { + "cell_type": "markdown", + "id": "c9af1d32", + "metadata": {}, + "source": [ + "## LangChain\n", + "Create Langchain chat, memory and retrievers.\n", + "\n", + "Trying a number of LLM's for ollama. They are not very good at sortingo out the relevant information from financial documents - they do provide results, but tend to be overly chatty and especially the specific numbers can be hallucinated or taken out of context. \n", + "\n", + "GPT-4o-mini provided much more accurate answers to specific questions, even with huggingface's embeddings for the vector store. \n", + "\n", + "Implemented (with Claude's help) a custom retriever and prompt to focus on financial statement analysis.\n", + "\n", + "### OpenAI rate limits\n", + "*Note*: If using OpenAI's embeddings, there's a limit of 300K tokens per request. This requires special handling when calling Chroma.from_documents.\n", + "###TO DO:\n", + "- [ ] Add rate limiter for encoding documents and encode in batches." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "59f75e5d", + "metadata": {}, + "outputs": [], + "source": [ + "# Specialized Retriever for consolidated financials\n", + "\n", + "from langchain.schema import BaseRetriever, Document\n", + "from typing import List\n", + "\n", + "from langchain.vectorstores.base import VectorStoreRetriever\n", + "\n", + "class EUFinancialRetriever(VectorStoreRetriever):\n", + " def _get_relevant_documents(self, query: str, *, run_manager=None) -> List[Document]:\n", + " query_lower = query.lower()\n", + " k = self.search_kwargs.get(\"k\", 5)\n", + " \n", + " # Section-aware search logic\n", + " section_queries = {\n", + " 'income': ['income', 'revenue', 'profit', 'earnings'],\n", + " 'balance': ['balance', 'assets', 'liabilities', 'equity'],\n", + " 'cash': ['cash flow', 'operating cash', 'free cash']\n", + " }\n", + " \n", + " for section, terms in section_queries.items():\n", + " if any(term in query_lower for term in terms):\n", + " try:\n", + " return self.vectorstore.similarity_search(\n", + " query, k=k, filter={\"section\": section}\n", + " )\n", + " except:\n", + " break\n", + " \n", + " # Fallback to standard search\n", + " return self.vectorstore.similarity_search(query, k=k)\n", + "\n", + "\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aca30d15", + "metadata": {}, + "outputs": [], + "source": [ + "# Specialized prompt for the retriever\n", + "\n", + "financial_prompt = \"\"\"\n", + "You are analyzing EU bank and corporate financial statements. When answering:\n", + "\n", + "1. For numerical data, ALWAYS cite the specific financial statement section\n", + "2. Consider regulatory context (IFRS, Basel III for banks)\n", + "3. Note if data spans multiple periods or segments\n", + "4. Highlight any footnotes or adjustments mentioned\n", + "5. Be precise about currency and units (EUR millions, thousands, etc.)\n", + "\n", + "Context from financial statements:\n", + "{context}\n", + "\n", + "Question: {question}\n", + "\n", + "Answer:\n", + "\"\"\"\n", + "# Updated chain with financial-aware prompt\n", + "from langchain.prompts import PromptTemplate\n", + "\n", + "prompt = PromptTemplate(\n", + " input_variables=[\"context\", \"question\"],\n", + " template=financial_prompt\n", + ")\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "2360006e", + "metadata": {}, + "outputs": [], + "source": [ + "# create a new Chat with OpenAI\n", + "llm = ChatOpenAI(temperature=0.7, model_name=MODEL)\n", + "\n", + "# Alternative - if you'd like to use Ollama locally, uncomment this line instead\n", + "#llm = ChatOpenAI(temperature=0.7, model_name='gemma3:4b', base_url='http://localhost:11434/v1', api_key='ollama')\n", + "\n", + "# set up the conversation memory for the chat\n", + "memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)\n", + "\n", + "# the retriever is an abstraction over the VectorStore that will be used during RAG\n", + "retriever = EUFinancialRetriever(\n", + " vectorstore=vectorstore, \n", + " search_kwargs={\"k\": 5}\n", + ")\n", + "# putting it together: set up the conversation chain with the GPT 3.5 LLM, the vector store and memory\n", + "conversation_chain = ConversationalRetrievalChain.from_llm(\n", + " llm=llm, \n", + " retriever=retriever, \n", + " memory=memory, \n", + " combine_docs_chain_kwargs={\"prompt\": prompt},\n", + " return_source_documents=False\n", + " )" + ] + }, + { + "cell_type": "markdown", + "id": "88a21bb3", + "metadata": {}, + "source": [ + "## UI part\n", + "Create Gradio interface\n", + "\n", + "Simple built-in chat interface\n", + "\n", + "###To Do: \n", + "- [ ] Add model selector for Claude 3.5 Haiku\n", + "- [ ] Update interface to handle sources (with **return_source_documents=True**)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0dfe7d75", + "metadata": {}, + "outputs": [], + "source": [ + "# Wrapping that in a function\n", + "\n", + "def chat(question, history):\n", + " result = conversation_chain.invoke({\"question\": question})\n", + " return result[\"answer\"]\n", + "\n", + "# And in Gradio:\n", + "\n", + "view = gr.ChatInterface(chat, type=\"messages\").launch(inbrowser=True)" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "llms", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.13" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week5/community-contributions/elchanio_rag_bot/spider_runner.py b/week5/community-contributions/elchanio_rag_bot/spider_runner.py new file mode 100644 index 0000000..08fa6e1 --- /dev/null +++ b/week5/community-contributions/elchanio_rag_bot/spider_runner.py @@ -0,0 +1,138 @@ +import scrapy +import os +from urllib.parse import urljoin, urlparse +from scrapy.crawler import CrawlerProcess + +class IRWebSpider(scrapy.Spider): + name= 'ir_web_spider' + custom_settings = { + 'LOG_LEVEL': 'INFO', # DEBUG, INFO, WARNING, ERROR + 'DOWNLOAD_DELAY': 1, # Be nice to the server + 'ROBOTSTXT_OBEY': True, + } + num_pages = 10 # how many links to follow per page (Excluding documents) + + def __init__(self, start_urls=None, allowed_domains=None, *args, **kwargs): + super(IRWebSpider, self).__init__(*args, **kwargs) + + # Handle start_urls + if start_urls: + if isinstance(start_urls, str): + self.start_urls = [start_urls] + else: + self.start_urls = list(start_urls) + else: + self.start_urls = [] + + # Handle allowed_domains + if allowed_domains: + if isinstance(allowed_domains, str): + self.allowed_domains = [allowed_domains] + else: + self.allowed_domains = list(allowed_domains) + else: + # Auto-extract domains from start_urls if not provided + self.allowed_domains = [] + for url in self.start_urls: + domain = urlparse(url).netloc + if domain and domain not in self.allowed_domains: + self.allowed_domains.append(domain) + # Log initialization + self.logger.info(f"Spider initialized with start_urls: {self.start_urls}") + self.logger.info(f"Allowed domains: {self.allowed_domains}") + + def start_requests(self): + urls = self.start_urls + if not urls: + raise ValueError("No URLs provided to scrape.") + for url in urls: + self.logger.info(f"Starting request to: {url}") + yield scrapy.Request(url=url, callback=self.parse) + + def parse(self, response): + self.logger.info(f"Parsing response from: {response.url}") + self.logger.info(f"Response status: {response.status}") + # Save the page content + + # Extract document links with better selectors + doc_selectors = [ + 'a[href$=".pdf"]::attr(href)', + 'a[href$=".xlsx"]::attr(href)', + 'a[href$=".xls"]::attr(href)', + 'a[href$=".docx"]::attr(href)', + 'a[href$=".doc"]::attr(href)', + 'a[href$=".pptx"]::attr(href)', + 'a[href$=".ppt"]::attr(href)', + ] + doc_links = [] + for selector in doc_selectors: + links = response.css(selector).getall() + doc_links.extend(links) + self.logger.debug(f"Found {len(links)} links with selector: {selector}") + + self.logger.info(f"Total document links found: {len(doc_links)}") + + if not doc_links: + self.logger.warning("No document links found. Checking page content...") + # Log some of the page content for debugging + self.logger.debug(f"Page title: {response.css('title::text').get()}") + self.logger.debug(f"First 500 chars: {response.text[:500]}") + + for link in doc_links: + full_url = urljoin(response.url, link) + self.logger.info(f"Queuing document: {full_url}") + yield scrapy.Request( + url=full_url, + callback=self.save_document + ) + + # Look for more investor relations pages + ir_links = response.css('a[href*="investor-relations/"]::attr(href)').getall() + + + for link in ir_links[:self.num_pages]: # Limit to avoid infinite crawling + full_url = urljoin(response.url, link) + if full_url != response.url: # Avoid self-loops + self.logger.info(f"Following IR link: {full_url}") + yield scrapy.Request(url=full_url, callback=self.parse) + + + def save_document(self, response): + """Save the document to the local file system. + Will create a directory structure based on the domain and save the file with its original name or a hash if no name is available. + All documents are saved in the 'kb' directory.""" + + self.logger.info(f"Downloading document from: {response.url}") + + parsed_url = urlparse(response.url) + domain = parsed_url.netloc.replace("www.", "") + filename = os.path.basename(parsed_url.path) + if not filename: + filename = f"document_{hash(response.url) % 10000}.bin" + + os.makedirs(f'kb/{domain}', exist_ok=True) + filepath = f'kb/{domain}/{filename}' + + with open(filepath, 'wb') as f: + f.write(response.body) + + file_size = len(response.body) + self.logger.info(f"Saved document: {filepath} ({file_size} bytes)") + +if __name__ == '__main__': + import sys + + start_urls = sys.argv[1] if len(sys.argv) > 1 else 'http://example.com/investor-relations' + allowed_domains = sys.argv[2] if len(sys.argv) > 2 else 'example.com' + + process = CrawlerProcess({ + 'LOG_LEVEL': 'INFO', + 'DOWNLOAD_DELAY': 1, + 'ROBOTSTXT_OBEY': True, + }) + + process.crawl(IRWebSpider, + start_urls=start_urls, + allowed_domains=allowed_domains) + + process.start() \ No newline at end of file diff --git a/week6/community-contributions/day2-improved.ipynb b/week6/community-contributions/day2-improved.ipynb new file mode 100644 index 0000000..f3a2a39 --- /dev/null +++ b/week6/community-contributions/day2-improved.ipynb @@ -0,0 +1,823 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "28a0673e-96b5-43f2-8a8b-bd033bf851b0", + "metadata": {}, + "source": [ + "# The Product Pricer Continued\n", + "\n", + "A model that can estimate how much something costs, from its description.\n", + "\n", + "## Data Curation Part 2\n", + "\n", + "Today we'll extend our dataset to a greater coverage, and craft it into an excellent dataset for training. \n", + "Data curation can seem less exciting than other things we work on, but it's a crucial part of the LLM engineers' responsibility and an important craft to hone, so that you can build your own commercial solutions with high quality datasets.\n", + "\n", + "The dataset is here: \n", + "https://huggingface.co/datasets/McAuley-Lab/Amazon-Reviews-2023\n", + "\n", + "And the folder with all the product datasets is here: \n", + "https://huggingface.co/datasets/McAuley-Lab/Amazon-Reviews-2023/tree/main/raw/meta_categories\n", + "\n", + "Handles Large Datasets: This notebook is designed to efficiently process large datasets like the Amazon Reviews 2023 data, even with limited local resources.\n", + "https://colab.research.google.com/drive/1KY55mHyM5weQMSzHxiDXKSCxB_hItCD2?usp=sharing\n", + "\n", + "## Important Note - read me first please\n", + "\n", + "We are about to craft a massive dataset of 400,000 items covering multiple types of product. In Week 7 we will be using this data to train our own model. It's a pretty big dataset, and depending on the GPU you select, training could take 20+ hours. It will be really good fun, but it could cost a few dollars in compute units.\n", + "\n", + "As an alternative, if you want to keep things quick & low cost, you can work with a smaller dataset focused only on Home Appliances. You'll be able to cover the same learning points; the results will be good -- not quite as good as the full dataset, but still pretty amazing! If you'd prefer to do this, I've set up an alternative jupyter notebook in this folder called `lite.ipynb` that you should use in place of this one.\n", + "\n", + "Also, if you'd prefer, you can shortcut running all this data curation by downloading the pickle files that we save in the last cell. The pickle files are available here: https://drive.google.com/drive/folders/1f_IZGybvs9o0J5sb3xmtTEQB3BXllzrW" + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "67cedf85-8125-4322-998e-9375fe745597", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import random\n", + "from dotenv import load_dotenv\n", + "from huggingface_hub import login\n", + "from datasets import load_dataset, Dataset, DatasetDict\n", + "import matplotlib.pyplot as plt\n", + "from collections import Counter, defaultdict\n", + "import numpy as np\n", + "import pickle" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "446bc939-62fe-4608-bec3-52ae1b2de322", + "metadata": {}, + "outputs": [], + "source": [ + "# Run this in your LOCAL environment to get the exact versions\n", + "import sys\n", + "print(f\"Python version: {sys.version}\")\n", + "print(\"=\"*50)\n", + "\n", + "# Check versions of all your dependencies\n", + "dependencies = [\n", + " 'datasets',\n", + " 'transformers', \n", + " 'huggingface_hub',\n", + " 'matplotlib',\n", + " 'numpy',\n", + " 'python-dotenv', # This is the package name for dotenv\n", + " 'tqdm' # Usually imported by datasets/transformers\n", + "]\n", + "\n", + "# Method 1: Using __version__ attribute\n", + "print(\"DEPENDENCY VERSIONS:\")\n", + "print(\"=\"*50)\n", + "\n", + "for dep in dependencies:\n", + " try:\n", + " if dep == 'python-dotenv':\n", + " import dotenv\n", + " version = dotenv.__version__\n", + " print(f\"python-dotenv: {version}\")\n", + " elif dep == 'huggingface_hub':\n", + " import huggingface_hub\n", + " version = huggingface_hub.