Insurellm RAG Chatbot

A Retrieval-Augmented Generation (RAG) chatbot application built with LangChain, Chroma, and Gradio. This application allows users to ask questions about Insurellm company information, employees, products, and contracts using semantic search.

🚀 Features

• Automatic Vector Store Building: Automatically builds the vector database on first launch
• Semantic Search: Uses Chroma vector database for intelligent document retrieval
• Conversational Memory: Maintains conversation context across queries
• Beautiful UI: Modern Gradio interface with examples and controls
• Production Ready: Includes error handling and logging

📋 Prerequisites

• Python 3.8 or higher
• OpenAI API key
• Conda (recommended) or pip

🛠️ Installation

1. Clone the repository:

   git clone [email protected]:deXterbed/Insurellm-RAG-Assistant.git
   cd Insurellm-RAG-Assistant

2. Create a conda environment (recommended):

   conda create -n llms python=3.11
   conda activate llms

3. Install dependencies:

   pip install -r requirements.txt

4. Set up environment variables: Create a .env file in the project root:

   echo "OPENAI_API_KEY=your-api-key-here" > .env

   Or manually create .env and add:

   OPENAI_API_KEY=sk-your-actual-api-key
   

🎯 Usage

1. Run the application:

   python app.py

2. Access the interface:

   • The app will automatically open in your browser
   • Or visit http://localhost:7860

3. First Launch:

   • On first launch, the app will automatically build the vector database from the knowledge-base/ directory
   • This may take a few minutes depending on the number of documents
   • Subsequent launches will use the existing vector database

4. Ask Questions:

   • Type your question in the chat interface
   • Examples:
     • "Who is Avery Lancaster?"
     • "What is Carllm?"
     • "Tell me about Insurellm"
     • "What contracts does Insurellm have?"

📁 Project Structure

week5/
├── app.py                 # Main application file
├── requirements.txt       # Python dependencies
├── .gitignore            # Git ignore file
├── README.md             # This file
├── .env                  # Environment variables (not in git)
├── knowledge-base/       # Source documents
│   ├── company/          # Company information
│   ├── employees/        # Employee profiles
│   ├── products/         # Product descriptions
│   └── contracts/        # Contract documents
└── vector_db/           # Vector database (auto-generated)

🏗️ Architecture

Components

1. Document Loading: Loads markdown files from knowledge-base/
2. Text Splitting: Splits documents into chunks (1000 chars, 200 overlap)
3. Embedding: Uses OpenAI embeddings to create vector representations
4. Vector Store: Stores embeddings in Chroma database
5. Retrieval: Semantic search finds relevant document chunks
6. Generation: GPT-4o-mini generates answers based on retrieved context
7. Memory: Maintains conversation history for context

Flow

User Question
    ↓
Embedding (Query Vector)
    ↓
Vector Similarity Search (Chroma)
    ↓
Retrieve Top-K Relevant Chunks
    ↓
Combine with Chat History
    ↓
LLM (GPT-4o-mini) Generation
    ↓
Response + Update Memory

🔧 Configuration

Model Settings

Edit app.py to change:

• MODEL: Change the OpenAI model (default: "gpt-4o-mini")
• temperature: Adjust creativity (default: 0.7)
• search_kwargs: Change number of retrieved chunks (default: k=4)

Vector Store Settings

Edit the RecursiveCharacterTextSplitter parameters:

• chunk_size: Size of text chunks (default: 1000)
• chunk_overlap: Overlap between chunks (default: 200)

📝 Example Queries

• Employee Information: "Who is Avery Lancaster?", "Tell me about Alex Chen"
• Product Information: "What is Carllm?", "Describe Rellm"
• Company Information: "What is Insurellm?", "Tell me about the company"
• Contracts: "What contracts does Insurellm have?", "Who is TechDrive Insurance?"

🐛 Troubleshooting

Vector Store Not Building

• Ensure knowledge-base/ directory exists with markdown files
• Check that you have write permissions in the project directory
• Review logs for specific error messages

API Key Issues

• Verify .env file exists and contains OPENAI_API_KEY
• Check that your API key is valid and has credits
• Ensure no extra spaces or quotes in the .env file

Memory Issues

• Use "Reset Memory" button to clear conversation history
• Restart the app if memory becomes too large

📚 Technologies Used

• LangChain: RAG pipeline and chain orchestration
• Chroma: Vector database for embeddings
• OpenAI: Embeddings and LLM (GPT-4o-mini)
• Gradio: Web interface
• Python: Core language

🤝 Contributing

Feel free to submit issues, fork the repository, and create pull requests!

📄 License

This project is for educational purposes.

🙏 Acknowledgments

• Built as part of LLM tutorials
• Uses LangChain's ConversationalRetrievalChain
• Inspired by RAG best practices

---

Note: Make sure to keep your .env file private and never commit it to version control!