brad/LLM_Engineering_OLD
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #805 from javieroteromarquez/jom/week5

Browse Source 
Javi Bootcamp Week 5 Exercise
This commit is contained in:
Ed Donner
2025-10-23 10:54:22 -04:00
committed by GitHub
parent 4605821c2a f12577b6fb
commit 903884c3d2
1 changed files with 623 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

623
week5/community-contributions/week5_jom/Exercise_week5_jom.ipynb Normal file
View File

@@ -0,0 +1,623 @@
{
"cells": [
{
"cell_type": "markdown",
"id": "6f0f38e7",
"metadata": {},
"source": [
"# Email Mindmap Demo (Week 5 Community Contribution)\n",
"\n",
"Welcome to the **Email Mindmap Demo** notebook! This demo walks you through a workflow for exploring and visualizing email relationships using embeddings and mindmaps.\n",
"\n",
"---\n",
"\n",
"## 📋 Workflow Overview\n",
"\n",
"1. **Load/Create Synthetic Email Data** \n",
" Generate or load varied types of emails: work, personal, family, subscriptions, etc.\n",
"\n",
"2. **Generate Embeddings** \n",
" Use an open-source model to create vector embeddings for email content.\n",
"\n",
"3. **Build & Visualize a Mindmap** \n",
" Construct a mindmap of email relationships and visualize it interactively using `networkx` and `matplotlib`.\n",
"\n",
"4. **Question-Answering Interface** \n",
" Query the email content and the mindmap using a simple Q&A interface powered by Gradio.\n",
"\n",
"---\n",
"\n",
"## ⚙️ Requirements\n",
"\n",
"> **Tip:** \n",
"> I'm including an example of the synthetic emails in case you don't want to run that part.\n",
"> Might need to install other libraries like pyvis, nbformat and faiss-cpu\n",
"\n",
"\n",
"## ✨ Features\n",
"\n",
"- Synthetic generation of varied emails (work, personal, family, subscriptions)\n",
"- Embedding generation with open-source models (hugging face sentence-transformer)\n",
"- Interactive mindmap visualization (`networkx`, `pyvis`)\n",
"- Simple chatbot interface (Gradio) and visualization of mindmap created\n",
"\n",
"---\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 1,
"id": "a9aeb363",
"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 AI\n",
"OLLAMA API Key exists and begins 36\n"
]
}
],
"source": [
"# imports\n",
"\n",
"import os\n",
"from dotenv import load_dotenv\n",
"from openai import OpenAI\n",
"import gradio as gr\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",
"ollama_api_key = os.getenv('OLLAMA_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 (and this is optional)\")\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 (and this is optional)\")\n",
"\n",
"if ollama_api_key:\n",
" print(f\"OLLAMA API Key exists and begins {ollama_api_key[:2]}\")\n",
"else:\n",
" print(\"OLLAMA API Key not set (and this is optional)\")\n",
"\n",
"# Connect to client libraries\n",
"\n",
"openai = OpenAI()\n",
"\n",
"anthropic_url = \"https://api.anthropic.com/v1/\"\n",
"gemini_url = \"https://generativelanguage.googleapis.com/v1beta/openai/\"\n",
"ollama_url = \"http://localhost:11434/v1\"\n",
"\n",
"anthropic = OpenAI(api_key=anthropic_api_key, base_url=anthropic_url)\n",
"gemini = OpenAI(api_key=google_api_key, base_url=gemini_url)\n",
"ollama = OpenAI(api_key=ollama_api_key, base_url=ollama_url)\n",
"\n"
]
},
{
"cell_type": "markdown",
"id": "b8ddce62",
"metadata": {},
"source": [
"## Preparation of synthetic data (could have been week2 work)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "2e250912",
"metadata": {},
"outputs": [],
"source": [
"#using ollama gpt oss 120b cloud i'm going to create synthetic emails using a persona.\n",
"#they are going to be saved in a json file with different keys\n",
"from pydantic import BaseModel, Field\n",
"from typing import List, Optional\n",
"\n",
"\n",
"class Email(BaseModel):\n",
" sender: str = Field(description=\"Email address of the sender\")\n",
" subject: str = Field(description=\"Email subject line\")\n",
" body: str = Field(description=\"Email body content\")\n",
" timestamp: str = Field(description=\"ISO 8601 timestamp when email was received\")\n",
" category: str = Field(description=\"Category of the email\")\n",
"\n",
"class EmailBatch(BaseModel):\n",
" emails: List[Email] = Field(description=\"List of generated emails\")\n"
]
},
{
"cell_type": "code",
"execution_count": 3,
"id": "1f67fdb3",
"metadata": {},
"outputs": [],
"source": [
"def create_persona(name: str, age: int, occupation: str, \n",
