RAG-MCP: Retrieval Augmented Generation with Model Control Protocol

A modular implementation of a Retrieval Augmented Generation server using the Model Control Protocol (MCP).

Features

• Vector database integration with ChromaDB
• PDF document loading and processing
• Semantic search over document contents
• Specialized prompts for document analysis
• MCP server implementation with tools, resources, and prompts

Installation

1. Clone the repository:

git clone https://github.com/yourusername/rag-mcp.git
cd rag-mcp

2. Install the package:

pip install -e .

3. Create a .env file based on the example:

cp .env.example .env

4. Edit the .env file with your API key and other settings:

API_KEY=your_api_key_here
CHROMA_DB_PATH=./chromadb
PDF_DIRECTORY=./testing

Usage

Running the Server

python -m rag_mcp.server

Directory Structure

• rag_mcp/: Main package
  • database/: ChromaDB client and vector database operations
  • tools/: MCP tools implementation
  • resources/: Document resource handlers
  • prompts/: Specialized prompts for document analysis
• tests/: Unit tests

Development

Running Tests

python -m unittest discover

Adding New Tools

1. Create a new file in the tools directory
2. Implement your tool function
3. Update the TOOL_HANDLERS and TOOL_DEFINITIONS in tools/__init__.py

Adding New Prompts

1. Add your prompt implementation in prompts/analysis_prompts.py
2. Update the PROMPT_HANDLERS and PROMPT_DEFINITIONS

License

MIT