Warplanes RAG

A Python Model Context Protocol (MCP) RAG server for searching information on modern warplanes.

Overview

This project implements a Retrieval Augmented Generation (RAG) system that allows users to query information about modern warplanes. The system consists of two main components:

1. Data Loader: Processes text files containing information about warplanes, chunks the text, creates embeddings, and stores them in a vector database.
2. MCP Server: Exposes an MCP tool that accepts queries, retrieves relevant information from the vector database, and generates responses using an LLM.

Requirements

• Python 3.10+
• OpenAI API key (set as environment variable OPENAI_API_KEY)

Installation

# Clone the repository
git clone https://github.com/yourusername/warplanes-rag.git
cd warplanes-rag

# Install dependencies
uv add openai mcp chromadb
uv add --dev pytest ruff

Usage

1. Load Data

First, place your text files containing information about warplanes in the data/ directory. Then run the loader script to process the data and create the vector database:

python src/loader.py

2. Start the MCP Server

Start the MCP server to expose the warplanes search tool:

python src/server.py

The server will start on localhost:8000 by default.

3. Connect to the MCP Server

You can connect to the MCP server using any MCP client. For example, using the MCP CLI:

npx -y mcp-cli@latest connect http://localhost:8000

Then you can use the warplanes_search tool to query information about warplanes:

npx -y mcp-cli@latest use warplanes_search --query "What are the capabilities of the F/A-18C?"

Project Structure

warplanes-rag/
├── data/                  # Text files containing warplane information
├── db/                    # Vector database (created by loader.py)
├── src/
│   ├── warplanes_rag/     # Core package
│   │   ├── __init__.py
│   │   ├── config.py      # Configuration settings
│   │   └── utils.py       # Utility functions
│   ├── loader.py          # Data loader script
│   └── server.py          # MCP server script
└── tests/                 # Unit tests
    ├── test_loader.py
    └── test_server.py

Configuration

Configuration settings are stored in src/warplanes_rag/config.py. You can modify these settings to customize the behavior of the system:

• CHUNK_SIZE: Size of text chunks in words (default: 200)
• CHUNK_OVERLAP: Overlap between chunks in words (default: 20)
• EMBEDDING_MODEL: OpenAI embedding model to use (default: "text-embedding-3-large")
• LLM_MODEL: OpenAI LLM model to use (default: "gpt-4o")
• SERVER_HOST: Host to bind the MCP server to (default: "localhost")
• SERVER_PORT: Port to bind the MCP server to (default: 8000)
• TOP_K_RESULTS: Number of chunks to retrieve from the vector database (default: 5)

Testing

Run the tests using pytest:

pytest

License

MIT