RAG-Enriched Voice Agent

This sample project demonstrates a Retrieval-Augmented Generation (RAG) enabled voice agent using LiveKit Agents 1.0. The example scrapes the LiveKit docs site, builds a local index, and then provides that data on demand to the assistant via a function tool.

Prerequisites

• Python 3.9 or higher
• OpenAI API key
• Deepgram API key
• LiveKit server

Installation

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

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Create a .env file in the project root with your API keys:

   OPENAI_API_KEY=your_openai_api_key
   DEEPGRAM_API_KEY=your_deepgram_api_key
   LIVEKIT_URL=your_livekit_url
   LIVEKIT_API_KEY=your_livekit_api_key
   LIVEKIT_API_SECRET=your_livekit_api_secret
   

Project Structure

• main.py: Main agent implementation
• scrape_docs.py: Scraper for the LiveKit docs site
• build_rag_data.py: Script to build the RAG database from scraped docs
• rag_db_builder.py: Database builder implementation
• rag_handler.py: RAG processing logic
• data/: Directory for vector database files

Usage

1. Scrape the docs site:

2. Build the RAG database:

3. Download model files:

   python main.py download-files

4. Run the agent:

The agent will start and be ready to handle voice interactions. It will use the RAG system to provide contextually relevant answers to user questions.