RAG-chatbot is a production-grade chatbot app that combines LLMs with RAG techniques. It leverages semantic text embeddings, vector search using ElasticSearch and LLM integration (via OpenAI API or open-source alternatives) to produce context-rich responses for customer service and automated content generation.

This project was some sort of PoC/demo as part of my freelance work.

Features

• LLM integration: communicate with LLMs (e.g., OpenAI GPT) for natural language understanding and generation.
• Retrieval augmentation: retrieve relevant documents from a knowledge base using vector search to augment chatbot responses.
• Text similarity & reranking: utilize SOTA text similarity methods to rerank retrieved content.
• Data preprocessing: clean and preprocess text and generate embeddings using modern transformer-based models.
• Modular architecture: clearly separated modules for LLM integration, vector database operations, retrieval augmentation and chatbot logic.
• Real-time chat interface: CLI for immediate testing and demonstration.
• Testing suite: unit tests covering major modules and components.

Architecture

For a detailed description of the system architecture, please refer to docs/architecture.md.

Installation

1. Clone the repository:

   git clone git@github.com:avrtt/RAG-chatbot.git
   cd RAG-chatbot

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate

3. Install the dependencies:

   pip install -r requirements.txt

4. Configure the application:
   Update config.py with your API keys, ElasticSearch settings and other configuration parameters as needed.

Usage

Start the chatbot by running:

python main.py

You will be prompted with a CLI where you can enter your queries. The chatbot will process your input, retrieve relevant documents and generate a context-aware response.

Testing

Run the full test suite using pytest:

pytest

Contributing

Contributions are welcome.

License

MIT