__version__\n", + " print(f\"huggingface_hub: {version}\")\n", + " else:\n", + " module = __import__(dep)\n", + " version = getattr(module, '__version__', 'Unknown')\n", + " print(f\"{dep}: {version}\")\n", + " except ImportError:\n", + " print(f\"{dep}: NOT INSTALLED\")\n", + " except AttributeError:\n", + " print(f\"{dep}: Version attribute not found\")\n", + "\n", + "print(\"\\n\" + \"=\"*50)\n", + "print(\"INSTALLATION COMMANDS FOR COLAB:\")\n", + "print(\"=\"*50)\n", + "\n", + "# Method 2: Using pip show (more reliable)\n", + "import subprocess\n", + "import json\n", + "\n", + "def get_pip_version(package):\n", + " try:\n", + " result = subprocess.run([sys.executable, '-m', 'pip', 'show', package], \n", + " capture_output=True, text=True)\n", + " if result.returncode == 0:\n", + " for line in result.stdout.split('\\n'):\n", + " if line.startswith('Version:'):\n", + " return line.split(':', 1)[1].strip()\n", + " except:\n", + " pass\n", + " return None\n", + "\n", + "print(\"# Run these commands in Google Colab:\")\n", + "print(\"# (Copy and paste the exact versions from your local environment)\")\n", + "print()\n", + "\n", + "for dep in dependencies:\n", + " version = get_pip_version(dep)\n", + " if version:\n", + " print(f\"!pip install {dep}=={version}\")\n", + " else:\n", + " print(f\"# !pip install {dep} # Version not found\")\n", + "\n", + "print()\n", + "print(\"# Alternative: Install all at once\")\n", + "install_commands = []\n", + "for dep in dependencies:\n", + " version = get_pip_version(dep)\n", + " if version:\n", + " install_commands.append(f\"{dep}=={version}\")\n", + " else:\n", + " install_commands.append(dep)\n", + "\n", + "print(f\"!pip install {' '.join(install_commands)}\")\n", + "\n", + "print(\"\\n\" + \"=\"*50)\n", + "print(\"ADDITIONAL INFO:\")\n", + "print(\"=\"*50)\n", + "\n", + "# Check if we're in a virtual environment\n", + "print(f\"Virtual environment: {sys.prefix != sys.base_prefix}\")\n", + "print(f\"Python executable: {sys.executable}\")\n", + "\n", + "# Show pip list for reference\n", + "print(\"\\nFull pip list (for reference):\")\n", + "try:\n", + " result = subprocess.run([sys.executable, '-m', 'pip', 'list'], \n", + " capture_output=True, text=True)\n", + " if result.returncode == 0:\n", + " lines = result.stdout.split('\\n')\n", + " relevant_packages = []\n", + " for line in lines:\n", + " for dep in dependencies + ['torch', 'tensorflow', 'tokenizers']:\n", + " if dep.lower() in line.lower():\n", + " relevant_packages.append(line.strip())\n", + " break\n", + " \n", + " for pkg in relevant_packages:\n", + " print(f\" {pkg}\")\n", + "except Exception as e:\n", + " print(f\"Could not get pip list: {e}\")\n", + "\n", + "print(\"\\n\" + \"=\"*50)\n", + "print(\"REQUIREMENTS.TXT FORMAT:\")\n", + "print(\"=\"*50)\n", + "print(\"# Copy this to create a requirements.txt file:\")\n", + "\n", + "for dep in dependencies:\n", + " version = get_pip_version(dep)\n", + " if version:\n", + " print(f\"{dep}=={version}\")\n", + " else:\n", + " print(f\"{dep}\")\n", + "\n", + "print(\"\\n\" + \"=\"*50)\n", + "print(\"COLAB SETUP SCRIPT:\")\n", + "print(\"=\"*50)\n", + "print(\"\"\"# Copy this entire block to run in Colab:\n", + "\n", + "# Install exact versions from local environment\n", + "!pip install --upgrade pip\n", + "\n", + "# Your specific versions (replace with actual versions from above)\"\"\")\n", + "\n", + "for dep in dependencies:\n", + " version = get_pip_version(dep)\n", + " if version:\n", + " print(f\"!pip install {dep}=={version}\")\n", + "\n", + "print(\"\"\"\n", + "# Restart runtime after installation\n", + "import os\n", + "os.kill(os.getpid(), 9) # This will restart the runtime\n", + "\"\"\")\n", + "\n", + "print(\"\\n\" + \"=\"*50)\n", + "print(\"VERIFICATION SCRIPT FOR COLAB:\")\n", + "print(\"=\"*50)\n", + "print(\"\"\"# Run this in Colab AFTER installing to verify versions match:\n", + "\n", + "import sys\n", + "dependencies_to_check = [\n", + " 'datasets', 'transformers', 'huggingface_hub', \n", + " 'matplotlib', 'numpy', 'dotenv', 'tqdm'\n", + "]\n", + "\n", + "print(\"Verification of installed versions:\")\n", + "print(\"=\"*40)\n", + "for dep in dependencies_to_check:\n", + " try:\n", + " if dep == 'dotenv':\n", + " import dotenv as module\n", + " else:\n", + " module = __import__(dep)\n", + " version = getattr(module, '__version__', 'Unknown')\n", + " print(f\"{dep}: {version}\")\n", + " except ImportError:\n", + " print(f\"{dep}: NOT INSTALLED\")\n", + "\n", + "print(\"\\\\nIf all versions match your local environment, the code should work!\")\n", + "\"\"\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7390a6aa-79cb-4dea-b6d7-de7e4b13e472", + "metadata": {}, + "outputs": [], + "source": [ + "# environment\n", + "\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n", + "os.environ['ANTHROPIC_API_KEY'] = os.getenv('ANTHROPIC_API_KEY', 'your-key-if-not-using-env')\n", + "os.environ['HF_TOKEN'] = os.getenv('HF_TOKEN', 'your-key-if-not-using-env')" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0732274a-aa6a-44fc-aee2-40dc8a8e4451", + "metadata": {}, + "outputs": [], + "source": [ + "# Log in to HuggingFace\n", + "\n", + "hf_token = os.environ['HF_TOKEN']\n", + "login(hf_token, add_to_git_credential=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6746144c-2e19-485a-8086-368c144722b4", + "metadata": {}, + "outputs": [], + "source": [ + "# More imports after HF login\n", + "\n", + "from loaders import ItemLoader\n", + "from items import Item" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "1adcf323-de9d-4c24-a9c3-d7ae554d06ca", + "metadata": {}, + "outputs": [], + "source": [ + "%matplotlib inline" + ] + }, + { + "cell_type": "markdown", + "id": "01065d69-765c-42c8-9f90-68b8c8754068", + "metadata": {}, + "source": [ + "## The ItemLoader code\n", + "\n", + "Look in loaders.py - there's some useful code to make life easier for us" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "049885d4-fdfa-4ff0-a932-4a2ed73928e2", + "metadata": {}, + "outputs": [], + "source": [ + "# Load in the same dataset as last time\n", + "\n", + "items = ItemLoader(\"All_Beauty\").load()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "ffba41b5-ddb6-4359-9790-9b2db900eee1", + "metadata": {}, + "outputs": [], + "source": [ + "# Look for a familiar item..\n", + "print(items[1].prompt)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7cc7f3e7-e98e-48c1-8eed-1608b42b0f65", + "metadata": {}, + "outputs": [], + "source": [ + "import datasets\n", + "print(datasets.__version__)" + ] + }, + { + "cell_type": "markdown", + "id": "e2b6dc50-ac5c-4cf2-af2e-968ed8ef86d7", + "metadata": {}, + "source": [ + "## Now to SCALE UP\n", + "\n", + "Let's look at all datasets of all the items that you might find in a large home retail store - electrical, electronic, office and related, but not clothes / beauty / books." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d1d06cd3-f3c2-44f0-a9f2-13b54ff8be5c", + "metadata": {}, + "outputs": [], + "source": [ + "dataset_names = [\n", + " \"Automotive\",\n", + " \"Electronics\",\n", + " \"Office_Products\",\n", + " \"Tools_and_Home_Improvement\",\n", + " \"Cell_Phones_and_Accessories\",\n", + " \"Toys_and_Games\",\n", + " \"Appliances\",\n", + " \"Musical_Instruments\",\n", + "]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "aa8fd0f0-509a-4298-8fcc-e499a061e1be", + "metadata": {}, + "outputs": [], + "source": [ + "items = []\n", + "for dataset_name in dataset_names:\n", + " loader = ItemLoader(dataset_name)\n", + " items.extend(loader.load())\n", + "\n", + "# Now, time for a coffee break!!\n", + "# By the way, I put the biggest datasets first.. it gets faster." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3e29a5ab-ca61-41cc-9b33-22d374681b85", + "metadata": {}, + "outputs": [], + "source": [ + "print(f\"A grand total of {len(items):,} items\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "89078cb1-9679-4eb0-b295-599b8586bcd1", + "metadata": {}, + "outputs": [], + "source": [ + "# Plot the distribution of token counts again\n", + "\n", + "tokens = [item.token_count for item in items]\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Token counts: Avg {sum(tokens)/len(tokens):,.1f} and highest {max(tokens):,}\\n\")\n", + "plt.xlabel('Length (tokens)')\n", + "plt.ylabel('Count')\n", + "plt.hist(tokens, rwidth=0.7, color=\"skyblue\", bins=range(0, 300, 10))\n", + "plt.show()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "c38e0c43-9f7a-450e-a911-c94d37d9b9c3", + "metadata": {}, + "outputs": [], + "source": [ + "# Plot the distribution of prices\n", + "\n", + "prices = [item.price for item in items]\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Prices: Avg {sum(prices)/len(prices):,.1f} and highest {max(prices):,}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"blueviolet\", bins=range(0, 1000, 10))\n", + "plt.show()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "eabc7c61-0cd2-41f4-baa1-b85400bbf87f", + "metadata": {}, + "outputs": [], + "source": [ + "category_counts = Counter()\n", + "for item in items:\n", + " category_counts[item.category]+=1\n", + "\n", + "categories = category_counts.keys()\n", + "counts = [category_counts[category] for category in categories]\n", + "\n", + "# Bar chart by category\n", + "plt.figure(figsize=(15, 6))\n", + "plt.bar(categories, counts, color=\"goldenrod\")\n", + "plt.title('How many in each category')\n", + "plt.xlabel('Categories')\n", + "plt.ylabel('Count')\n", + "\n", + "plt.xticks(rotation=30, ha='right')\n", + "\n", + "# Add value labels on top of each bar\n", + "for i, v in enumerate(counts):\n", + " plt.text(i, v, f\"{v:,}\", ha='center', va='bottom')\n", + "\n", + "# Display the chart\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "e5b6e987-83ba-4262-a082-57c6b0741062", + "metadata": {}, + "source": [ + "# Objective\n", + "\n", + "Craft a dataset which is more balanced in terms of prices. Less heavily scewed to cheap items, with an average that's higher than $60. Try to balance out the categories - fewer Automotive items." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b9424c1-44e0-499a-b45e-a35246655469", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a dict with a key of each price from $1 to $999\n", + "# And in the value, put a list of items with that price (to nearest round number)\n", + "\n", + "slots = defaultdict(list)\n", + "for item in items:\n", + " slots[round(item.price)].append(item)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7805a7f1-4ad8-48f6-bea3-d64b64894804", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a dataset called \"sample\" which tries to more evenly take from the range of prices\n", + "# And gives more weight to items from categories other than Automotive\n", + "# Set random seed for reproducibility\n", + "\n", + "np.random.seed(42)\n", + "random.seed(42)\n", + "sample = []\n", + "for i in range(1, 1000):\n", + " slot = slots[i]\n", + " if i>=240:\n", + " sample.extend(slot)\n", + " elif len(slot) <= 1200:\n", + " sample.extend(slot)\n", + " else:\n", + " weights = np.array([1 if item.category=='Automotive' else 5 for item in slot])\n", + " weights = weights / np.sum(weights)\n", + " selected_indices = np.random.choice(len(slot), size=1200, replace=False, p=weights)\n", + " selected = [slot[i] for i in selected_indices]\n", + " sample.extend(selected)\n", + "\n", + "print(f\"There are {len(sample):,} items in the sample\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "430b432f-b769-41da-9506-a238cb5cf1b6", + "metadata": {}, + "outputs": [], + "source": [ + "# Plot the distribution of prices in sample\n", + "\n", + "prices = [float(item.price) for item in sample]\n", + "plt.figure(figsize=(15, 10))\n", + "plt.title(f\"Avg {sum(prices)/len(prices):.2f} and highest {max(prices):,.2f}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"darkblue\", bins=range(0, 1000, 10))\n", + "plt.show()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "0d570794-6f1d-462e-b567-a46bae3556a1", + "metadata": {}, + "outputs": [], + "source": [ + "# OK, we did well in terms of raising the average price and having a smooth-ish population of prices\n", + "# Let's see the categories\n", + "\n", + "category_counts = Counter()\n", + "for item in sample:\n", + " category_counts[item.category]+=1\n", + "\n", + "categories = category_counts.keys()\n", + "counts = [category_counts[category] for category in categories]\n", + "\n", + "# Create bar chart\n", + "plt.figure(figsize=(15, 6))\n", + "plt.bar(categories, counts, color=\"lightgreen\")\n", + "\n", + "# Customize the chart\n", + "plt.title('How many in each category')\n", + "plt.xlabel('Categories')\n", + "plt.ylabel('Count')\n", + "\n", + "plt.xticks(rotation=30, ha='right')\n", + "\n", + "# Add value labels on top of each bar\n", + "for i, v in enumerate(counts):\n", + " plt.text(i, v, f\"{v:,}\", ha='center', va='bottom')\n", + "\n", + "# Display the chart\n", + "plt.show()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "6609d77c-3e0a-4679-9129-c7cdc3273070", + "metadata": {}, + "outputs": [], + "source": [ + "# Automotive still in the lead, but improved somewhat\n", + "# For another perspective, let's look at a pie\n", + "\n", + "plt.figure(figsize=(12, 10))\n", + "plt.pie(counts, labels=categories, autopct='%1.0f%%', startangle=90)\n", + "\n", + "# Add a circle at the center to create a donut chart (optional)\n", + "centre_circle = plt.Circle((0,0), 0.70, fc='white')\n", + "fig = plt.gcf()\n", + "fig.gca().add_artist(centre_circle)\n", + "plt.title('Categories')\n", + "\n", + "# Equal aspect ratio ensures that pie is drawn as a circle\n", + "plt.axis('equal') \n", + "\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "ac046cc1-2717-415b-96ad-b73b2950d235", + "metadata": {}, + "source": [ + "# Dataset Curated!\n", + "\n", + "We've crafted an excellent dataset.\n", + "\n", + "Let's do some final checks" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "70219e99-22cc-4e08-9121-51f9707caef0", + "metadata": {}, + "outputs": [], + "source": [ + "# How does the price vary with the character count of the prompt?\n", + "\n", + "sizes = [len(item.prompt) for item in sample]\n", + "prices = [item.price for item in sample]\n", + "\n", + "# Create the scatter plot\n", + "plt.figure(figsize=(15, 8))\n", + "plt.scatter(sizes, prices, s=0.2, color=\"red\")\n", + "\n", + "# Add labels and title\n", + "plt.xlabel('Size')\n", + "plt.ylabel('Price')\n", + "plt.title('Is there a simple correlation?')\n", + "\n", + "# Display the plot\n", + "plt.show()" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "30ae1453-b9fc-40db-8310-65d850c4b1da", + "metadata": {}, + "outputs": [], + "source": [ + "def report(item):\n", + " prompt = item.prompt\n", + " tokens = Item.tokenizer.encode(item.prompt)\n", + " print(prompt)\n", + " print(tokens[-10:])\n", + " print(Item.tokenizer.batch_decode(tokens[-10:]))" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "d9998b8d-d746-4541-9ac2-701108e0e8fb", + "metadata": {}, + "outputs": [], + "source": [ + "report(sample[398000])" + ] + }, + { + "cell_type": "markdown", + "id": "7aa0a3fc-d2fe-4e6e-8fdb-96913df2f588", + "metadata": {}, + "source": [ + "## Observation\n", + "\n", + "An interesting thing about the Llama tokenizer is that every number from 1 to 999 gets mapped to 1 token, much as we saw with gpt-4o. The same is not true of qwen2, gemma and phi3, which all map individual digits to tokens. This does turn out to be a bit useful for our project, although it's not an essential requirement." + ] + }, + { + "cell_type": "markdown", + "id": "0f03c0ee-3103-4603-af5c-b484884a3aa2", + "metadata": {}, + "source": [ + "# Finally\n", + "\n", + "It's time to break down our data into a training, test and validation dataset.\n", + "\n", + "It's typical to use 5%-10% of your data for testing purposes, but actually we have far more than we need at this point. We'll take 400,000 points for training, and we'll reserve 2,000 for testing, although we won't use all of them.\n" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "3b163ca2-18ef-4c26-8e9d-88eb55f114f6", + "metadata": {}, + "outputs": [], + "source": [ + "random.seed(42)\n", + "random.shuffle(sample)\n", + "train = sample[:400_000]\n", + "test = sample[400_000:402_000]\n", + "print(f\"Divided into a training set of {len(train):,} items and test set of {len(test):,} items\")" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "299b9816-8885-4798-829a-69d66d60eb01", + "metadata": {}, + "outputs": [], + "source": [ + "print(train[0].prompt)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "97222da3-9f2c-4d15-a5cd-5e5f8dbde6cc", + "metadata": {}, + "outputs": [], + "source": [ + "print(test[0].test_prompt())" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "7a116369-335a-412b-b70c-2add6675c2e3", + "metadata": {}, + "outputs": [], + "source": [ + "# Plot the distribution of prices in the first 250 test points\n", + "\n", + "prices = [float(item.price) for item in test[:250]]\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Avg {sum(prices)/len(prices):.2f} and highest {max(prices):,.2f}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"darkblue\", bins=range(0, 1000, 10))\n", + "plt.show()" + ] + }, + { + "cell_type": "markdown", + "id": "d522d752-6f66-4786-a4dc-8ef51842558c", + "metadata": {}, + "source": [ + "# Finally - upload your brand new dataset\n", + "\n", + "Convert to prompts and upload to HuggingFace hub" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "fa11b3e5-fcf4-4efc-a573-f6f67fec3e73", + "metadata": {}, + "outputs": [], + "source": [ + "train_prompts = [item.prompt for item in train]\n", + "train_prices = [item.price for item in train]\n", + "test_prompts = [item.test_prompt() for item in test]\n", + "test_prices = [item.price for item in test]" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b020ab1b-7153-4e5f-b8a3-d5bc2fafb6df", + "metadata": {}, + "outputs": [], + "source": [ + "# Create a Dataset from the lists\n", + "\n", + "train_dataset = Dataset.from_dict({\"text\": train_prompts, \"price\": train_prices})\n", + "test_dataset = Dataset.from_dict({\"text\": test_prompts, \"price\": test_prices})\n", + "dataset = DatasetDict({\n", + " \"train\": train_dataset,\n", + " \"test\": test_dataset\n", + "})" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "17639641-fb55-44e2-a463-b0b394d00f32", + "metadata": {}, + "outputs": [], + "source": [ + "# Uncomment these lines if you're ready to push to the hub, and replace my name with your HF username\n", + "\n", + "# HF_USER = \"ed-donner\"\n", + "# DATASET_NAME = f\"{HF_USER}/pricer-data\"\n", + "# dataset.push_to_hub(DATASET_NAME, private=True)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "b85733ba-d165-4f07-b055-46803543edfe", + "metadata": {}, + "outputs": [], + "source": [ + "# One more thing!