" interests: List[str], family_status: str) -> str:\n",
" persona = f\"\"\"\n",
" You are generating synthetic emails for a realistic inbox simulation.\n",
"\n",
" **Person Profile:**\n",
" - Name: {name}\n",
" - Age: {age}\n",
" - Occupation: {occupation}\n",
" - Interests: {', '.join(interests)}\n",
" - Family Status: {family_status}\n",
"\n",
" **Email Categories to Include:**\n",
" 1. **Work Emails**: Project updates, meeting invitations, colleague communications, \n",
" performance reviews, company announcements\n",
" 2. **Purchases**: Order confirmations, shipping notifications, delivery updates, \n",
" receipts from various retailers (Amazon, local shops, etc.)\n",
" 3. **Subscriptions**: Newsletter updates, streaming services (Netflix, Spotify), \n",
" software subscriptions (Adobe, Microsoft 365), magazine subscriptions\n",
" 4. **Family**: Communications with parents, siblings, children, extended family members,\n",
" family event planning, photo sharing\n",
" 5. **Friends**: Social plans, birthday wishes, casual conversations, group hangouts,\n",
" catching up messages\n",
" 6. **Finance**: Bank statements, credit card bills, investment updates, tax documents,\n",
" payment reminders\n",
" 7. **Social Media**: Facebook notifications, LinkedIn updates, Instagram activity,\n",
" Twitter mentions\n",
" 8. **Personal**: Doctor appointments, gym memberships, utility bills, insurance updates\n",
"\n",
" **Instructions:**\n",
" - Generate realistic email content that reflects the person's life over time\n",
" - Include temporal patterns (more work emails on weekdays, more personal on weekends)\n",
" - Create realistic sender names and email addresses\n",
" - Vary email length and formality based on context\n",
" - Include realistic subject lines\n",
" - Make emails interconnected when appropriate (e.g., follow-up emails, conversation threads)\n",
" - Include seasonal events (holidays, birthdays, annual renewals)\n",
" \"\"\"\n",
" return persona\n",
"\n",
"persona_description = create_persona(\n",
" name=\"John Doe\",\n",
" age=30,\n",
" occupation=\"Software Engineer\",\n",
" interests=[\"technology\", \"reading\", \"traveling\"],\n",
" family_status=\"single\"\n",
")\n"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "cec185e3",
"metadata": {},
"outputs": [],
"source": [
"from openai import OpenAI\n",
"from datetime import datetime, timedelta\n",
"import random\n",
"from typing import List\n",
"\n",
"def generate_synthetic_emails(\n",
" persona_description: str,\n",
" num_emails: int,\n",
" start_date: str,\n",
" end_date: str,\n",
" model: str = \"gpt-4o-2024-08-06\"\n",
") -> List[Email]:\n",
" \"\"\"\n",
" NEEDS TO WORK WITH OPENAI MODELS BECAUSE OF PARSED (STRUC OUTPUT) MODELS\n",
" Generates synthetic emails using OpenAI's structured output feature.\n",
" \n",
" Args:\n",
" persona_description: Detailed persona description\n",
" num_emails: Number of emails to generate per batch\n",
" start_date: Start date for email timestamps\n",
" end_date: End date for email timestamps\n",
" model: OpenAI model to use (must support structured outputs)\n",
" \n",
" Returns:\n",
" List of Email objects\n",
" \"\"\"\n",
" \n",
" # Calculate date range for context\n",
" date_range_context = f\"\"\"\n",
" Generate emails with timestamps between {start_date} and {end_date}.\n",
" Distribute emails naturally across this time period, with realistic patterns:\n",
" - More emails during business hours on weekdays\n",
" - Fewer emails late at night\n",
" - Occasional weekend emails\n",
" - Bursts of activity around events or busy periods\n",
" \"\"\"\n",
" \n",
" # System message combining persona and structure instructions\n",
" system_message = f\"\"\"\n",
" {persona_description}\n",
"\n",
" {date_range_context}\n",
"\n",
" Generate {num_emails} realistic emails that fit this person's life. \n",
" Ensure variety in categories, senders, and content while maintaining realism.\n",
" \"\"\"\n",
" \n",
" try:\n",
" client = OpenAI()\n",
"\n",
" response = client.chat.completions.parse(\n",
" model=model,\n",
" messages=[\n",
" {\n",
" \"role\": \"system\",\n",
" \"content\": system_message\n",
" },\n",
" {\n",
" \"role\": \"user\",\n",
" \"content\": f\"Generate {num_emails} diverse, realistic emails for this person's inbox.\"\n",
" }\n",
" ],\n",
" response_format=EmailBatch,\n",
" )\n",
" return response.choices[0].message.parsed.emails\n",
" \n",
