RAG MCP Server for GitHub Copilot

A Model Context Protocol (MCP) server that provides RAG (Retrieval-Augmented Generation) capabilities for GitHub Copilot. This allows you to process PDFs, create vector embeddings, and retrieve relevant context without using external LLM APIs.

Features

• PDF Processing: Extract text from PDF documents and chunk them for indexing
• Vector Embeddings: Generate embeddings using Hugging Face transformers
• Semantic Search: Find relevant document chunks using vector similarity
• MCP Integration: Connect directly with GitHub Copilot through MCP protocol

Setup

1. Install dependencies:

npm install

2. Build the project:

npm run build

3. Configure GitHub Copilot to use this MCP server by adding to your VS Code settings.json:

{
  "mcp": {
    "inputs": [],
    "servers": {
      "rag-server": {
        "command": "node",
        "args": [
          "D:\\YASH\\Projects GGC\\capgemini projects\\mcp part 2\\dist\\index.js"
        ],
        "env": {}
      }
    }
  }
}

Usage

Once configured, you can use these tools in GitHub Copilot:

Add PDF Documents

@rag-server add_pdf file_path="/path/to/document.pdf"

Search Documents

@rag-server search_documents query="your search query" max_results=5

Get Context for Copilot

@rag-server get_context query="what you're working on"

How It Works

1. PDF Processing: Uses pdf-parse to extract text from PDFs and chunks it into manageable pieces
2. Embeddings: Generates vector embeddings using the all-MiniLM-L6-v2 model from Hugging Face
3. Vector Store: Uses FAISS for efficient similarity search
4. MCP Protocol: Exposes functionality through MCP tools that GitHub Copilot can call

Architecture

• src/pdf-processor.ts: Handles PDF text extraction and embedding generation
• src/vector-store.ts: Manages vector storage and similarity search
• src/index.ts: Main MCP server implementation
• src/mcp-config.json: Configuration template for GitHub Copilot

This system runs locally without requiring external API keys or cloud services.