\n", + "# Let's pickle the training and test dataset so we don't have to execute all this code next time!\n", + "\n", + "with open('train.pkl', 'wb') as file:\n", + " pickle.dump(train, file)\n", + "\n", + "with open('test.pkl', 'wb') as file:\n", + " pickle.dump(test, file)" + ] + }, + { + "cell_type": "markdown", + "id": "2b58dc61-747f-46f7-b9e0-c205db4f3e5e", + "metadata": {}, + "source": [ + "## Todos for you:\n", + "\n", + "- Investigate the dataset more!\n", + "- Confirm that the tokenizer tokenizes all 3 digit prices into 1 token" + ] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.12" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week6/community-contributions/day5-improved.ipynb b/week6/community-contributions/day5-improved.ipynb new file mode 100644 index 0000000..152abaa --- /dev/null +++ b/week6/community-contributions/day5-improved.ipynb @@ -0,0 +1,1097 @@ +{ + "cells": [ + { + "cell_type": "markdown", + "id": "db8736a7-ed94-441c-9556-831fa57b5a10", + "metadata": {}, + "source": [ + "# The Product Pricer Continued\n", + "\n", + "A model that can estimate how much something costs, from its description.\n", + "\n", + "## AT LAST - it's time for Fine Tuning!\n", + "\n", + "After all this data preparation, and old school machine learning, we've finally arrived at the moment you've been waiting for. Fine-tuning a model." + ] + }, + { + "cell_type": "code", + "execution_count": 1, + "id": "681c717b-4c24-4ac3-a5f3-3c5881d6e70a", + "metadata": {}, + "outputs": [], + "source": [ + "# imports\n", + "\n", + "import os\n", + "import re\n", + "import math\n", + "import json\n", + "import random\n", + "from dotenv import load_dotenv\n", + "from huggingface_hub import login\n", + "import matplotlib.pyplot as plt\n", + "import numpy as np\n", + "import pickle\n", + "from collections import Counter\n", + "from openai import OpenAI\n", + "from anthropic import Anthropic" + ] + }, + { + "cell_type": "code", + "execution_count": 2, + "id": "36d05bdc-0155-4c72-a7ee-aa4e614ffd3c", + "metadata": {}, + "outputs": [], + "source": [ + "# environment\n", + "\n", + "load_dotenv(override=True)\n", + "os.environ['OPENAI_API_KEY'] = os.getenv('OPENAI_API_KEY', 'your-key-if-not-using-env')\n", + "os.environ['ANTHROPIC_API_KEY'] = os.getenv('ANTHROPIC_API_KEY', 'your-key-if-not-using-env')\n", + "os.environ['HF_TOKEN'] = os.getenv('HF_TOKEN', 'your-key-if-not-using-env')" + ] + }, + { + "cell_type": "code", + "execution_count": 3, + "id": "4dd3aad2-6f99-433c-8792-e461d2f06622", + "metadata": {}, + "outputs": [ + { + "name": "stderr", + "output_type": "stream", + "text": [ + "Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.\n" + ] + } + ], + "source": [ + "# Log in to HuggingFace\n", + "\n", + "hf_token = os.environ['HF_TOKEN']\n", + "login(hf_token, add_to_git_credential=True)" + ] + }, + { + "cell_type": "code", + "execution_count": 51, + "id": "884a50bd-8cae-425e-8e56-f079fc3e65ce", + "metadata": {}, + "outputs": [], + "source": [ + "# moved our Tester into a separate package\n", + "# call it with Tester.test(function_name, test_dataset)\n", + "\n", + "from items import Item\n", + "from testing import Tester" + ] + }, + { + "cell_type": "code", + "execution_count": 5, + "id": "b0a6fb86-74a4-403c-ab25-6db2d74e9d2b", + "metadata": {}, + "outputs": [], + "source": [ + "openai = OpenAI()" + ] + }, + { + "cell_type": "code", + "execution_count": 6, + "id": "c830ed3e-24ee-4af6-a07b-a1bfdcd39278", + "metadata": {}, + "outputs": [], + "source": [ + "%matplotlib inline" + ] + }, + { + "cell_type": "code", + "execution_count": 7, + "id": "5c9b05f4-c9eb-462c-8d86-de9140a2d985", + "metadata": {}, + "outputs": [], + "source": [ + "# Let's avoid curating all our data again! Load in the pickle files:\n", + "\n", + "with open('train.pkl', 'rb') as file:\n", + " train = pickle.load(file)\n", + "\n", + "with open('test.pkl', 'rb') as file:\n", + " test = pickle.load(file)" + ] + }, + { + "cell_type": "code", + "execution_count": 8, + "id": "e8367135-f40e-43e1-8f3c-09e990ab1194", + "metadata": {}, + "outputs": [], + "source": [ + "# OpenAI recommends fine-tuning with populations of 50-100 examples\n", + "# But as our examples are very small, I'm suggesting we go with 200 examples (and 1 epoch)\n", + "\n", + "fine_tune_train = train[:200]\n", + "fine_tune_validation = train[200:250]" + ] + }, + { + "cell_type": "markdown", + "id": "8be4a889-81c3-42b1-a2fc-034cdc7321a6", + "metadata": {}, + "source": [ + "# Step 1\n", + "\n", + "Prepare our data for fine-tuning in JSONL (JSON Lines) format and upload to OpenAI" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ae2fb3c-1cff-4ce3-911e-627c970edd7b", + "metadata": {}, + "outputs": [], + "source": [ + "# # First let's work on a good prompt for a Frontier model\n", + "# # Notice that I'm removing the \" to the nearest dollar\"\n", + "# # When we train our own models, we'll need to make the problem as easy as possible, \n", + "# # but a Frontier model needs no such simplification.\n", + "\n", + "# def messages_for(item):\n", + "# system_message = \"You estimate prices of items. Reply only with the price, no explanation\"\n", + "# user_prompt = item.test_prompt().replace(\" to the nearest dollar\",\"\").replace(\"\\n\\nPrice is $\",\"\")\n", + "# return [\n", + "# {\"role\": \"system\", \"content\": system_message},\n", + "# {\"role\": \"user\", \"content\": user_prompt},\n", + "# {\"role\": \"assistant\", \"content\": f\"Price is ${item.price:.2f}\"}\n", + "# ]" + ] + }, + { + "cell_type": "code", + "execution_count": 9, + "id": "ca3c0910-1919-47f8-8800-b12be4e983e9", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(item):\n", + " system_message = \"\"\"You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\n", + "\n", + "Key Amazon pricing factors to evaluate:\n", + "- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\n", + "- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\n", + "- Pack size and quantity (bulk/multi-packs often better per-unit value)\n", + "- Prime eligibility indicators and fulfillment method signals\n", + "- Product variations (color, size, model) affecting price tiers\n", + "- Feature density and specification richness\n", + "- Amazon's Choice or bestseller indicators in description\n", + "- Customer rating implications (4.5+ stars = premium pricing power)\n", + "- Seasonal/trending product indicators\n", + "\n", + "Amazon-specific pricing patterns:\n", + "- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\n", + "- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\n", + "- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\n", + "- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\n", + "- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\n", + "- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\n", + "\n", + "Consider Amazon's psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\n", + "\n", + "Output format: Respond with only the price including dollar sign and cents (e.g., \"$24.99\"). No explanations or additional text.\"\"\"\n", + " \n", + " user_prompt = item.test_prompt().replace(\" to the nearest dollar\",\"\").replace(\"\\n\\nPrice is $\",\"\")\n", + " \n", + " return [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt},\n", + " {\"role\": \"assistant\", \"content\": f\"${item.price:.2f}\"}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 10, + "id": "1aa280f6-1227-426a-a2e2-1ce985feba1e", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "[{'role': 'system',\n", + " 'content': 'You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\\n\\nKey Amazon pricing factors to evaluate:\\n- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\\n- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\\n- Pack size and quantity (bulk/multi-packs often better per-unit value)\\n- Prime eligibility indicators and fulfillment method signals\\n- Product variations (color, size, model) affecting price tiers\\n- Feature density and specification richness\\n- Amazon\\'s Choice or bestseller indicators in description\\n- Customer rating implications (4.5+ stars = premium pricing power)\\n- Seasonal/trending product indicators\\n\\nAmazon-specific pricing patterns:\\n- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\\n- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\\n- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\\n- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\\n- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\\n- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\\n\\nConsider Amazon\\'s psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\\n\\nOutput format: Respond with only the price including dollar sign and cents (e.g., \"$24.99\"). No explanations or additional text.'},\n", + " {'role': 'user',\n", + " 'content': 'How much does this cost?\\n\\nDelphi FG0166 Fuel Pump Module\\nDelphi brings 80 years of OE Heritage into each Delphi pump, ensuring quality and fitment for each Delphi part. Part is validated, tested and matched to the right vehicle application Delphi brings 80 years of OE Heritage into each Delphi assembly, ensuring quality and fitment for each Delphi part Always be sure to check and clean fuel tank to avoid unnecessary returns Rigorous OE-testing ensures the pump can withstand extreme temperatures Brand Delphi, Fit Type Vehicle Specific Fit, Dimensions LxWxH 19.7 x 7.7 x 5.1 inches, Weight 2.2 Pounds, Auto Part Position Unknown, Operation Mode Mechanical, Manufacturer Delphi, Model FUEL PUMP, Dimensions 19.7'},\n", + " {'role': 'assistant', 'content': '$226.95'}]" + ] + }, + "execution_count": 10, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "messages_for(train[0])" + ] + }, + { + "cell_type": "code", + "execution_count": 11, + "id": "c0e5b56c-8a0b-4d8e-a112-ce87efb4e152", + "metadata": {}, + "outputs": [], + "source": [ + "# Convert the items into a list of json objects - a \"jsonl\" string\n", + "# Each row represents a message in the form:\n", + "# {\"messages\" : [{\"role\": \"system\", \"content\": \"You estimate prices...\n", + "\n", + "\n", + "def make_jsonl(items):\n", + " result = \"\"\n", + " for item in items:\n", + " messages = messages_for(item)\n", + " messages_str = json.dumps(messages)\n", + " result += '{\"messages\": ' + messages_str +'}\\n'\n", + " return result.strip()" + ] + }, + { + "cell_type": "code", + "execution_count": 12, + "id": "5e72de93-a6a6-4b35-855e-15786b97bf5f", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "{\"messages\": [{\"role\": \"system\", \"content\": \"You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\\n\\nKey Amazon pricing factors to evaluate:\\n- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\\n- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\\n- Pack size and quantity (bulk/multi-packs often better per-unit value)\\n- Prime eligibility indicators and fulfillment method signals\\n- Product variations (color, size, model) affecting price tiers\\n- Feature density and specification richness\\n- Amazon's Choice or bestseller indicators in description\\n- Customer rating implications (4.5+ stars = premium pricing power)\\n- Seasonal/trending product indicators\\n\\nAmazon-specific pricing patterns:\\n- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\\n- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\\n- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\\n- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\\n- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\\n- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\\n\\nConsider Amazon's psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\\n\\nOutput format: Respond with only the price including dollar sign and cents (e.g., \\\"$24.99\\\"). No explanations or additional text.\"}, {\"role\": \"user\", \"content\": \"How much does this cost?\\n\\nDelphi FG0166 Fuel Pump Module\\nDelphi brings 80 years of OE Heritage into each Delphi pump, ensuring quality and fitment for each Delphi part. Part is validated, tested and matched to the right vehicle application Delphi brings 80 years of OE Heritage into each Delphi assembly, ensuring quality and fitment for each Delphi part Always be sure to check and clean fuel tank to avoid unnecessary returns Rigorous OE-testing ensures the pump can withstand extreme temperatures Brand Delphi, Fit Type Vehicle Specific Fit, Dimensions LxWxH 19.7 x 7.7 x 5.1 inches, Weight 2.2 Pounds, Auto Part Position Unknown, Operation Mode Mechanical, Manufacturer Delphi, Model FUEL PUMP, Dimensions 19.7\"}, {\"role\": \"assistant\", \"content\": \"$226.95\"}]}\n", + "{\"messages\": [{\"role\": \"system\", \"content\": \"You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\\n\\nKey Amazon pricing factors to evaluate:\\n- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\\n- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\\n- Pack size and quantity (bulk/multi-packs often better per-unit value)\\n- Prime eligibility indicators and fulfillment method signals\\n- Product variations (color, size, model) affecting price tiers\\n- Feature density and specification richness\\n- Amazon's Choice or bestseller indicators in description\\n- Customer rating implications (4.5+ stars = premium pricing power)\\n- Seasonal/trending product indicators\\n\\nAmazon-specific pricing patterns:\\n- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\\n- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\\n- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\\n- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\\n- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\\n- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\\n\\nConsider Amazon's psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\\n\\nOutput format: Respond with only the price including dollar sign and cents (e.g., \\\"$24.99\\\"). No explanations or additional text.\"}, {\"role\": \"user\", \"content\": \"How much does this cost?\\n\\nPower Stop Rear Z36 Truck and Tow Brake Kit with Calipers\\nThe Power Stop Z36 Truck & Tow Performance brake kit provides the superior stopping power demanded by those who tow boats, haul loads, tackle mountains, lift trucks, and play in the harshest conditions. The brake rotors are drilled to keep temperatures down during extreme braking and slotted to sweep away any debris for constant pad contact. Combined with our Z36 Carbon-Fiber Ceramic performance friction formulation, you can confidently push your rig to the limit and look good doing it with red powder brake calipers. Components are engineered to handle the stress of towing, hauling, mountainous driving, and lifted trucks. Dust-free braking performance. Z36 Carbon-Fiber Ceramic formula provides the extreme braking performance demanded by your truck or 4x\"}, {\"role\": \"assistant\", \"content\": \"$506.98\"}]}\n", + "{\"messages\": [{\"role\": \"system\", \"content\": \"You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\\n\\nKey Amazon pricing factors to evaluate:\\n- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\\n- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\\n- Pack size and quantity (bulk/multi-packs often better per-unit value)\\n- Prime eligibility indicators and fulfillment method signals\\n- Product variations (color, size, model) affecting price tiers\\n- Feature density and specification richness\\n- Amazon's Choice or bestseller indicators in description\\n- Customer rating implications (4.5+ stars = premium pricing power)\\n- Seasonal/trending product indicators\\n\\nAmazon-specific pricing patterns:\\n- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\\n- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\\n- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\\n- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\\n- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\\n- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\\n\\nConsider Amazon's psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\\n\\nOutput format: Respond with only the price including dollar sign and cents (e.g., \\\"$24.99\\\"). No explanations or additional text.