{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "dc8af57c-23a9-452e-9fc3-0e5027edda14",
   "metadata": {},
   "source": [
    "# AI-powered Brochure Generator\n",
    "---\n",
    "- 🌍 Task: Generate a company brochure using its name and website for clients, investors, and recruits.\n",
    "- 🧠 Model: Toggle `USE_OPENAI` to switch between OpenAI and Ollama models\n",
    "- 🕵️‍♂️ Data Extraction: Scraping website content and filtering key links (About, Products, Careers, Contact).\n",
    "- 📌 Output Format: a Markdown-formatted brochure streamed in real-time.\n",
    "- 🚀 Tools: BeautifulSoup, OpenAI API, and IPython display, ollama.\n",
    "- 🧑‍💻 Skill Level: Intermediate.\n",
    "\n",
    "🛠️ Requirements\n",
    "- ⚙️ Hardware: ✅ CPU is sufficient — no GPU required\n",
    "- 🔑 OpenAI API Key \n",
    "- Install Ollama and pull llama3.2:3b or another lightweight model\n",
    "---\n",
    "📢 Find more LLM notebooks on my [GitHub repository](https://github.com/lisekarimi/lexo)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec869f2c",
   "metadata": {},
   "source": [
    "## 🧩 System Design Overview\n",
    "\n",
    "### Class Structure\n",
    "\n",
    "![](https://github.com/lisekarimi/lexo/blob/main/assets/02_brochure_class_diagram.png?raw=true)\n",
    "\n",
    "This code consists of three main classes:\n",
    "\n",
    "1. **`Website`**:  \n",
    "   - Scrapes and processes webpage content.  \n",
    "   - Extracts **text** and **links** from a given URL.  \n",
    "\n",
    "2. **`LLMClient`**:  \n",
    "   - Handles interactions with **OpenAI or Ollama (`llama3`, `deepseek`, `qwen`)**.  \n",
    "   - Uses `get_relevant_links()` to filter webpage links.  \n",
    "   - Uses `generate_brochure()` to create and stream a Markdown-formatted brochure.  \n",
    "\n",
    "3. **`BrochureGenerator`**:  \n",
    "   - Uses `Website` to scrape the main webpage and relevant links.  \n",
    "   - Uses `LLMClient` to filter relevant links and generate a brochure.  \n",
    "   - Calls `generate()` to run the entire process.\n",
    "\n",
    "### Workflow\n",
    "\n",
    "1. **`main()`** initializes `BrochureGenerator` and calls `generate()`.  \n",
    "2. **`generate()`** calls **`LLMClient.get_relevant_links()`** to extract relevant links using **LLM (OpenAI/Ollama)**.  \n",
    "3. **`Website` scrapes the webpage**, extracting **text and links** from the given URL.  \n",
    "4. **Relevant links are re-scraped** using `Website` to collect additional content.  \n",
    "5. **All collected content is passed to `LLMClient.generate_brochure()`**.  \n",
    "6. **`LLMClient` streams the generated brochure** using **OpenAI or Ollama**.  \n",
    "7. **The final brochure is displayed in Markdown format.**\n",
    "\n",
    "![](https://github.com/lisekarimi/lexo/blob/main/assets/02_brochure_process.png?raw=true)\n",
    "\n",
    "\n",
    "### Intermediate reasoning\n",
    "\n",
    "In this workflow, we have intermediate reasoning because the LLM is called twice:\n",
    "\n",
    "1. **First LLM call**: Takes raw links → filters/selects relevant ones (reasoning step).\n",
    "2. **Second LLM call**: Takes selected content → generates final brochure.\n",
    "\n",
    "🧠 **LLM output becomes LLM input** — that’s intermediate reasoning.\n",
    "\n",
    "![](https://github.com/lisekarimi/lexo/blob/main/assets/02_llm_intermd_reasoning.png?raw=true)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b286461-35ee-4bc5-b07d-af554923e36d",
   "metadata": {},
   "source": [
    "## 📦 Import Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3fe5670c-5146-474b-9e75-484210533f55",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import requests\n",
    "import json\n",
    "import ollama\n",
    "from dotenv import load_dotenv\n",
    "from bs4 import BeautifulSoup\n",
    "from IPython.display import display, Markdown, update_display\n",
    "from openai import OpenAI"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f3e23181-1e66-410d-a910-1fb4230f8088",
   "metadata": {},
   "source": [
    "## 🧠 Define the Model\n",
    "\n",
    "The user can switch between OpenAI and Ollama by changing a single variable (`USE_OPENAI`). The model selection is dynamic."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fa2bd452-0cf4-4fec-9542-e1c86584c23f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load API key\n",
