Company Brochure Generator

This project automatically generates a short company brochure from a website.
It scrapes the company’s landing page and relevant links (About, Careers, etc.), then summarizes them into a structured Markdown brochure.

The system supports two backends for the LLMs:

• Ollama (local model inference, e.g., LLaMA 3.2)
• OpenAI API (cloud-based model, e.g., GPT-4o-mini)

---

📂 Project Structure

├── website.py                # Scrapes webpages and extracts text + links
├── ollamaLinksExtractor.py   # Extracts relevant links using Ollama
├── openAILinksExtractor.py   # Extracts relevant links using OpenAI
├── run_w_ollama.py           # Full pipeline with Ollama
├── run_w_open_api.py         # Full pipeline with OpenAI
├── requirements.txt          # Python dependencies
├── pipeline_diagram_two_llms.png   # Pipeline diagram (with 2 LLM stages)

---

⚙️ Installation

1. Clone this repository
2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate   # macOS/Linux
   venv\Scripts\activate      # Windows

3. Install requirements:

   pip install -r requirements.txt

4. (For OpenAI only) Add your API key to a .env file:

   OPENAI_API_KEY=sk-proj-xxxx
   

---

▶️ Usage

Run with Ollama

Before running, make sure you have pulled the model you want to use (for example, LLaMA 3.2):

ollama pull llama3.2

Then run:

python run_w_ollama.py

Run with OpenAI

python run_w_open_api.py

Both scripts will:

1. Scrape the target website
2. LLM #1 filters the relevant links (About, Careers, etc.)
3. Fetch contents of those links
4. LLM #2 generates a company brochure in Markdown format
5. Save the result as <CompanyName>_brochure.md

---

📊 Pipeline

The workflow uses two LLM stages: one for link filtering and another for brochure generation.

---

✅ Example Output

After running the pipeline on https://huggingface.co, you’ll get a Markdown file like:

=== Generated Company Brochure ===

# Hugging Face

## About
...

## Careers
...