\"}, {\"role\": \"user\", \"content\": \"How much does this cost?\\n\\nABBA 36 Gas Cooktop with 5 Sealed Burners - Tempered Glass Surface with SABAF Burners, Natural Gas Stove for Countertop, Home Improvement Essentials, Easy to Clean, 36 x 4.1 x 20.5\\ncooktop Gas powered with 4 fast burners and 1 ultra-fast center burner Tempered glass surface with removable grid for easy cleaning Lightweight for easy installation. Installation Manual Included Counter cutout Dimensions 19 3/8 x 34 1/2 (see diagram) Insured shipping for your satisfaction and peace of mind Brand Name ABBA EST. 1956, Weight 30 pounds, Dimensions 20.5\\\\ D x 36\\\\ W x 4.1\\\\ H, Installation Type Count\"}, {\"role\": \"assistant\", \"content\": \"$405.00\"}]}\n" + ] + } + ], + "source": [ + "print(make_jsonl(train[:3]))" + ] + }, + { + "cell_type": "code", + "execution_count": 13, + "id": "7734bff0-95c4-4e67-a87e-7e2254e2c67d", + "metadata": {}, + "outputs": [], + "source": [ + "# Convert the items into jsonl and write them to a file\n", + "\n", + "def write_jsonl(items, filename):\n", + " with open(filename, \"w\") as f:\n", + " jsonl = make_jsonl(items)\n", + " f.write(jsonl)" + ] + }, + { + "cell_type": "code", + "execution_count": 14, + "id": "393d3ad8-999a-4f99-8c04-339d9166d604", + "metadata": {}, + "outputs": [], + "source": [ + "write_jsonl(fine_tune_train, \"fine_tune_train.jsonl\")" + ] + }, + { + "cell_type": "code", + "execution_count": 15, + "id": "8e23927f-d73e-4668-ac20-abe6f14a56cb", + "metadata": {}, + "outputs": [], + "source": [ + "write_jsonl(fine_tune_validation, \"fine_tune_validation.jsonl\")" + ] + }, + { + "cell_type": "code", + "execution_count": 16, + "id": "d59ad8d2-c61a-448e-b7ed-232f1606970f", + "metadata": {}, + "outputs": [], + "source": [ + "with open(\"fine_tune_train.jsonl\", \"rb\") as f:\n", + " train_file = openai.files.create(file=f, purpose=\"fine-tune\")" + ] + }, + { + "cell_type": "code", + "execution_count": 17, + "id": "083fefba-fd54-47ce-9ff3-aabbc200846f", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "FileObject(id='file-8nECKUWKSHTorbKVrcLxGc', bytes=480142, created_at=1753339843, filename='fine_tune_train.jsonl', object='file', purpose='fine-tune', status='processed', expires_at=None, status_details=None)" + ] + }, + "execution_count": 17, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "train_file" + ] + }, + { + "cell_type": "code", + "execution_count": 18, + "id": "97df3360-0760-4422-a556-5f26d23de6dc", + "metadata": {}, + "outputs": [], + "source": [ + "with open(\"fine_tune_validation.jsonl\", \"rb\") as f:\n", + " validation_file = openai.files.create(file=f, purpose=\"fine-tune\")" + ] + }, + { + "cell_type": "code", + "execution_count": 19, + "id": "a1abb8f3-9e52-4061-970c-fcf399d8ffa3", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "FileObject(id='file-TkMc2SaZVLPmoP37ucXDPH', bytes=119935, created_at=1753339852, filename='fine_tune_validation.jsonl', object='file', purpose='fine-tune', status='processed', expires_at=None, status_details=None)" + ] + }, + "execution_count": 19, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "validation_file" + ] + }, + { + "cell_type": "markdown", + "id": "466052b9-9fb9-48f6-8cf9-c74e6ddc1394", + "metadata": {}, + "source": [ + "# Step 2\n", + "\n", + "I love Weights and Biases - a beautiful, free platform for monitoring training runs. \n", + "Weights and Biases is integrated with OpenAI for fine-tuning.\n", + "\n", + "First set up your weights & biases free account at:\n", + "\n", + "https://wandb.ai\n", + "\n", + "From the Avatar >> Settings menu, near the bottom, you can create an API key.\n", + "\n", + "Then visit the OpenAI dashboard at:\n", + "\n", + "https://platform.openai.com/account/organization\n", + "\n", + "In the integrations section, you can add your Weights & Biases key.\n", + "\n", + "## And now time to Fine-tune!" + ] + }, + { + "cell_type": "code", + "execution_count": 20, + "id": "c7add1a7-a746-4d6e-a5f8-e25629b8b527", + "metadata": {}, + "outputs": [], + "source": [ + "wandb_integration = {\"type\": \"wandb\", \"wandb\": {\"project\": \"gpt-pricer\"}}" + ] + }, + { + "cell_type": "code", + "execution_count": 21, + "id": "49801e69-9277-4deb-9f33-99efb6b45ac2", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'file-8nECKUWKSHTorbKVrcLxGc'" + ] + }, + "execution_count": 21, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "train_file.id" + ] + }, + { + "cell_type": "code", + "execution_count": 23, + "id": "45421b86-5531-4e42-ab19-d6abbb8f4c13", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "FineTuningJob(id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX', created_at=1753341041, error=Error(code=None, message=None, param=None), fine_tuned_model=None, finished_at=None, hyperparameters=Hyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1), model='gpt-4o-mini-2024-07-18', object='fine_tuning.job', organization_id='org-du4k6X6j1Eu0j5xNKaVVIO3O', result_files=[], seed=42, status='validating_files', trained_tokens=None, training_file='file-8nECKUWKSHTorbKVrcLxGc', validation_file='file-TkMc2SaZVLPmoP37ucXDPH', estimated_finish=None, integrations=[FineTuningJobWandbIntegrationObject(type='wandb', wandb=FineTuningJobWandbIntegration(project='gpt-pricer', entity=None, name=None, tags=None, run_id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX'))], metadata=None, method=Method(type='supervised', dpo=None, reinforcement=None, supervised=SupervisedMethod(hyperparameters=SupervisedHyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1))), user_provided_suffix='pricer', usage_metrics=None, shared_with_openai=False, eval_id=None)" + ] + }, + "execution_count": 23, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "openai.fine_tuning.jobs.create(\n", + " training_file=train_file.id,\n", + " validation_file=validation_file.id,\n", + " model=\"gpt-4o-mini-2024-07-18\",\n", + " seed=42,\n", + " hyperparameters={\"n_epochs\": 1},\n", + " integrations = [wandb_integration],\n", + " suffix=\"pricer\"\n", + ")" + ] + }, + { + "cell_type": "code", + "execution_count": 24, + "id": "aeb9de2e-542c-4e83-81c7-b6745133e48b", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "SyncCursorPage[FineTuningJob](data=[FineTuningJob(id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX', created_at=1753341041, error=Error(code=None, message=None, param=None), fine_tuned_model=None, finished_at=None, hyperparameters=Hyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1), model='gpt-4o-mini-2024-07-18', object='fine_tuning.job', organization_id='org-du4k6X6j1Eu0j5xNKaVVIO3O', result_files=[], seed=42, status='validating_files', trained_tokens=None, training_file='file-8nECKUWKSHTorbKVrcLxGc', validation_file='file-TkMc2SaZVLPmoP37ucXDPH', estimated_finish=None, integrations=[FineTuningJobWandbIntegrationObject(type='wandb', wandb=FineTuningJobWandbIntegration(project='gpt-pricer', entity=None, name=None, tags=None, run_id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX'))], metadata=None, method=Method(type='supervised', dpo=None, reinforcement=None, supervised=SupervisedMethod(hyperparameters=SupervisedHyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1))), user_provided_suffix='pricer', usage_metrics=None, shared_with_openai=False, eval_id=None)], has_more=False, object='list')" + ] + }, + "execution_count": 24, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "openai.fine_tuning.jobs.list(limit=1)" + ] + }, + { + "cell_type": "code", + "execution_count": 25, + "id": "40d24873-8ff5-413f-b0d4-8f77c28f18e1", + "metadata": {}, + "outputs": [], + "source": [ + "job_id = openai.fine_tuning.jobs.list(limit=1).data[0].id" + ] + }, + { + "cell_type": "code", + "execution_count": 26, + "id": "a32aef35-4b38-436c-ad00-d082f758efa7", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'ftjob-IBBaP9CY5ovNGnsueXoCLjeX'" + ] + }, + "execution_count": 26, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "job_id" + ] + }, + { + "cell_type": "code", + "execution_count": 27, + "id": "a7e01247-c133-48e1-93d3-c79c399e6178", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "FineTuningJob(id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX', created_at=1753341041, error=Error(code=None, message=None, param=None), fine_tuned_model=None, finished_at=None, hyperparameters=Hyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1), model='gpt-4o-mini-2024-07-18', object='fine_tuning.job', organization_id='org-du4k6X6j1Eu0j5xNKaVVIO3O', result_files=[], seed=42, status='validating_files', trained_tokens=None, training_file='file-8nECKUWKSHTorbKVrcLxGc', validation_file='file-TkMc2SaZVLPmoP37ucXDPH', estimated_finish=None, integrations=[FineTuningJobWandbIntegrationObject(type='wandb', wandb=FineTuningJobWandbIntegration(project='gpt-pricer', entity=None, name=None, tags=None, run_id='ftjob-IBBaP9CY5ovNGnsueXoCLjeX'))], metadata=None, method=Method(type='supervised', dpo=None, reinforcement=None, supervised=SupervisedMethod(hyperparameters=SupervisedHyperparameters(batch_size='auto', learning_rate_multiplier='auto', n_epochs=1))), user_provided_suffix='pricer', usage_metrics=None, shared_with_openai=False, eval_id=None)" + ] + }, + "execution_count": 27, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "openai.fine_tuning.jobs.retrieve(job_id)" + ] + }, + { + "cell_type": "code", + "execution_count": 28, + "id": "0f5150e1-b8de-485f-8eba-cf1e5b00c117", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "[FineTuningJobEvent(id='ftevent-Hen3aW1QhNBeK3fNUmtTnBop', created_at=1753341041, level='info', message='Validating training file: file-8nECKUWKSHTorbKVrcLxGc and validation file: file-TkMc2SaZVLPmoP37ucXDPH', object='fine_tuning.job.event', data={}, type='message'),\n", + " FineTuningJobEvent(id='ftevent-osUdwcFOjzf1HG99p1q4ivBm', created_at=1753341041, level='info', message='Created fine-tuning job: ftjob-IBBaP9CY5ovNGnsueXoCLjeX', object='fine_tuning.job.event', data={}, type='message')]" + ] + }, + "execution_count": 28, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "openai.fine_tuning.jobs.list_events(fine_tuning_job_id=job_id, limit=10).data" + ] + }, + { + "cell_type": "markdown", + "id": "066fef03-8338-4526-9df3-89b649ad4f0a", + "metadata": {}, + "source": [ + "# Step 3\n", + "\n", + "Test our fine tuned model" + ] + }, + { + "cell_type": "code", + "execution_count": 43, + "id": "fa4488cb-3c17-4eda-abd1-53c1c68a491b", + "metadata": {}, + "outputs": [], + "source": [ + "fine_tuned_model_name = openai.fine_tuning.jobs.retrieve(job_id).fine_tuned_model" + ] + }, + { + "cell_type": "code", + "execution_count": 44, + "id": "e9370937-5a6f-4724-8265-b208663b4450", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "'ft:gpt-4o-mini-2024-07-18:aaron:pricer:BwkX85YV'" + ] + }, + "execution_count": 44, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "fine_tuned_model_name" + ] + }, + { + "cell_type": "code", + "execution_count": 31, + "id": "66ea68e8-ab1b-4f0d-aba4-a59574d8f85e", + "metadata": {}, + "outputs": [], + "source": [ + "def messages_for(item):\n", + " system_message = \"\"\"You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\n", + "\n", + "Key Amazon pricing factors to evaluate:\n", + "- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\n", + "- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\n", + "- Pack size and quantity (bulk/multi-packs often better per-unit value)\n", + "- Prime eligibility indicators and fulfillment method signals\n", + "- Product variations (color, size, model) affecting price tiers\n", + "- Feature density and specification richness\n", + "- Amazon's Choice or bestseller indicators in description\n", + "- Customer rating implications (4.5+ stars = premium pricing power)\n", + "- Seasonal/trending product indicators\n", + "\n", + "Amazon-specific pricing patterns:\n", + "- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\n", + "- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\n", + "- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\n", + "- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\n", + "- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\n", + "- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\n", + "\n", + "Consider Amazon's psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\n", + "\n", + "Output format: Respond with only the price including dollar sign and cents (e.g., \"$24.99\"). No explanations or additional text.\"\"\"\n", + " \n", + " user_prompt = item.test_prompt().replace(\" to the nearest dollar\",\"\").replace(\"\\n\\nPrice is $\",\"\")\n", + " \n", + " return [\n", + " {\"role\": \"system\", \"content\": system_message},\n", + " {\"role\": \"user\", \"content\": user_prompt},\n", + " {\"role\": \"assistant\", \"content\": f\"${item.price:.2f}\"}\n", + " ]" + ] + }, + { + "cell_type": "code", + "execution_count": 32, + "id": "4ff92d61-0d27-4b0d-8b32-c9891016509b", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "[{'role': 'system',\n", + " 'content': 'You are an Amazon marketplace pricing expert. Analyze product descriptions to predict accurate Amazon selling prices based on typical marketplace dynamics and consumer behavior patterns.\\n\\nKey Amazon pricing factors to evaluate:\\n- Brand strength (Apple, Samsung, Nike = premium; generic/unbranded = budget)\\n- Product category positioning (Home & Kitchen, Electronics, Sports, Beauty, etc.)\\n- Pack size and quantity (bulk/multi-packs often better per-unit value)\\n- Prime eligibility indicators and fulfillment method signals\\n- Product variations (color, size, model) affecting price tiers\\n- Feature density and specification richness\\n- Amazon\\'s Choice or bestseller indicators in description\\n- Customer rating implications (4.5+ stars = premium pricing power)\\n- Seasonal/trending product indicators\\n\\nAmazon-specific pricing patterns:\\n- Electronics: $10-50 (accessories), $50-200 (mid-tier), $200+ (premium)\\n- Home/Kitchen: $15-40 (small items), $40-150 (appliances), $150+ (major items)\\n- Beauty/Personal Care: $8-25 (drugstore), $25-60 (prestige), $60+ (luxury)\\n- Sports/Outdoors: $20-80 (equipment), $30-120 (apparel), $100+ (specialized gear)\\n- Books/Media: $10-20 (paperback), $15-35 (hardcover), $25-50 (specialty)\\n- Toys/Games: $15-40 (standard), $40-100 (premium/electronic)\\n\\nConsider Amazon\\'s psychological pricing (ends in .99, .95, .49) and competitive marketplace pressure.\\n\\nOutput format: Respond with only the price including dollar sign and cents (e.g., \"$24.99\"). No explanations or additional text.'},\n", + " {'role': 'user',\n", + " 'content': \"How much does this cost?