    "load_dotenv()\n",
    "api_key = os.getenv('OPENAI_API_KEY')\n",
    "if not api_key or not api_key.startswith('sk-'):\n",
    "    raise ValueError(\"Invalid OpenAI API key. Check your .env file.\")\n",
    "\n",
    "# Define the model dynamically\n",
    "USE_OPENAI = True  # True to use openai and False to use Ollama\n",
    "MODEL = 'gpt-4o-mini' if USE_OPENAI else 'llama3.2:3b'\n",
    "\n",
    "openai_client = OpenAI() if USE_OPENAI else None"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4fd997b7-1b89-4817-b53a-078164f5f71f",
   "metadata": {},
   "source": [
    "## 🏗️ Define Classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aed1af59-8b8f-4add-98dc-a9f1b5b511a5",
   "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",
    "    A utility class to scrape and process website content.\n",
    "    \"\"\"\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.string if soup.title else \"No title found\"\n",
    "        self.text = self.extract_text(soup)\n",
    "        self.links = self.extract_links(soup)\n",
    "\n",
    "    def extract_text(self, soup):\n",
    "        if soup.body:\n",
    "            for irrelevant in soup.body([\"script\", \"style\", \"img\", \"input\"]):\n",
    "                irrelevant.decompose()\n",
    "            return soup.body.get_text(separator=\"\\n\", strip=True)\n",
    "        return \"\"\n",
    "\n",
    "    def extract_links(self, soup):\n",
    "        links = [link.get('href') for link in soup.find_all('a')]\n",
    "        return [link for link in links if link and 'http' in 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": "ea04dc7e-ff4c-4113-83b7-0bddcf5072b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LLMClient:\n",
    "    def __init__(self, model=MODEL):\n",
    "        self.model = model\n",
    "\n",
    "    def get_relevant_links(self, website):\n",
    "        link_system_prompt = \"\"\"\n",
    "        You are given a list of links from a company website.\n",
    "        Select only relevant links for a brochure (About, Company, Careers, Products, Contact).\n",
    "        Exclude login, terms, privacy, and emails.\n",
    "\n",
    "        ### **Instructions**\n",
    "        - Return **only valid JSON**.\n",
    "        - **Do not** include explanations, comments, or Markdown.\n",
    "        - Example output:\n",
    "        {\n",
    "            \"links\": [\n",
    "                {\"type\": \"about\", \"url\": \"https://company.com/about\"},\n",
    "                {\"type\": \"contact\", \"url\": \"https://company.com/contact\"},\n",
    "                {\"type\": \"product\", \"url\": \"https://company.com/products\"}\n",
    "            ]\n",
    "        }\n",
    "        \"\"\"\n",
    "\n",
    "        user_prompt = f\"\"\"\n",
    "        Here is the list of links on the website of {website.url}:\n",
    "        Please identify the relevant web links for a company brochure. Respond in JSON format.\n",
    "        Do not include login, terms of service, privacy, or email links.\n",
    "        Links (some might be relative links):\n",
    "        {', '.join(website.links)}\n",
    "        \"\"\"\n",
    "\n",
    "        if USE_OPENAI:\n",
    "            response = openai_client.chat.completions.create(\n",
    "                model=self.model,\n",
    "                messages=[\n",
    "                    {\"role\": \"system\", \"content\": link_system_prompt},\n",
    "                    {\"role\": \"user\", \"content\": user_prompt}\n",
    "                ]\n",
    "            )\n",
    "            return json.loads(response.choices[0].message.content.strip())\n",
    "        else:\n",
    "            response = ollama.chat(\n",
    "                model=self.model,\n",
    "                messages=[\n",
    "                    {\"role\": \"system\", \"content\": link_system_prompt},\n",
    "                    {\"role\": \"user\", \"content\": user_prompt}\n",
    "                ]\n",
    "            )\n",
    "            result = response.get(\"message\", {}).get(\"content\", \"\").strip()\n",
    "            try:\n",
    "                return json.loads(result)  # Attempt to parse JSON\n",
    "            except json.JSONDecodeError:\n",
    "                print(\"Error: Response is not valid JSON\")\n",