" except Exception as e:\n",
" print(f\"Error generating emails: {e}\")\n",
" return []\n",
"\n",
"\n",
"def save_emails_to_json(emails: List[Email], filename: str):\n",
" \"\"\"\n",
" Saves emails to a JSON file.\n",
" \"\"\"\n",
" import json\n",
" \n",
" emails_dict = [email.model_dump() for email in emails]\n",
" \n",
" with open(filename, 'w', encoding='utf-8') as f:\n",
" json.dump(emails_dict, f, indent=2, ensure_ascii=False)\n",
" \n",
" print(f\"Saved {len(emails)} emails to {filename}\")\n"
]
},
{
"cell_type": "code",
"execution_count": 51,
"id": "be31f352",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"now\n"
]
}
],
"source": [
"mails_2 = generate_synthetic_emails(\n",
" persona_description = persona_description,\n",
" num_emails = 100,\n",
" start_date = '2024-06-01',\n",
" end_date = '2025-01-01',\n",
" model = \"gpt-4o\"\n",
" )"
]
},
{
"cell_type": "code",
"execution_count": 52,
"id": "24d844f2",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Saved 101 emails to emails2.json\n"
]
}
],
"source": [
"save_emails_to_json(mails_2, 'emails2.json')"
]
},
{
"cell_type": "markdown",
"id": "2b9c704e",
"metadata": {},
"source": [
"## Create embeddings for the mails\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "777012f8",
"metadata": {},
"outputs": [],
"source": [
"# imports for langchain, plotly and Chroma\n",
"\n",
"from langchain.document_loaders import DirectoryLoader, TextLoader\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",
"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\n",
"import json\n",
"from langchain.vectorstores import FAISS\n",
"\n",
"#MODEL = \"gpt-4o-mini\"\n",
"db_name = \"vector_db\""
]
},
{
"cell_type": "code",
"execution_count": 38,
"id": "ce95d9c7",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Total number of chunks: 206\n",
"Sample metadata fields: ['sender', 'timestamp', 'category']\n"
]
}
],
"source": [
"# Read in emails from the emails.json file and construct LangChain documents\n",
"\n",
"\n",
"with open(\"emails.json\", \"r\", encoding=\"utf-8\") as f:\n",
" emails = json.load(f)\n",
"\n",
"documents = []\n",
"for email in emails:\n",
" # Extract metadata (all fields except 'content')\n",
" metadata = {k: v for k, v in email.items() if k in ['sender','category','timestamp']}\n",
" body = email.get(\"body\", \"\")\n",
" documents.append(Document(page_content=body, metadata=metadata))\n",
"\n",
"text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=100)\n",
"chunks = text_splitter.split_documents(documents)\n",
"\n",
"print(f\"Total number of chunks: {len(chunks)}\")\n",
"print(f\"Sample metadata fields: {list(documents[0].metadata.keys()) if documents else []}\")\n",
"\n",
"embeddings_model = HuggingFaceEmbeddings(model_name=\"sentence-transformers/all-MiniLM-L6-v2\")\n",
"\n",
"if os.path.exists(db_name):\n",
" Chroma(persist_directory=db_name, embedding_function=embeddings_model).delete_collection()\n",
"\n",
"vectorstore = FAISS.from_documents(chunks, embedding=embeddings_model)\n",
"\n",
"all_embeddings = [vectorstore.index.reconstruct(i) for i in range(vectorstore.index.ntotal)]\n",
"\n",
"total_vectors = vectorstore.index.ntotal\n",
"dimensions = vectorstore.index.d\n"
]
},
{
"cell_type": "markdown",
"id": "78ca65bb",
"metadata": {},
"source": [
"## Visualizing mindmap"
]
},
{
"cell_type": "code",
"execution_count": 44,
"id": "a99dd2d6",
"metadata": {},
"outputs": [],
"source": [
"import networkx as nx\n",
"import matplotlib.pyplot as plt\n",
"from sklearn.metrics.pairwise import cosine_similarity\n",
"import plotly.graph_objects as go\n",
"import numpy as np\n",
"from sklearn.cluster import KMeans\n",
"from sklearn.manifold import TSNE # Or use UMAP\n",
"from pyvis.network import Network\n",
"\n",
"# Here, emails is your list of email objects, with .subject or .body\n",
"\n",
"# Build similarity graph\n",
"def build_mindmap_html(emails, all_embeddings, threshold=0.6):\n",
" similarity = cosine_similarity(all_embeddings)\n",
"\n",
" G = nx.Graph()\n",
" for i, email in enumerate(emails):\n",
" G.add_node(i, label=email['subject'][:80], title=email['body'][:50]) # Custom hover text\n",
"\n",
" for i in range(len(emails)):\n",
" for j in range(i+1, len(emails)):\n",
" if similarity[i][j] > threshold:\n",
" G.add_edge(i, j, weight=float(similarity[i][j]))\n",
"\n",
" # Convert to pyvis network\n",
" nt = Network(notebook=True, height='700px', width='100%', bgcolor='#222222', font_color='white')\n",