\\n\\nOEM AC Compressor w/A/C Repair Kit For Ford F150 F-150 V8 & Lincoln Mark LT 2007 2008 - BuyAutoParts NEW\\nAs one of the world's largest automotive parts suppliers, our parts are trusted every day by mechanics and vehicle owners worldwide. This A/C Compressor and Components Kit is manufactured and tested to the strictest OE standards for unparalleled performance. Built for trouble-free ownership and 100% visually inspected and quality tested, this A/C Compressor and Components Kit is backed by our 100% satisfaction guarantee. Guaranteed Exact Fit for easy installation 100% BRAND NEW, premium ISO/TS 16949 quality - tested to meet or exceed OEM specifications Engineered for superior durability, backed by industry-leading unlimited-mileage warranty Included in this K\"},\n", + " {'role': 'assistant', 'content': '$374.41'}]" + ] + }, + "execution_count": 32, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "# Try this out\n", + "\n", + "messages_for(test[0])" + ] + }, + { + "cell_type": "code", + "execution_count": 38, + "id": "b1af1888-f94a-4106-b0d8-8a70939eec4e", + "metadata": {}, + "outputs": [], + "source": [ + "# A utility function to extract the price from a string\n", + "\n", + "def get_price(s):\n", + " s = s.replace('$','').replace(',','')\n", + " match = re.search(r\"[-+]?\\d*\\.\\d+|\\d+\", s)\n", + " return float(match.group()) if match else 0" + ] + }, + { + "cell_type": "code", + "execution_count": 39, + "id": "f138c5b7-bcc1-4085-aced-68dad1bf36b4", + "metadata": {}, + "outputs": [ + { + "data": { + "text/plain": [ + "99.99" + ] + }, + "execution_count": 39, + "metadata": {}, + "output_type": "execute_result" + } + ], + "source": [ + "get_price(\"The price is roughly $99.99 because blah blah\")" + ] + }, + { + "cell_type": "code", + "execution_count": 45, + "id": "501a2a7a-69c8-451b-bbc0-398bcb9e1612", + "metadata": {}, + "outputs": [], + "source": [ + "# The function for gpt-4o-mini\n", + "\n", + "def gpt_fine_tuned(item):\n", + " response = openai.chat.completions.create(\n", + " model=fine_tuned_model_name, \n", + " messages=messages_for(item),\n", + " seed=42,\n", + " max_tokens=7\n", + " )\n", + " reply = response.choices[0].message.content\n", + " return get_price(reply)" + ] + }, + { + "cell_type": "code", + "execution_count": 46, + "id": "843d88b4-364a-431b-b48b-8a7c1f68b786", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "374.41\n", + "400.06\n" + ] + } + ], + "source": [ + "print(test[0].price)\n", + "print(gpt_fine_tuned(test[0]))" + ] + }, + { + "cell_type": "code", + "execution_count": 47, + "id": "edd7ada0-15b7-42ec-bbbb-1250e0eb9af1", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "How much does this cost to the nearest dollar?\n", + "\n", + "OEM AC Compressor w/A/C Repair Kit For Ford F150 F-150 V8 & Lincoln Mark LT 2007 2008 - BuyAutoParts NEW\n", + "As one of the world's largest automotive parts suppliers, our parts are trusted every day by mechanics and vehicle owners worldwide. This A/C Compressor and Components Kit is manufactured and tested to the strictest OE standards for unparalleled performance. Built for trouble-free ownership and 100% visually inspected and quality tested, this A/C Compressor and Components Kit is backed by our 100% satisfaction guarantee. Guaranteed Exact Fit for easy installation 100% BRAND NEW, premium ISO/TS 16949 quality - tested to meet or exceed OEM specifications Engineered for superior durability, backed by industry-leading unlimited-mileage warranty Included in this K\n", + "\n", + "Price is $\n" + ] + } + ], + "source": [ + "print(test[0].test_prompt())" + ] + }, + { + "cell_type": "code", + "execution_count": 53, + "id": "36bdd2c9-1859-4f99-a09f-3ec83b845b30", + "metadata": {}, + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Testing Gpt Fine Tuned on 250 samples...\n", + "Thresholds: Excellent ≤15%, Good ≤30%\n", + "--------------------------------------------------------------------------------\n", + "\u001b[94m1: Guess: $374.41 Truth: $374.41 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: OEM AC Compressor w/A/C Repair Kit For F...\u001b[0m\n", + "\u001b[95m2: Guess: $161.72 Truth: $225.11 Abs Error: $63.39 % Error: 28.2% SLE: 0.11 Item: Motorcraft YB3125 Fan Clutch\u001b[0m\n", + "\u001b[94m3: Guess: $61.68 Truth: $61.68 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dorman 603-159 Front Washer Fluid Reserv...\u001b[0m\n", + "\u001b[94m4: Guess: $599.99 Truth: $599.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: HP Premium 17.3-inch HD Plus Touchscreen...\u001b[0m\n", + "\u001b[94m5: Guess: $16.99 Truth: $16.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: 5-Position Super Switch Pickup Selector ...\u001b[0m\n", + "\u001b[94m6: Guess: $33.00 Truth: $31.99 Abs Error: $1.01 % Error: 3.2% SLE: 0.00 Item: Horror Bookmarks, Resin Horror Bookmarks...\u001b[0m\n", + "\u001b[94m7: Guess: $101.79 Truth: $101.79 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SK6241 - Stinger 4 Gauge 6000 Series Pow...\u001b[0m\n", + "\u001b[94m8: Guess: $289.00 Truth: $289.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Godox ML60Bi LED Light Kit, Handheld LED...\u001b[0m\n", + "\u001b[94m9: Guess: $629.99 Truth: $635.86 Abs Error: $5.87 % Error: 0.9% SLE: 0.00 Item: Randall RG75DG3PLUS G3 Plus 100-Watt Com...\u001b[0m\n", + "\u001b[94m10: Guess: $61.99 Truth: $65.99 Abs Error: $4.00 % Error: 6.1% SLE: 0.00 Item: HOLDWILL 6 Pack LED Shop Light, 4FT 24W ...\u001b[0m\n", + "\u001b[94m11: Guess: $228.06 Truth: $254.21 Abs Error: $26.15 % Error: 10.3% SLE: 0.01 Item: Viking Horns V103C/1005ATK 3 Gallon Air ...\u001b[0m\n", + "\u001b[94m12: Guess: $412.99 Truth: $412.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: CURT 70110 Custom Tow Bar Base Plate Bra...\u001b[0m\n", + "\u001b[94m13: Guess: $210.57 Truth: $205.50 Abs Error: $5.07 % Error: 2.5% SLE: 0.00 Item: 10-Pack Solar HAMMERED BRONZE Finish Pos...\u001b[0m\n", + "\u001b[94m14: Guess: $248.23 Truth: $248.23 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: COSTWAY Electric Tumble Dryer, Sliver\u001b[0m\n", + "\u001b[94m15: Guess: $399.00 Truth: $399.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: FREE SIGNAL TV Transit 32\" 12 Volt DC Po...\u001b[0m\n", + "\u001b[94m16: Guess: $371.96 Truth: $373.94 Abs Error: $1.98 % Error: 0.5% SLE: 0.00 Item: Bilstein 5100 Monotube Gas Shock Set com...\u001b[0m\n", + "\u001b[94m17: Guess: $81.99 Truth: $92.89 Abs Error: $10.90 % Error: 11.7% SLE: 0.02 Item: Sangean K-200 Multi-Function Upright AM/...\u001b[0m\n", + "\u001b[94m18: Guess: $51.99 Truth: $51.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Charles Leonard Magnetic Lapboard Class ...\u001b[0m\n", + "\u001b[94m19: Guess: $179.00 Truth: $179.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Gigabyte AMD Radeon HD 7870 2 GB GDDR5 D...\u001b[0m\n", + "\u001b[94m20: Guess: $19.00 Truth: $19.42 Abs Error: $0.42 % Error: 2.2% SLE: 0.00 Item: 3dRose LLC 8 x 8 x 0.25 Inches Bull Terr...\u001b[0m\n", + "\u001b[94m21: Guess: $539.95 Truth: $539.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ROKINON 85mm F1.4 Auto Focus Full Frame ...\u001b[0m\n", + "\u001b[94m22: Guess: $147.67 Truth: $147.67 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: AUTOSAVER88 Headlight Assembly Compatibl...\u001b[0m\n", + "\u001b[91m23: Guess: $47.97 Truth: $24.99 Abs Error: $22.98 % Error: 92.0% SLE: 0.40 Item: ASI NAUTICAL 2.5 Inches Opera Glasses Bi...\u001b[0m\n", + "\u001b[94m24: Guess: $149.00 Truth: $149.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Behringer TUBE OVERDRIVE TO100 Authentic...\u001b[0m\n", + "\u001b[94m25: Guess: $16.99 Truth: $16.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Fun Express Insect Finger Puppets - 24 f...\u001b[0m\n", + "\u001b[94m26: Guess: $7.99 Truth: $7.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: WAFJAMF Roller Stamp Identity Theft Stam...\u001b[0m\n", + "\u001b[94m27: Guess: $199.99 Truth: $199.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Capulina Tiffany Floor Lamp 2-Light 16\" ...\u001b[0m\n", + "\u001b[94m28: Guess: $251.45 Truth: $251.45 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Apple Watch Series 6 (GPS, 44mm) - Space...\u001b[0m\n", + "\u001b[94m29: Guess: $231.62 Truth: $231.62 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ICON 01725 Tandem Axle Fender Skirt FS17...\u001b[0m\n", + "\u001b[94m30: Guess: $135.00 Truth: $135.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SanDisk 128GB Ultra (10 Pack) MicroSD Cl...\u001b[0m\n", + "\u001b[94m31: Guess: $356.62 Truth: $356.62 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Velvac 2020,L,C/Hr,W,E2003,102\",Bk - 715...\u001b[0m\n", + "\u001b[94m32: Guess: $257.99 Truth: $257.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: TCMT Passenger Backrest Sissy Bar & Lugg...\u001b[0m\n", + "\u001b[94m33: Guess: $27.99 Truth: $27.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Alnicov 63.5MM Brass Tremolo Block，Tremo...\u001b[0m\n", + "\u001b[94m34: Guess: $171.20 Truth: $171.20 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Subaru Forester Outback Legacy OEM Engin...\u001b[0m\n", + "\u001b[94m35: Guess: $225.00 Truth: $225.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Richmond Auto Upholstery - 2012 Dodge Ra...\u001b[0m\n", + "\u001b[94m36: Guess: $105.00 Truth: $105.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: AP-39 Automotive Paint Primer Grey 2K Ur...\u001b[0m\n", + "\u001b[94m37: Guess: $299.99 Truth: $299.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Road Top Wireless Carplay Retrofit Kit D...\u001b[0m\n", + "\u001b[94m38: Guess: $535.09 Truth: $535.09 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Gibson Performance Exhaust 5658 Aluminiz...\u001b[0m\n", + "\u001b[94m39: Guess: $12.33 Truth: $12.33 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Bella Tunno Happy Links - Baby Montessor...\u001b[0m\n", + "\u001b[94m40: Guess: $81.99 Truth: $84.99 Abs Error: $3.00 % Error: 3.5% SLE: 0.00 Item: CANMORE H300 Handheld GPS Golf Device, S...\u001b[0m\n", + "\u001b[94m41: Guess: $15.99 Truth: $15.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: DCPOWER AC Adapter Compatible Replacemen...\u001b[0m\n", + "\u001b[94m42: Guess: $62.44 Truth: $62.44 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Sharp, VX2128V, Commercial Desktop Calcu...\u001b[0m\n", + "\u001b[94m43: Guess: $82.99 Truth: $82.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Melissa & Doug Lifelike Plush Stork Gian...\u001b[0m\n", + "\u001b[94m44: Guess: $599.95 Truth: $599.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Sony SSCS8 2-Way 3-Driver Center Channel...\u001b[0m\n", + "\u001b[94m45: Guess: $184.99 Truth: $194.99 Abs Error: $10.00 % Error: 5.1% SLE: 0.00 Item: ASUS Chromebook CX1, 14\" Full HD NanoEdg...\u001b[0m\n", + "\u001b[94m46: Guess: $344.99 Truth: $344.95 Abs Error: $0.04 % Error: 0.0% SLE: 0.00 Item: FiiO X7 32GB Hi-Res Lossless Music Playe...\u001b[0m\n", + "\u001b[94m47: Guess: $37.99 Truth: $37.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: TORRO Leather Case Compatible with iPhon...\u001b[0m\n", + "\u001b[94m48: Guess: $228.22 Truth: $224.35 Abs Error: $3.87 % Error: 1.7% SLE: 0.00 Item: Universal Air Conditioner KT 1031 A/C Co...\u001b[0m\n", + "\u001b[94m49: Guess: $814.00 Truth: $814.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Street Series Stainless Performance Cat-...\u001b[0m\n", + "\u001b[94m50: Guess: $399.99 Truth: $439.88 Abs Error: $39.89 % Error: 9.1% SLE: 0.01 Item: Lenovo IdeaPad 3 14-inch Laptop, 14.0-in...\u001b[0m\n", + "\u001b[94m51: Guess: $341.43 Truth: $341.43 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Access Bed Covers TonnoSport 22050219 - ...\u001b[0m\n", + "\u001b[94m52: Guess: $46.78 Truth: $46.78 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: G.I. JOE Hasbro 3 3/4\" Wave 5 Action Fig...\u001b[0m\n", + "\u001b[94m53: Guess: $171.44 Truth: $171.44 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: T&S Brass B-0232-BST Double Pantry Fauce...\u001b[0m\n", + "\u001b[94m54: Guess: $458.00 Truth: $458.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ZTUOAUMA Fuel Injection Pump 3090942 309...\u001b[0m\n", + "\u001b[94m55: Guess: $130.00 Truth: $130.75 Abs Error: $0.75 % Error: 0.6% SLE: 0.00 Item: 2AP18AA#ABA Hp Prime Graphing Calculator...\u001b[0m\n", + "\u001b[94m56: Guess: $81.73 Truth: $83.81 Abs Error: $2.08 % Error: 2.5% SLE: 0.00 Item: Lowrance 000-0119-83 Nmea 2000 25' Exten...\u001b[0m\n", + "\u001b[91m57: Guess: $47.97 Truth: $386.39 Abs Error: $338.42 % Error: 87.6% SLE: 4.28 Item: Jeep Genuine Accessories 82213051 Hood L...\u001b[0m\n", + "\u001b[94m58: Guess: $169.00 Truth: $169.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: GODOX CB-06 Hard Carrying Case with Whee...\u001b[0m\n", + "\u001b[94m59: Guess: $17.95 Truth: $17.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Au-Tomotive Gold, INC. Ford Black Valet ...\u001b[0m\n", + "\u001b[94m60: Guess: $269.00 Truth: $269.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Snailfly Black Roof Rack Rail + Cross Ba...\u001b[0m\n", + "\u001b[94m61: Guess: $77.77 Truth: $77.77 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: KING SHA Anti Glare LED Track Lighting H...\u001b[0m\n", + "\u001b[94m62: Guess: $81.99 Truth: $88.99 Abs Error: $7.00 % Error: 7.9% SLE: 0.01 Item: APS Compatible with Chevy Silverado 1500...\u001b[0m\n", + "\u001b[94m63: Guess: $364.41 Truth: $364.41 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Wilwood Engineering 14011291R Brake Cali...\u001b[0m\n", + "\u001b[94m64: Guess: $127.03 Truth: $127.03 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ACDelco Gold 336-1925A Starter, Remanufa...\u001b[0m\n", + "\u001b[94m65: Guess: $778.95 Truth: $778.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: UWS EC10783 69-Inch Matte Black Heavy-Wa...\u001b[0m\n", + "\u001b[94m66: Guess: $206.66 Truth: $206.66 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dell Latitude E5440 14in Business Laptop...\u001b[0m\n", + "\u001b[94m67: Guess: $35.94 Truth: $35.94 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: (Plug and Play) Spare Tire Brake Light W...\u001b[0m\n", + "\u001b[94m68: Guess: $149.00 Truth: $149.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: The Ultimate Roadside Rescue Assistant\u001b[0m\n", + "\u001b[94m69: Guess: $251.98 Truth: $251.98 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Brand New 18\" x 8.5\" Replacement Wheel f...\u001b[0m\n", + "\u001b[94m70: Guess: $160.00 Truth: $160.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Headlight Headlamp LH Left & RH Right Pa...\u001b[0m\n", + "\u001b[91m71: Guess: $64.90 Truth: $39.99 Abs Error: $24.91 % Error: 62.3% SLE: 0.23 Item: Lilo And Stitch Deluxe Oversize Print La...\u001b[0m\n", + "\u001b[94m72: Guess: $362.41 Truth: $362.41 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: AC Compressor & A/C Clutch For Hyundai A...\u001b[0m\n", + "\u001b[94m73: Guess: $344.00 Truth: $344.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: House Of Troy PIN475-AB Pinnacle Collect...\u001b[0m\n", + "\u001b[94m74: Guess: $22.99 Truth: $25.09 Abs Error: $2.10 % Error: 8.4% SLE: 0.01 Item: Juno T29 WH Floating Electrical Feed Sin...\u001b[0m\n", + "\u001b[94m75: Guess: $175.95 Truth: $175.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Sherman GO-PARTS - for 2013-2016 Toyota ...\u001b[0m\n", + "\u001b[94m76: Guess: $132.64 Truth: $132.64 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Roland RPU-3 Electronic Keyboard Pedal o...\u001b[0m\n", + "\u001b[94m77: Guess: $422.99 Truth: $422.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Rockland VMI14 12,000 Pound 12 Volt DC E...\u001b[0m\n", + "\u001b[94m78: Guess: $146.48 Truth: $146.48 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Max Advanced Brakes Elite XDS Front Cros...\u001b[0m\n", + "\u001b[94m79: Guess: $156.83 Truth: $156.83 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Quality-Built 11030 Premium Quality Alte...\u001b[0m\n", + "\u001b[94m80: Guess: $251.99 Truth: $251.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Lucida LG-510 Student Classical Guitar, ...\u001b[0m\n", + "\u001b[94m81: Guess: $940.33 Truth: $940.33 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Longacre 52-79800 Aluminum Turn Plates\u001b[0m\n", + "\u001b[95m82: Guess: $64.22 Truth: $52.99 Abs Error: $11.23 % Error: 21.2% SLE: 0.04 Item: Motion Pro 08-0380 Adjustable Torque Wre...