    "                return {\"links\": []}  # Return empty list if parsing fails\n",
    "\n",
    "\n",
    "    def generate_brochure(self, company_name, content, language):\n",
    "        system_prompt = \"\"\"\n",
    "        You are a professional translator and writer who creates fun and engaging brochures.\n",
    "        Your task is to read content from a company’s website and write a short, humorous, joky,\n",
    "        and entertaining brochure for potential customers, investors, and job seekers.\n",
    "        Include details about the company’s culture, customers, and career opportunities if available.\n",
    "        Respond in Markdown format.\n",
    "        \"\"\"\n",
    "\n",
    "        user_prompt = f\"\"\"\n",
    "        Create a fun brochure for '{company_name}' using the following content:\n",
    "        {content[:5000]}\n",
    "        Respond in {language} only, and format your response correctly in Markdown.\n",
    "        Do NOT escape characters or return extra backslashes.\n",
    "        \"\"\"\n",
    "\n",
    "        if USE_OPENAI:\n",
    "            response_stream = openai_client.chat.completions.create(\n",
    "                model=self.model,\n",
    "                messages=[\n",
    "                    {\"role\": \"system\", \"content\": system_prompt},\n",
    "                    {\"role\": \"user\", \"content\": user_prompt}\n",
    "                ],\n",
    "                stream=True\n",
    "            )\n",
    "            response = \"\"\n",
    "            display_handle = display(Markdown(\"\"), display_id=True)\n",
    "            for chunk in response_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",
    "        else:\n",
    "            response_stream = ollama.chat(\n",
    "                model=self.model,\n",
    "                messages=[\n",
    "                    {\"role\": \"system\", \"content\": system_prompt},\n",
    "                    {\"role\": \"user\", \"content\": user_prompt}\n",
    "                ],\n",
    "                stream=True\n",
    "            )\n",
    "            display_handle = display(Markdown(\"\"), display_id=True)\n",
    "            full_text = \"\"\n",
    "            for chunk in response_stream:\n",
    "                if \"message\" in chunk:\n",
    "                        content = chunk[\"message\"][\"content\"] or \"\"\n",
    "                        full_text += content\n",
    "                        update_display(Markdown(full_text), display_id=display_handle.display_id)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1c69651f-e004-421e-acc5-c439e57a8762",
   "metadata": {},
   "outputs": [],
   "source": [
    "class BrochureGenerator:\n",
    "    \"\"\"\n",
    "    Main class to generate a company brochure.\n",
    "    \"\"\"\n",
    "    def __init__(self, company_name, url, language='English'):\n",
    "        self.company_name = company_name\n",
    "        self.url = url\n",
    "        self.language = language\n",
    "        self.website = Website(url)\n",
    "        self.llm_client = LLMClient()\n",
    "\n",
    "    def generate(self):\n",
    "        links = self.llm_client.get_relevant_links(self.website)\n",
    "        content = self.website.get_contents()\n",
    "\n",
    "        for link in links['links']:\n",
    "            linked_website = Website(link['url'])\n",
    "            content += f\"\\n\\n{link['type']}:\\n\"\n",
    "            content += linked_website.get_contents()\n",
    "\n",
    "        self.llm_client.generate_brochure(self.company_name, content, self.language)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1379d39d",
   "metadata": {},
   "source": [
    "## 📝 Generate Brochure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1a63519a-1981-477b-9de1-f1ff9be94201",
   "metadata": {},
   "outputs": [],
   "source": [
    "def main():\n",
    "    company_name = \"Tour Eiffel\"\n",
    "    url = \"https://www.toureiffel.paris/fr\"\n",
    "    language = \"French\"\n",
    "\n",
    "    generator = BrochureGenerator(company_name, url, language)\n",
    "    generator.generate()\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    main()"
   ]
  }
 ],
 "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": "3.12.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}