" nt.from_nx(G)\n",
" html = nt.generate_html().replace(\"'\", \"\\\"\")\n",
" return html\n"
]
},
{
"cell_type": "markdown",
"id": "53a2fbaf",
"metadata": {},
"source": [
"## Putting it all together in a gradio.\n",
"It needs to have an interface to make questions, and the visual to see the mindmap.\n"
]
},
{
"cell_type": "code",
"execution_count": 45,
"id": "161144ac",
"metadata": {},
"outputs": [],
"source": [
"# create a new Chat with OpenAI\n",
"MODEL=\"gpt-4o-mini\"\n",
"llm = ChatOpenAI(temperature=0.7, model_name=MODEL)\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()\n",
"from langchain_core.callbacks import StdOutCallbackHandler\n",
"\n",
"# putting it together: set up the conversation chain with the GPT 3.5 LLM, the vector store and memory\n",
"conversation_chain_debug = ConversationalRetrievalChain.from_llm(llm=llm, retriever=retriever, memory=memory, callbacks=[StdOutCallbackHandler()])\n",
"conversation_chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=retriever, memory=memory)\n",
"\n",
"# Wrapping that in a function\n",
"\n",
"def chat(question, history):\n",
" result = conversation_chain.invoke({\"question\": question})\n",
" return result[\"answer\"]"
]
},
{
"cell_type": "code",
"execution_count": 60,
"id": "16a4d8d1",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"c:\\Users\\Javi\\Desktop\\course\\llm_engineering\\.venv\\Lib\\site-packages\\gradio\\chat_interface.py:347: UserWarning:\n",
"\n",
"The 'tuples' format for chatbot messages is deprecated and will be removed in a future version of Gradio. Please set type='messages' instead, which uses openai-style 'role' and 'content' keys.\n",
"\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Warning: When cdn_resources is 'local' jupyter notebook has issues displaying graphics on chrome/safari. Use cdn_resources='in_line' or cdn_resources='remote' if you have issues viewing graphics in a notebook.\n",
"* Running on local URL: http://127.0.0.1:7878\n",
"* To create a public link, set `share=True` in `launch()`.\n"
]
},
{
"data": {
"text/html": [
"<div><iframe src=\"http://127.0.0.1:7878/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
],
"text/plain": [
"<IPython.core.display.HTML object>"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/plain": []
},
"execution_count": 60,
"metadata": {},
"output_type": "execute_result"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Warning: When cdn_resources is 'local' jupyter notebook has issues displaying graphics on chrome/safari. Use cdn_resources='in_line' or cdn_resources='remote' if you have issues viewing graphics in a notebook.\n",
"Warning: When cdn_resources is 'local' jupyter notebook has issues displaying graphics on chrome/safari. Use cdn_resources='in_line' or cdn_resources='remote' if you have issues viewing graphics in a notebook.\n"
]
}
],
"source": [
"\n",
"import gradio as gr\n",
"\n",
"def show_mindmap():\n",
" # Call build_mindmap_html to generate the HTML\n",
" html = build_mindmap_html(emails, all_embeddings)\n",
" return f\"\"\"<iframe style=\"width: 100%; height: 600px;margin:0 auto\" name=\"result\" allow=\"midi; geolocation; microphone; camera; \n",
" display-capture; encrypted-media;\" sandbox=\"allow-modals allow-forms \n",
" allow-scripts allow-same-origin allow-popups \n",
" allow-top-navigation-by-user-activation allow-downloads\" allowfullscreen=\"\" \n",
" allowpaymentrequest=\"\" frameborder=\"0\" srcdoc='{html}'></iframe>\"\"\"\n",
"\n",
"\n",
"with gr.Blocks(title=\"Mindmap & Email Chatbot\") as demo:\n",
" gr.Markdown(\"# 📧 Mindmap Visualization & Email QA Chatbot\")\n",
" with gr.Row():\n",
" chatbot = gr.ChatInterface(fn=chat, title=\"Ask about your emails\",\n",
" examples=[\n",
" \"What is my most important message?\",\n",
" \"Who have I been communicating with?\",\n",
" \"Summarize recent emails\"\n",
" ],\n",
")\n",
" mindmap_html = gr.HTML(\n",
" show_mindmap,\n",
" label=\"🧠 Mindmap of Your Emails\",\n",
" )\n",
" # Reduce height: update show_mindmap (elsewhere) to ~400px, or do inline replace for the demo here:\n",
" # mindmap_html = gr.HTML(lambda: show_mindmap().replace(\"height: 600px\", \"height: 400px\"))\n",
" \n",
"demo.launch(inbrowser=True)\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "221a9d98",
"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": "3.12.8"
}
},
"nbformat": 4,
"nbformat_minor": 5
}