\u001b[0m\n", + "\u001b[94m83: Guess: $219.99 Truth: $219.95 Abs Error: $0.04 % Error: 0.0% SLE: 0.00 Item: Glyph Thunderbolt 3 NVMe Dock (0 GB)\u001b[0m\n", + "\u001b[94m84: Guess: $441.03 Truth: $441.03 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: TOYO Open Country MT Performance Radial ...\u001b[0m\n", + "\u001b[94m85: Guess: $168.00 Truth: $168.98 Abs Error: $0.98 % Error: 0.6% SLE: 0.00 Item: Razer Seiren X USB Streaming Microphone ...\u001b[0m\n", + "\u001b[95m86: Guess: $3.09 Truth: $2.49 Abs Error: $0.60 % Error: 24.1% SLE: 0.03 Item: Happy Birthday to Dad From Your Daughter...\u001b[0m\n", + "\u001b[94m87: Guess: $97.99 Truth: $98.62 Abs Error: $0.63 % Error: 0.6% SLE: 0.00 Item: Little Tikes My Real Jam First Concert S...\u001b[0m\n", + "\u001b[91m88: Guess: $156.97 Truth: $256.95 Abs Error: $99.98 % Error: 38.9% SLE: 0.24 Item: Studio M Peace and Harmony Art Pole Comm...\u001b[0m\n", + "\u001b[94m89: Guess: $30.99 Truth: $30.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: MyVolts 12V Power Supply Adaptor Compati...\u001b[0m\n", + "\u001b[94m90: Guess: $569.84 Truth: $569.84 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dell Latitude 7212 Rugged Extreme Tablet...\u001b[0m\n", + "\u001b[94m91: Guess: $177.99 Truth: $177.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Covermates Contour Fit Car Cover - Light...\u001b[0m\n", + "\u001b[94m92: Guess: $992.99 Truth: $997.99 Abs Error: $5.00 % Error: 0.5% SLE: 0.00 Item: Westin 57-4025 Black HDX Grille Guard fi...\u001b[0m\n", + "\u001b[94m93: Guess: $219.00 Truth: $219.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Fieldpiece JL2 Job Link Wireless App Tra...\u001b[0m\n", + "\u001b[94m94: Guess: $210.57 Truth: $225.55 Abs Error: $14.98 % Error: 6.6% SLE: 0.00 Item: hansgrohe Talis S Modern Premium Easy Cl...\u001b[0m\n", + "\u001b[94m95: Guess: $495.99 Truth: $495.95 Abs Error: $0.04 % Error: 0.0% SLE: 0.00 Item: G-Technology G-SPEED eS PRO High-Perform...\u001b[0m\n", + "\u001b[94m96: Guess: $942.37 Truth: $942.37 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: DreamLine SHDR-1960723L-01 Shower Door, ...\u001b[0m\n", + "\u001b[94m97: Guess: $1.99 Truth: $1.94 Abs Error: $0.05 % Error: 2.6% SLE: 0.00 Item: Sanctuary Square Backplate Finish: Oiled...\u001b[0m\n", + "\u001b[94m98: Guess: $284.34 Truth: $284.34 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Pelican Protector 1750 Long Case - Multi...\u001b[0m\n", + "\u001b[94m99: Guess: $171.90 Truth: $171.90 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Brock Replacement Driver and Passenger H...\u001b[0m\n", + "\u001b[95m100: Guess: $169.97 Truth: $144.99 Abs Error: $24.98 % Error: 17.2% SLE: 0.02 Item: Carlinkit Ai Box Mini, Android 11, Multi...\u001b[0m\n", + "\u001b[94m101: Guess: $470.47 Truth: $470.47 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: StarDot NetCamLIVE2 YouTube Live Stream ...\u001b[0m\n", + "\u001b[94m102: Guess: $66.95 Truth: $66.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Atomic Compatible FILXXCAR0016 16x25x5 M...\u001b[0m\n", + "\u001b[94m103: Guess: $130.97 Truth: $117.00 Abs Error: $13.97 % Error: 11.9% SLE: 0.01 Item: Bandai Awakening of S. H. s.h.figuarts s...\u001b[0m\n", + "\u001b[94m104: Guess: $172.14 Truth: $172.14 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Fit System 62135G Passenger Side Towing ...\u001b[0m\n", + "\u001b[94m105: Guess: $392.74 Truth: $392.74 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Black Horse Black Aluminum Exceed Runnin...\u001b[0m\n", + "\u001b[94m106: Guess: $16.99 Truth: $16.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dearsun Twinkle Star Color Night Light P...\u001b[0m\n", + "\u001b[91m107: Guess: $1.99 Truth: $1.34 Abs Error: $0.65 % Error: 48.5% SLE: 0.06 Item: Pokemon - Gallade Spirit Link (83/108) -...\u001b[0m\n", + "\u001b[94m108: Guess: $349.00 Truth: $349.98 Abs Error: $0.98 % Error: 0.3% SLE: 0.00 Item: Ibanez GA34STCE-NT GIO Series Classical ...\u001b[0m\n", + "\u001b[94m109: Guess: $370.71 Truth: $370.71 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Set 2 Heavy Duty 12-16.5 12x16.5 12 Ply ...\u001b[0m\n", + "\u001b[94m110: Guess: $57.99 Truth: $65.88 Abs Error: $7.89 % Error: 12.0% SLE: 0.02 Item: Hairpin Table Legs 28\" Heavy Duty Hairpi...\u001b[0m\n", + "\u001b[94m111: Guess: $229.99 Truth: $229.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Marada Racing Seat with Adjustable Slide...\u001b[0m\n", + "\u001b[91m112: Guess: $22.57 Truth: $9.14 Abs Error: $13.43 % Error: 146.9% SLE: 0.71 Item: Remington Industries 24UL1007STRWHI25 24...\u001b[0m\n", + "\u001b[94m113: Guess: $199.00 Truth: $199.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Acer S3-391-6046 13.3-inch Ultrabook, In...\u001b[0m\n", + "\u001b[94m114: Guess: $109.99 Truth: $109.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ICBEAMER 7\" RGB LED Headlights Bulb Halo...\u001b[0m\n", + "\u001b[94m115: Guess: $570.42 Truth: $570.42 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: R1 Concepts Front Rear Brakes and Rotors...\u001b[0m\n", + "\u001b[94m116: Guess: $279.99 Truth: $279.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Camplux 2.64 GPM Tankless , Outdoor Port...\u001b[0m\n", + "\u001b[94m117: Guess: $30.99 Truth: $30.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: KNOKLOCK 10 Pack 3.75 Inch(96mm) Kitchen...\u001b[0m\n", + "\u001b[94m118: Guess: $31.99 Truth: $31.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Valley Enterprises Yaesu USB FTDI CT-62 ...\u001b[0m\n", + "\u001b[94m119: Guess: $15.90 Truth: $15.90 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: G9 LED Light Bulbs，8W，75W 100W replaceme...\u001b[0m\n", + "\u001b[94m120: Guess: $45.99 Truth: $45.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ZCHAOZ 4 Lights Antique White Farmhouse ...\u001b[0m\n", + "\u001b[94m121: Guess: $113.52 Truth: $113.52 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Honeywell TH8320R1003 Honeywell VisionPr...\u001b[0m\n", + "\u001b[94m122: Guess: $516.99 Truth: $516.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Patriot Exhaust H8013-1 1-7/8\" Clippster...\u001b[0m\n", + "\u001b[95m123: Guess: $166.22 Truth: $196.99 Abs Error: $30.77 % Error: 15.6% SLE: 0.03 Item: Fitrite Autopart New Front Left Driver S...\u001b[0m\n", + "\u001b[94m124: Guess: $46.99 Truth: $46.55 Abs Error: $0.44 % Error: 0.9% SLE: 0.00 Item: Technical Precision Replacement for GE G...\u001b[0m\n", + "\u001b[94m125: Guess: $356.99 Truth: $356.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Covercraft Carhartt SeatSaver Front Row ...\u001b[0m\n", + "\u001b[94m126: Guess: $319.95 Truth: $319.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Sennheiser SD Pro 2 (506008) - Double-Si...\u001b[0m\n", + "\u001b[94m127: Guess: $96.06 Truth: $96.06 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Hitachi MAF0110 Mass Air Flow Sensor\u001b[0m\n", + "\u001b[94m128: Guess: $190.99 Truth: $190.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: AmScope SE305R-P-LED-PS36A 10X-30X LED C...\u001b[0m\n", + "\u001b[94m129: Guess: $257.99 Truth: $257.95 Abs Error: $0.04 % Error: 0.0% SLE: 0.00 Item: Front Left Driver Side Window Regulator ...\u001b[0m\n", + "\u001b[94m130: Guess: $57.95 Truth: $62.95 Abs Error: $5.00 % Error: 7.9% SLE: 0.01 Item: Premium Replica Hubcap Set, Fits Nissan ...\u001b[0m\n", + "\u001b[94m131: Guess: $47.66 Truth: $47.66 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Excellerations Phonics Spelling Game for...\u001b[0m\n", + "\u001b[94m132: Guess: $226.99 Truth: $226.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: RC4WD BigDog Dual Axle Scale Car/Truck T...\u001b[0m\n", + "\u001b[94m133: Guess: $359.95 Truth: $359.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Unknown Stage 2 Clutch Kit - Low Altitud...\u001b[0m\n", + "\u001b[94m134: Guess: $78.40 Truth: $78.40 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: 2002-2008 Dodge Ram 1500 Mopar 4X4 Emble...\u001b[0m\n", + "\u001b[94m135: Guess: $172.77 Truth: $172.77 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Pro Comp Alloys Series 89 Wheel with Pol...\u001b[0m\n", + "\u001b[94m136: Guess: $316.45 Truth: $316.45 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Detroit Axle - Front Rear Strut & Coil S...\u001b[0m\n", + "\u001b[94m137: Guess: $87.99 Truth: $87.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ECCPP Rear Wheel Axle Replacement fit fo...\u001b[0m\n", + "\u001b[94m138: Guess: $226.63 Truth: $226.63 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dell Latitude E6520 Intel i7-2720QM 2.20...\u001b[0m\n", + "\u001b[94m139: Guess: $31.49 Truth: $31.49 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: F FIERCE CYCLE 251pcs Black Universal Mo...\u001b[0m\n", + "\u001b[94m140: Guess: $196.00 Truth: $196.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Flash Furniture 4 Pk. HERCULES Series 88...\u001b[0m\n", + "\u001b[95m141: Guess: $57.99 Truth: $78.40 Abs Error: $20.41 % Error: 26.0% SLE: 0.09 Item: B&M 30287 Throttle Valve/Kickdown Cable,...\u001b[0m\n", + "\u001b[94m142: Guess: $116.25 Truth: $116.25 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Gates TCK226 PowerGrip Premium Timing Be...\u001b[0m\n", + "\u001b[94m143: Guess: $112.78 Truth: $112.78 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Monroe Shocks & Struts Quick-Strut 17149...\u001b[0m\n", + "\u001b[95m144: Guess: $22.47 Truth: $27.32 Abs Error: $4.85 % Error: 17.8% SLE: 0.04 Item: Feit Electric BPMR16/GU10/930CA/6 35W EQ...\u001b[0m\n", + "\u001b[94m145: Guess: $145.91 Truth: $145.91 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Yellow Jacket 2806 Contractor Extension ...\u001b[0m\n", + "\u001b[94m146: Guess: $171.09 Truth: $171.09 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Garage-Pro Tailgate SET Compatible with ...\u001b[0m\n", + "\u001b[94m147: Guess: $169.97 Truth: $167.95 Abs Error: $2.02 % Error: 1.2% SLE: 0.00 Item: 3M Perfect It Buffing and Polishing Kit ...\u001b[0m\n", + "\u001b[91m148: Guess: $57.99 Truth: $28.49 Abs Error: $29.50 % Error: 103.5% SLE: 0.48 Item: Chinese Style Dollhouse Model DIY Miniat...\u001b[0m\n", + "\u001b[94m149: Guess: $122.23 Truth: $122.23 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Generic NRG Innovations SRK-161H Steerin...\u001b[0m\n", + "\u001b[91m150: Guess: $57.99 Truth: $32.99 Abs Error: $25.00 % Error: 75.8% SLE: 0.30 Item: Learning Resources Coding Critters Range...\u001b[0m\n", + "\u001b[94m151: Guess: $81.47 Truth: $71.20 Abs Error: $10.27 % Error: 14.4% SLE: 0.02 Item: Bosch Automotive 15463 Oxygen Sensor, OE...\u001b[0m\n", + "\u001b[94m152: Guess: $112.75 Truth: $112.75 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Case of 24-2 Inch Blue Painters Tape - 6...\u001b[0m\n", + "\u001b[94m153: Guess: $142.43 Truth: $142.43 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: MOCA Engine Water Pump & Fan Clutch fit ...\u001b[0m\n", + "\u001b[94m154: Guess: $398.99 Truth: $398.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SAREMAS Foot Step Bars for Hyundai Palis...\u001b[0m\n", + "\u001b[94m155: Guess: $449.00 Truth: $449.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Gretsch G9210 Square Neck Boxcar Mahogan...\u001b[0m\n", + "\u001b[94m156: Guess: $169.97 Truth: $189.00 Abs Error: $19.03 % Error: 10.1% SLE: 0.01 Item: NikoMaku Mirror Dash Cam Front and Rear ...\u001b[0m\n", + "\u001b[94m157: Guess: $120.91 Truth: $120.91 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Fenix HP25R v2.0 USB-C Rechargeable Head...\u001b[0m\n", + "\u001b[94m158: Guess: $203.53 Truth: $203.53 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: R&L Racing Heavy Duty Roll-Up Soft Tonne...\u001b[0m\n", + "\u001b[94m159: Guess: $399.99 Truth: $349.99 Abs Error: $50.00 % Error: 14.3% SLE: 0.02 Item: Garmin 010-02258-10 GPSMAP 64sx, Handhel...\u001b[0m\n", + "\u001b[91m160: Guess: $22.99 Truth: $34.35 Abs Error: $11.36 % Error: 33.1% SLE: 0.15 Item: Brown 5-7/8\" X 8-1/2\" X 3/16\" Thick Heav...\u001b[0m\n", + "\u001b[94m161: Guess: $384.99 Truth: $384.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: GAOMON PD2200 Pen Display & 20 Pen Nibs ...\u001b[0m\n", + "\u001b[94m162: Guess: $211.00 Truth: $211.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: VXMOTOR for 97-03 Ford F150/F250 Lightdu...\u001b[0m\n", + "\u001b[94m163: Guess: $129.00 Truth: $129.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: HP EliteBook 2540p Intel Core i7-640LM X...\u001b[0m\n", + "\u001b[94m164: Guess: $111.45 Truth: $111.45 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Green EPX Mixing Nozzles 100-Pack-fits 3...\u001b[0m\n", + "\u001b[94m165: Guess: $81.12 Truth: $81.12 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Box Partners 6 1/4 x 3 1/8\" 13 Pt. Manil...\u001b[0m\n", + "\u001b[94m166: Guess: $457.08 Truth: $457.08 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Vixen Air 1/2\" NPT Air Ride Suspension H...\u001b[0m\n", + "\u001b[94m167: Guess: $49.99 Truth: $49.49 Abs Error: $0.50 % Error: 1.0% SLE: 0.00 Item: Smart Floor Lamp, 2700-6500K+RGBPink Mul...\u001b[0m\n", + "\u001b[94m168: Guess: $80.56 Truth: $80.56 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SOZG 324mm Wheelbase Body Shell RC Car B...\u001b[0m\n", + "\u001b[94m169: Guess: $278.39 Truth: $278.39 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Mickey Thompson ET Street S/S Racing Rad...\u001b[0m\n", + "\u001b[94m170: Guess: $364.50 Truth: $364.50 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Pirelli 275/40R20 106W XL RFT P0 PZ4-LUX...\u001b[0m\n", + "\u001b[94m171: Guess: $378.99 Truth: $378.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Torklift C3212 Rear Tie Down\u001b[0m\n", + "\u001b[94m172: Guess: $165.28 Truth: $165.28 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Cardone 78-4226 Remanufactured Ford Comp...\u001b[0m\n", + "\u001b[94m173: Guess: $57.33 Truth: $56.74 Abs Error: $0.59 % Error: 1.0% SLE: 0.00 Item: Kidde AccessPoint 001798 Supra TouchPoin...\u001b[0m\n", + "\u001b[94m174: Guess: $307.95 Truth: $307.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: 3M Protecta 3100414 Self Retracting Life...\u001b[0m\n", + "\u001b[94m175: Guess: $38.00 Truth: $38.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Plantronics 89435-01 Wired Headset, Blac...\u001b[0m\n", + "\u001b[95m176: Guess: $63.88 Truth: $53.00 Abs Error: $10.88 % Error: 20.5% SLE: 0.03 Item: Logitech K750 Wireless Solar Keyboard fo...\u001b[0m\n", + "\u001b[94m177: Guess: $498.00 Truth: $498.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Olympus PEN E-PL9 Body Only with 3-Inch ...\u001b[0m\n", + "\u001b[94m178: Guess: $53.99 Truth: $53.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Beck/Arnley 051-6066 Hub & Bearing Assem...\u001b[0m\n", + "\u001b[94m179: Guess: $319.88 Truth: $350.00 Abs Error: $30.12 % Error: 8.6% SLE: 0.01 Item: Eibach Pro-Kit Performance Springs E10-6...\u001b[0m\n", + "\u001b[94m180: Guess: $299.99 Truth: $299.95 Abs Error: $0.04 % Error: 0.0% SLE: 0.00 Item: LEGO DC Batman 1989 Batwing 76161 Displa...\u001b[0m\n", + "\u001b[94m181: Guess: $94.93 Truth: $94.93 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Kingston Brass KS3608PL Restoration 4-In...\u001b[0m\n", + "\u001b[94m182: Guess: $379.00 Truth: $379.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Polk Vanishing Series 265-LS In-Wall 3-W...\u001b[0m\n", + "\u001b[94m183: Guess: $299.95 Truth: $299.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Spec-D Tuning LED Projector Headlights G...\u001b[0m\n", + "\u001b[94m184: Guess: $24.99 Truth: $24.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: RICHMOND & FINCH Airpod Pro Case, Green ...\u001b[0m\n", + "\u001b[94m185: Guess: $41.04 Truth: $41.04 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: LFA Industries 43B-5A-33JT 1/16-1/2-1.5-...\u001b[0m\n", + "\u001b[94m186: Guess: $327.90 Truth: $327.90 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SAUTVS LED Headlight Assembly for Slings...\u001b[0m\n", + "\u001b[94m187: Guess: $10.99 Truth: $10.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: 2 Pack Combo Womens Safety Glasses Impac...\u001b[0m\n", + "\u001b[94m188: Guess: $14.99 Truth: $14.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Arepa - Venezuelan cuisine - Venezuela P...\u001b[0m\n", + "\u001b[91m189: Guess: $22.57 Truth: $84.95 Abs Error: $62.38 % Error: 73.4% SLE: 1.67 Item: Schlage Lock Company KS23D2300 Padlock, ...\u001b[0m\n", + "\u001b[94m190: Guess: $111.00 Truth: $111.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Techni Mobili White Sit to Stand Mobile ...\u001b[0m\n", + "\u001b[94m191: Guess: $123.73 Truth: $123.73 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Special Lite Products Contemporary Wall ...\u001b[0m\n", + "\u001b[94m192: Guess: $557.38 Truth: $557.38 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Tascam DP-24SD 24-Track Digital Portastu...\u001b[0m\n", + "\u001b[94m193: Guess: $95.55 Truth: $95.55 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Glow Lighting 636CC10SP Vista Crystal Fl...\u001b[0m\n", + "\u001b[94m194: Guess: $154.00 Truth: $154.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Z3 Wind Deflector, Smoke Tint, Lexan, Wi...\u001b[0m\n", + "\u001b[94m195: Guess: $198.99 Truth: $198.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Olympus E-20 5MP Digital Camera w/ 4x Op...\u001b[0m\n", + "\u001b[94m196: Guess: $430.44 Truth: $430.44 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: PHYNEDI 1:1000 World Trade Center (1973-...\u001b[0m\n", + "\u001b[94m197: Guess: $45.67 Truth: $45.67 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: YANGHUAN Unstable Unicorns Adventure Car...\u001b[0m\n", + "\u001b[94m198: Guess: $249.00 Truth: $249.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Interlogix NX-1820E NetworX Touch Screen...\u001b[0m\n", + "\u001b[94m199: Guess: $42.99 Truth: $42.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Steering Damper,Universal Motorcycle Han...\u001b[0m\n", + "\u001b[94m200: Guess: $181.33 Truth: $181.33 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Amprobe TIC 410A Hot Stick Attachment\u001b[0m\n", + "\u001b[95m201: Guess: $6.99 Truth: $6.03 Abs Error: $0.96 % Error: 15.9% SLE: 0.02 Item: MyCableMart 3.5mm Plug/Jack, 4 Conductor...\u001b[0m\n", + "\u001b[94m202: Guess: $33.94 Truth: $29.99 Abs Error: $3.95 % Error: 13.2% SLE: 0.01 Item: OtterBox + Pop Symmetry Series Case for ...\u001b[0m\n", + "\u001b[94m203: Guess: $899.00 Truth: $899.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dell XPS X8700-1572BLK Desktop ( Intel C...\u001b[0m\n", + "\u001b[94m204: Guess: $399.99 Truth: $399.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Franklin Iron Works Sperry Industrial Br...\u001b[0m\n", + "\u001b[94m205: Guess: $4.66 Truth: $4.66 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Avery Legal Dividers, Standard Collated ...\u001b[0m\n", + "\u001b[94m206: Guess: $261.41 Truth: $261.41 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Moen 8346 Commercial Posi-Temp Pressure ...\u001b[0m\n", + "\u001b[94m207: Guess: $136.97 Truth: $136.97 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Carlisle Versa Trail ATR All Terrain Rad...\u001b[0m\n", + "\u001b[94m208: Guess: $79.00 Truth: $79.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SUNWAYFOTO 44mm Tripod Ball Head Arca Co...\u001b[0m\n", + "\u001b[94m209: Guess: $444.99 Truth: $444.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: NanoBeam AC NBE-5AC-Gen2-US 4 Units 5GHz...\u001b[0m\n", + "\u001b[94m210: Guess: $411.94 Truth: $411.94 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: WULF 4\" Front 2\" Rear Leveling Lift Kit ...\u001b[0m\n", + "\u001b[94m211: Guess: $148.40 Truth: $148.40 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Alera ALEVABFMC Valencia Series Mobile B...\u001b[0m\n", + "\u001b[94m212: Guess: $244.99 Truth: $244.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: YU-GI-OH! Ignition Assault Booster Box\u001b[0m\n", + "\u001b[94m213: Guess: $86.50 Truth: $86.50 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: 48\" x 36\" Extra-Large Framed Magnetic Bl...\u001b[0m\n", + "\u001b[94m214: Guess: $297.95 Truth: $297.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Dell Latitude D620 Renewed Notebook PC\u001b[0m\n", + "\u001b[94m215: Guess: $399.99 Truth: $399.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: acer Aspire 5 Laptop, AMD Ryzen 3 5300U ...\u001b[0m\n", + "\u001b[94m216: Guess: $599.00 Truth: $599.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Elk 31080/6RC-GRN 30 by 6-Inch Viva 6-Li...\u001b[0m\n", + "\u001b[91m217: Guess: $64.90 Truth: $105.99 Abs Error: $41.09 % Error: 38.8% SLE: 0.23 Item: Barbie Top Model Doll\u001b[0m\n", + "\u001b[94m218: Guess: $629.97 Truth: $689.00 Abs Error: $59.03 % Error: 8.6% SLE: 0.01 Item: Danby Designer 20-In. Electric Range wit...\u001b[0m\n", + "\u001b[94m219: Guess: $404.99 Truth: $404.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: FixtureDisplays® Metal Truss Podium Doub...\u001b[0m\n", + "\u001b[94m220: Guess: $207.76 Truth: $207.76 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: ACDelco 13597235 GM Original Equipment A...\u001b[0m\n", + "\u001b[94m221: Guess: $171.82 Truth: $171.82 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: EBC S1KF1135 Stage-1 Premium Street Brak...\u001b[0m\n", + "\u001b[94m222: Guess: $293.24 Truth: $293.24 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: FXR Men's Boost FX Jacket (Black/Orange/...\u001b[0m\n", + "\u001b[94m223: Guess: $374.95 Truth: $374.95 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SuperATV Scratch Resistant 3-in-1 Flip W...\u001b[0m\n", + "\u001b[94m224: Guess: $111.99 Truth: $111.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: SBU 3 Layer All Weather Mini Van Car Cov...\u001b[0m\n", + "\u001b[91m225: Guess: $22.99 Truth: $42.99 Abs Error: $20.00 % Error: 46.5% SLE: 0.37 Item: 2 Pack Outdoor Brochure Holder Advertisi...\u001b[0m\n", + "\u001b[94m226: Guess: $116.71 Truth: $116.71 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Monroe Shocks & Struts Quick-Strut 17158...\u001b[0m\n", + "\u001b[94m227: Guess: $118.61 Truth: $118.61 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Elements of Design Magellan EB235AL Thre...\u001b[0m\n", + "\u001b[94m228: Guess: $147.12 Truth: $147.12 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: GM Genuine Parts 15-62961 Air Conditioni...\u001b[0m\n", + "\u001b[94m229: Guess: $119.99 Truth: $119.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Baseus 17-in-1 USB C Docking Station to ...\u001b[0m\n", + "\u001b[94m230: Guess: $369.98 Truth: $369.98 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Whitehall™ Personalized Whitehall Capito...\u001b[0m\n", + "\u001b[94m231: Guess: $315.55 Truth: $315.55 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Pro Circuit Works Pipe PY05250 for 02-19...\u001b[0m\n", + "\u001b[94m232: Guess: $210.00 Truth: $190.99 Abs Error: $19.01 % Error: 10.0% SLE: 0.01 Item: HYANKA 15 \"1200W Professional DJ Speaker...\u001b[0m\n", + "\u001b[94m233: Guess: $155.00 Truth: $155.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Bluetooth X6BT Card Reader Writer Encode...\u001b[0m\n", + "\u001b[94m234: Guess: $349.99 Truth: $349.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: AIRAID Cold Air Intake System by K&N: In...\u001b[0m\n", + "\u001b[94m235: Guess: $249.99 Truth: $249.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Bostingner Shower Faucets Sets Complete,...\u001b[0m\n", + "\u001b[94m236: Guess: $42.99 Truth: $42.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: PIT66 Front Bumper Turn Signal Lights, C...\u001b[0m\n", + "\u001b[95m237: Guess: $22.99 Truth: $17.99 Abs Error: $5.00 % Error: 27.8% SLE: 0.05 Item: Caseology Bumpy Compatible with Google P...\u001b[0m\n", + "\u001b[94m238: Guess: $425.00 Truth: $425.00 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Fleck 2510 Timer Mechanical Filter Contr...\u001b[0m\n", + "\u001b[94m239: Guess: $249.99 Truth: $249.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Haloview MC7108 Wireless RV Backup Camer...\u001b[0m\n", + "\u001b[94m240: Guess: $138.23 Truth: $138.23 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Schmidt Spiele - Manhattan\u001b[0m\n", + "\u001b[94m241: Guess: $354.99 Truth: $414.99 Abs Error: $60.00 % Error: 14.5% SLE: 0.02 Item: Corsa 14333 Tip Kit (Ford Mustang GT)\u001b[0m\n", + "\u001b[94m242: Guess: $168.28 Truth: $168.28 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Hoshizaki FM116A Fan Motor Kit 1\u001b[0m\n", + "\u001b[94m243: Guess: $199.99 Truth: $199.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: BAINUO Antler Chandelier Lighting,6 Ligh...\u001b[0m\n", + "\u001b[94m244: Guess: $126.70 Truth: $126.70 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: DNA MOTORING HL-OH-FEXP06-SM-AM Smoke Le...\u001b[0m\n", + "\u001b[94m245: Guess: $5.91 Truth: $5.91 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Wera Stainless 3840/1 TS 2.5mm Hex Inser...\u001b[0m\n", + "\u001b[94m246: Guess: $193.06 Truth: $193.06 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Celestron - PowerSeeker 127EQ Telescope ...\u001b[0m\n", + "\u001b[94m247: Guess: $249.99 Truth: $249.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: NHOPEEW 10.1inch Android Car Radio Carpl...\u001b[0m\n", + "\u001b[94m248: Guess: $64.12 Truth: $64.12 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Other Harmonica (Suzuki-2Timer24- A)\u001b[0m\n", + "\u001b[94m249: Guess: $114.99 Truth: $114.99 Abs Error: $0.00 % Error: 0.0% SLE: 0.00 Item: Harley Air Filter Venturi Intake Air Cle...\u001b[0m\n", + "\u001b[94m250: Guess: $928.33 Truth: $926.00 Abs Error: $2.33 % Error: 0.3% SLE: 0.00 Item: Elite Screens Edge Free Ambient Light Re...\u001b[0m\n", + "\n", + "------------------------------------------------------------\n", + "TEST SUMMARY: Gpt Fine Tuned\n", + "------------------------------------------------------------\n", + "Samples tested: 250\n", + "Average absolute error: $5.18\n", + "Average percentage error: 5.1%\n", + "RMSLE: 0.1979\n", + "\n", + "Performance Breakdown:\n", + " 🔵 Excellent (<=15% error): 228 (91.2%)\n", + " 🟣 Good (<=30% error): 10 (4.0%)\n", + " 🔴 Poor (>30% error): 12 (4.8%)\n", + "------------------------------------------------------------\n" + ] + }, + { + "data": { + "image/png": "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", + "text/plain": [ + "
" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], + "source": [ + "from dynamic_testing import Tester # Imports Tester from dynamic_testing module.\n", + "\n", + "# Tester.test(my_predictor, data, excellent_threshold=0.10, good_threshold=0.25) # Example usage with custom thresholds:\n", + "\n", + "Tester.test(gpt_fine_tuned, test)" + ] + }, + { + "cell_type": "code", + "execution_count": null, + "id": "8ce2d93d-13f0-49bd-9928-c60b7bd22ff5", + "metadata": {}, + "outputs": [], + "source": [] + } + ], + "metadata": { + "kernelspec": { + "display_name": "Python 3 (ipykernel)", + "language": "python", + "name": "python3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.11.12" + } + }, + "nbformat": 4, + "nbformat_minor": 5 +} diff --git a/week6/community-contributions/week6_day2_google_colab.ipynb b/week6/community-contributions/week6_day2_google_colab.ipynb new file mode 100644 index 0000000..c164af9 --- /dev/null +++ b/week6/community-contributions/week6_day2_google_colab.ipynb @@ -0,0 +1,676 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "provenance": [] + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "language_info": { + "name": "python" + } + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "91ae778b" + }, + "source": [ + "# Getting Started\n", + "\n", + "Before running this notebook, please ensure you have the following:\n", + "\n", + "1. **Local Modules:** Upload the necessary local Python files (`items.py`, `loaders.py`, `testing.py`) to the Colab runtime's temporary storage. You can do this by clicking the folder icon on the left sidebar, then the upload icon, and selecting the files.\n", + "2. **Hugging Face Access Token:** Add your Hugging Face access token to Colab's user data secrets. Click the key icon on the left sidebar, click \"New secret\", and add your token with the name `HF_TOKEN`.\n", + "3. **Install Dependencies:** Run the first code cell to install the required libraries with the specified versions.\n", + "\n", + "Once these steps are completed, you can run the rest of the notebook cells sequentially." + ] + }, + { + "cell_type": "code", + "execution_count": null, + "metadata": { + "id": "_fj3pImYM5dw" + }, + "outputs": [], + "source": [ + "# Install exact versions from local environment to match the course's environment\n", + "!pip install --upgrade pip\n", + "\n", + "# Install specific versions of required libraries\n", + "!pip install datasets==3.6.0\n", + "!pip install transformers==4.51.3\n", + "!pip install huggingface_hub==0.31.2\n", + "!pip install matplotlib==3.10.3\n", + "!pip install numpy==1.26.4\n", + "!pip install python-dotenv==1.1.0\n", + "!pip install tqdm==4.67.1" + ] + }, + { + "cell_type": "code", + "source": [ + "# Import necessary libraries\n", + "import os\n", + "import random\n", + "from dotenv import load_dotenv\n", + "from huggingface_hub import login\n", + "from datasets import load_dataset, Dataset, DatasetDict\n", + "import matplotlib.pyplot as plt\n", + "from collections import Counter, defaultdict\n", + "import numpy as np\n", + "import pickle" + ], + "metadata": { + "id": "YQHruTKgPMRX" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Retrieve the Hugging Face access token from Colab's user data secrets\n", + "# This token is needed to interact with the Hugging Face Hub\n", + "from google.colab import userdata\n", + "userdata.get('HF_TOKEN')" + ], + "metadata": { + "id": "jBdHkdyVNj_f" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Import custom classes from local files (items.py and loaders.py)\n", + "# These files were manually added to the Colab runtime's temporary storage\n", + "from loaders import ItemLoader\n", + "from items import Item" + ], + "metadata": { + "id": "FdBT3PPzNmq3" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Set the backend for matplotlib to display plots inline in the notebook\n", + "%matplotlib inline" + ], + "metadata": { + "id": "vynEBaq6OGEZ" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Load a single dataset (\"All_Beauty\") using the custom ItemLoader\n", + "# This was likely an initial test or example loading step\n", + "items = ItemLoader(\"Appliances\").load()" + ], + "metadata": { + "id": "OFOJtH6FOG2u" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Define a list of dataset names (Amazon product categories) to be loaded\n", + "dataset_names = [\n", + " \"Automotive\",\n", + " \"Electronics\",\n", + " \"Office_Products\",\n", + " \"Tools_and_Home_Improvement\",\n", + " \"Cell_Phones_and_Accessories\",\n", + " \"Toys_and_Games\",\n", + " \"Appliances\",\n", + " \"Musical_Instruments\",\n", + "]" + ], + "metadata": { + "id": "rkLXYtfhOJNn" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Check and print the available CPU cores and RAM in the Colab runtime environment\n", + "# This helps understand the resources available for data processing\n", + "import psutil\n", + "print(f\"CPU cores: {psutil.cpu_count()}\")\n", + "print(f\"Available RAM: {psutil.virtual_memory().available / (1024**3):.1f} GB\")" + ], + "metadata": { + "id": "1oQSUpovOfKs" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "items = []\n", + "for dataset_name in dataset_names:\n", + " loader = ItemLoader(dataset_name)\n", + " items.extend(loader.load(workers=8))\n", + "\n", + "# Now, time for a coffee break!!\n", + "# By the way, I put the biggest datasets first.. it gets faster." + ], + "metadata": { + "id": "UcV9RB2Go8nC" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Print the total number of items loaded from all datasets\n", + "print(f\"A grand total of {len(items):,} items\")" + ], + "metadata": { + "id": "YdkGJ_X3oI1g" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract token counts from all loaded items\n", + "tokens = [item.token_count for item in items]\n", + "# Create and display a histogram of token counts\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Token counts: Avg {sum(tokens)/len(tokens):,.1f} and highest {max(tokens):,}\\n\")\n", + "plt.xlabel('Length (tokens)')\n", + "plt.ylabel('Count')\n", + "plt.hist(tokens, rwidth=0.7, color=\"skyblue\", bins=range(0, 300, 10))\n", + "plt.show()" + ], + "metadata": { + "id": "8VzKJ7neo-wp" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract prices from all loaded items\n", + "prices = [item.price for item in items]\n", + "# Create and display a histogram of item prices\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Prices: Avg {sum(prices)/len(prices):,.1f} and highest {max(prices):,}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"blueviolet\", bins=range(0, 1000, 10))\n", + "plt.show()" + ], + "metadata": { + "id": "ZLFJycNZpDak" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Count the occurrences of each category in the loaded items\n", + "category_counts = Counter()\n", + "for item in items:\n", + " category_counts[item.category]+=1\n", + "\n", + "# Extract categories and their counts for plotting\n", + "categories = category_counts.keys()\n", + "counts = [category_counts[category] for category in categories]\n", + "\n", + "# Create and display a bar chart showing the count of items per category\n", + "plt.figure(figsize=(15, 6))\n", + "plt.bar(categories, counts, color=\"goldenrod\")\n", + "plt.title('How many in each category')\n", + "plt.xlabel('Categories')\n", + "plt.ylabel('Count')\n", + "\n", + "# Rotate x-axis labels for better readability\n", + "plt.xticks(rotation=30, ha='right')\n", + "\n", + "# Add value labels on top of each bar for clarity\n", + "for i, v in enumerate(counts):\n", + " plt.text(i, v, f\"{v:,}\", ha='center', va='bottom')\n", + "\n", + "# Display the chart\n", + "plt.show()" + ], + "metadata": { + "id": "6oRa8rI6pGb0" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Create a dictionary where keys are rounded prices and values are lists of items with that price\n", + "# This is done to group items by price for sampling\n", + "slots = defaultdict(list)\n", + "for item in items:\n", + " slots[round(item.price)].append(item)" + ], + "metadata": { + "id": "7mT5ZubkpJ06" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Create a curated sample dataset with a more even distribution of prices and reduced bias towards 'Automotive' category\n", + "# Items with price >= $240 are included entirely\n", + "# For prices < $240, if the number of items is <= 1200, all are included\n", + "# If the number of items > 1200, a weighted random sample of 1200 items is taken,\n", + "# giving non-Automotive items higher weight\n", + "\n", + "# Set random seeds for reproducibility\n", + "np.random.seed(42)\n", + "random.seed(42)\n", + "sample = []\n", + "for i in range(1, 1000):\n", + " slot = slots[i]\n", + " if i>=240:\n", + " sample.extend(slot)\n", + " elif len(slot) <= 1200:\n", + " sample.extend(slot)\n", + " else:\n", + " # Assign weights: 1 for 'Automotive', 5 for other categories\n", + " weights = np.array([1 if item.category=='Automotive' else 5 for item in slot])\n", + " # Normalize weights\n", + " weights = weights / np.sum(weights)\n", + " # Randomly select 1200 indices based on weights\n", + " selected_indices = np.random.choice(len(slot), size=1200, replace=False, p=weights)\n", + " # Select the items corresponding to the chosen indices\n", + " selected = [slot[i] for i in selected_indices]\n", + " sample.extend(selected)\n", + "\n", + "# Print the total number of items in the curated sample\n", + "print(f\"There are {len(sample):,} items in the sample\")" + ], + "metadata": { + "id": "qHJdXynopMBp" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract prices from the curated sample\n", + "prices = [float(item.price) for item in sample]\n", + "# Create and display a histogram of prices for the sample dataset\n", + "# This helps visualize the effect of the sampling process on the price distribution\n", + "plt.figure(figsize=(15, 10))\n", + "plt.title(f\"Avg {sum(prices)/len(prices):.2f} and highest {max(prices):,.2f}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"darkblue\", bins=range(0, 1000, 10))\n", + "plt.show()" + ], + "metadata": { + "id": "gtBkOdPGpOou" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Count the occurrences of each category in the curated sample\n", + "category_counts = Counter()\n", + "for item in sample:\n", + " category_counts[item.category]+=1\n", + "\n", + "# Extract categories and their counts for plotting\n", + "categories = category_counts.keys()\n", + "counts = [category_counts[category] for category in categories]\n", + "\n", + "# Create and display a bar chart showing the count of items per category in the sample\n", + "# This helps visualize the effect of weighted sampling on category distribution\n", + "plt.figure(figsize=(15, 6))\n", + "plt.bar(categories, counts, color=\"lightgreen\")\n", + "\n", + "# Customize the chart\n", + "plt.title('How many in each category')\n", + "plt.xlabel('Categories')\n", + "plt.ylabel('Count')\n", + "\n", + "# Rotate x-axis labels for better readability\n", + "plt.xticks(rotation=30, ha='right')\n", + "\n", + "# Add value labels on top of each bar for clarity\n", + "for i, v in enumerate(counts):\n", + " plt.text(i, v, f\"{v:,}\", ha='center', va='bottom')\n", + "\n", + "# Display the chart\n", + "plt.show()" + ], + "metadata": { + "id": "-lYpt40jpTE1" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Create and display a pie chart showing the percentage distribution of items across categories in the sample\n", + "plt.figure(figsize=(12, 10))\n", + "plt.pie(counts, labels=categories, autopct='%1.0f%%', startangle=90)\n", + "\n", + "# Add a circle at the center to create a donut chart (optional)\n", + "centre_circle = plt.Circle((0,0), 0.70, fc='white')\n", + "fig = plt.gcf()\n", + "fig.gca().add_artist(centre_circle)\n", + "plt.title('Categories')\n", + "\n", + "# Equal aspect ratio ensures that pie is drawn as a circle\n", + "plt.axis('equal')\n", + "\n", + "plt.show()" + ], + "metadata": { + "id": "5QPV4m2LpV3g" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Markdown cell indicates that the dataset curation is complete and ready for final checks\n", + "# Dataset Curated!\n", + "\n", + "# We've crafted an excellent dataset.\n", + "\n", + "# Let's do some final checks" + ], + "metadata": { + "id": "3Xc2ZxjapZ0a" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract prompt lengths (character counts) and prices from the curated sample\n", + "sizes = [len(item.prompt) for item in sample]\n", + "prices = [item.price for item in sample]\n", + "\n", + "# Create and display a scatter plot to visualize the relationship between prompt size and price\n", + "# This helps check for any simple correlation between the two\n", + "plt.figure(figsize=(15, 8))\n", + "plt.scatter(sizes, prices, s=0.2, color=\"red\")\n", + "\n", + "# Add labels and title\n", + "plt.xlabel('Size')\n", + "plt.ylabel('Price')\n", + "plt.title('Is there a simple correlation?')\n", + "\n", + "# Display the plot\n", + "plt.show()" + ], + "metadata": { + "id": "VXYQkVarpceE" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Define a helper function to report information about an item\n", + "# It prints the item's prompt, the last 10 token IDs, and the decoded last 10 tokens\n", + "def report(item):\n", + " prompt = item.prompt\n", + " tokens = Item.tokenizer.encode(item.prompt)\n", + " print(prompt)\n", + " print(tokens[-10:])\n", + " print(Item.tokenizer.batch_decode(tokens[-10:]))" + ], + "metadata": { + "id": "1BBJNDAKpgL_" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Use the report function to display information about a specific item in the sample\n", + "# This helps inspect the data and the tokenizer's behavior\n", + "report(sample[398000])" + ], + "metadata": { + "id": "ZO2zF09wpiPp" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "source": [ + "## Observation\n", + "\n", + "An interesting thing about the Llama tokenizer is that every number from 1 to 999 gets mapped to 1 token, much as we saw with gpt-4o. The same is not true of qwen2, gemma and phi3, which all map individual digits to tokens. This does turn out to be a bit useful for our project, although it's not an essential requirement." + ], + "metadata": { + "id": "GCkwmt_VpsaU" + } + }, + { + "cell_type": "markdown", + "source": [ + "# Finally\n", + "\n", + "It's time to break down our data into a training, test and validation dataset.\n", + "\n", + "It's typical to use 5%-10% of your data for testing purposes, but actually we have far more than we need at this point. We'll take 400,000 points for training, and we'll reserve 2,000 for testing, although we won't use all of them.\n" + ], + "metadata": { + "id": "dy6WGVAmpx0g" + } + }, + { + "cell_type": "code", + "source": [ + "# Set random seed for reproducibility before shuffling and splitting the sample\n", + "random.seed(42)\n", + "# Shuffle the curated sample dataset\n", + "random.shuffle(sample)\n", + "# Split the shuffled sample into training (400,000 items) and testing (2,000 items) sets\n", + "train = sample[:400_000]\n", + "test = sample[400_000:402_000]\n", + "# Print the sizes of the training and testing sets\n", + "print(f\"Divided into a training set of {len(train):,} items and test set of {len(test):,} items\")" + ], + "metadata": { + "id": "oY1ZSkW7p0VS" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract prices from the first 250 items of the test set\n", + "prices = [float(item.price) for item in test[:250]]\n", + "# Create and display a histogram of prices for the first 250 test items\n", + "# This provides a quick look at the price distribution in a small portion of the test data\n", + "plt.figure(figsize=(15, 6))\n", + "plt.title(f\"Avg {sum(prices)/len(prices):.2f} and highest {max(prices):,.2f}\\n\")\n", + "plt.xlabel('Price ($)')\n", + "plt.ylabel('Count')\n", + "plt.hist(prices, rwidth=0.7, color=\"darkblue\", bins=range(0, 1000, 10))\n", + "plt.show()" + ], + "metadata": { + "id": "nLnRpUbtp17N" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Extract prompts from the training set\n", + "train_prompts = [item.prompt for item in train]\n", + "# Extract prices from the training set\n", + "train_prices = [item.price for item in train]\n", + "# Extract test prompts (using the test_prompt method) from the test set\n", + "test_prompts = [item.test_prompt() for item in test]\n", + "# Extract prices from the test set\n", + "test_prices = [item.price for item in test]" + ], + "metadata": { + "id": "kpw1Y8IIp6kw" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Create Hugging Face Dataset objects from the training and testing data\n", + "train_dataset = Dataset.from_dict({\"text\": train_prompts, \"price\": train_prices})\n", + "test_dataset = Dataset.from_dict({\"text\": test_prompts, \"price\": test_prices})\n", + "# Create a DatasetDict containing the training and testing datasets\n", + "dataset = DatasetDict({\n", + " \"train\": train_dataset,\n", + " \"test\": test_dataset\n", + "})" + ], + "metadata": { + "id": "WtEFiTlvp8hL" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Push the created DatasetDict to the Hugging Face Hub\n", + "# Replace \"aaron-official\" with your Hugging Face username\n", + "# The dataset will be named \"your-username/pricer-data\" and will be private\n", + "# HF_USER = \"aaron-official\" # Uncomment and replace with your HF username\n", + "# DATASET_NAME = f\"{HF_USER}/pricer-data\" # Uncomment\n", + "# dataset.push_to_hub(DATASET_NAME, private=True) # Uncomment to push to hub" + ], + "metadata": { + "id": "sSnwZIxHp-VJ" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "source": [ + "# Pickle (serialize) the training and testing datasets and save them as files\n", + "# This allows for quick loading of the processed data in future sessions\n", + "with open('train.pkl', 'wb') as file:\n", + " pickle.dump(train, file)\n", + "\n", + "with open('test.pkl', 'wb') as file:\n", + " pickle.dump(test, file)" + ], + "metadata": { + "id": "WRawIsrOqMQ-" + }, + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "bd72e246" + }, + "source": [ + "# Mount Google Drive to access files in your Drive\n", + "from google.colab import drive\n", + "drive.mount('/content/drive')" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "6fc5d915" + }, + "source": [ + "Once your Google Drive is mounted, you can copy the file to a folder in your Drive. Replace `My Drive/your_folder_name` with the path to the folder where you want to save the file." + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "f319129b" + }, + "source": [ + "# Import the shutil module for file operations\n", + "import shutil\n", + "\n", + "# Define the destination path in Google Drive and the source path of the pickled training data\n", + "# Replace 'My Drive/your_folder_name' with your desired folder path in Google Drive\n", + "destination_path = '/content/drive/My Drive/train.pkl'\n", + "source_path = '/content/train.pkl'\n", + "\n", + "# Copy the pickled training data file from the Colab environment to Google Drive\n", + "shutil.copyfile(source_path, destination_path)\n", + "\n", + "# Print a confirmation message\n", + "print(f\"Copied {source_path} to {destination_path}\")" + ], + "execution_count": null, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "d23d6cf0" + }, + "source": [ + "# Import the shutil module for file operations\n", + "import shutil\n", + "\n", + "# Define the destination path in Google Drive and the source path of the pickled test data\n", + "# Replace 'My Drive/your_folder_name' with your desired folder path in Google Drive\n", + "destination_path = '/content/drive/My Drive/test.pkl'\n", + "source_path = '/content/test.pkl'\n", + "\n", + "# Copy the pickled test data file from the Colab environment to Google Drive\n", + "shutil.copyfile(source_path, destination_path)\n", + "\n", + "# Print a confirmation message\n", + "print(f\"Copied {source_path} to {destination_path}\")" + ], + "execution_count": null, + "outputs": [] + } + ] +} \ No newline at end of file diff --git a/week6/day5.ipynb b/week6/day5.ipynb index 5d5619e..14abeab 100644 --- a/week6/day5.ipynb +++ b/week6/day5.ipynb @@ -149,7 +149,7 @@ "source": [ "# First let's work on a good prompt for a Frontier model\n", "# Notice that I'm removing the \" to the nearest dollar\"\n", - "# When we train our own models, we'll need to make the problem as easy as possible, \n", + "# When we train our own models, we'll need to make the problem as easy as possible,\n", "# but a Frontier model needs no such simplification.\n", "\n", "def messages_for(item):\n", @@ -393,6 +393,22 @@ "openai.fine_tuning.jobs.list_events(fine_tuning_job_id=job_id, limit=10).data" ] }, + { + "cell_type": "code", + "execution_count": null, + "id": "b19ea9e9", + "metadata": {}, + "outputs": [], + "source": [ + "import wandb\n", + "from wandb.integration.openai.fine_tuning import WandbLogger\n", + "\n", + "# Log in to Weights & Biases.\n", + "wandb.login()\n", + "# Sync the fine-tuning job with Weights & Biases.\n", + "WandbLogger.sync(fine_tune_job_id=job_id, project=\"gpt-pricer\")" + ] + }, { "cell_type": "markdown", "id": "066fef03-8338-4526-9df3-89b649ad4f0a", @@ -490,7 +506,7 @@ "\n", "def gpt_fine_tuned(item):\n", " response = openai.chat.completions.create(\n", - " model=fine_tuned_model_name, \n", + " model=fine_tuned_model_name,\n", " messages=messages_for(item),\n", " seed=42,\n", " max_tokens=7\n",