Docs RAG

🚀 MCP Documentation RAG Server

This is a Retrieval Augmented Generation (RAG) system for documents stored in a local directory. It's a TypeScript-based MCP server that enables users to query documents via a Large Language Model (LLM) and retrieve context from local repositories and text files.

🚀 Quick Start

This RAG server allows users to interact with local documents using a Large Language Model (LLM). It provides various tools to list, access, and query documents stored in a local directory.

✨ Features

Resources

• List and Access Documents: Accessed via the docs:// URI.
  • Documents can be Git repositories or text files.
  • Content is accessed using the plain text MIME type.

Tools

• list_documents: Lists all available documents in the DOCS_PATH directory.
  • Returns a formatted list of documents.
  • Displays the total number of available documents.
• rag_query: Queries documents using RAG.
  • Accepts document_id and query parameters.
  • Returns an AI-generated response with document context.
• add_git_repository: Clones a Git repository to the docs directory, with optional sparse checkout.
  • Accepts the repository_url parameter.
  • Optionally accepts the document_name parameter to customize the document name (use a simple descriptive name without the '-docs' suffix).
  • Optionally accepts a subdirectory parameter for sparse checkout of a specific directory.
  • Automatically pulls the latest changes if the repository already exists.
• add_text_file: Downloads a text file to the docs directory.
  • Accepts the file_url parameter.
  • Downloads the file using wget.

Prompts

• guide_documents_usage: Guides users on how to use documents and the RAG feature.
  • Includes a list of available documents.
  • Provides tips on using the RAG feature.

📦 Installation

Install Dependencies

npm install

Build the Server

npm run build

Development with Automatic Rebuild

npm run watch

📚 Documentation

Setup

The server requires a local directory to store documents. By default, it uses ~/docs, but you can configure a different location by setting the DOCS_PATH environment variable.

Document Structure

The document directory can contain:

• Git repositories (cloned directories)
• Plain text files with the .txt extension

Each document is individually indexed using llama-index.ts and Google's Gemini embeddings.

API Key

The server uses Google's Gemini API for document indexing and querying. You need to set your Gemini API key as an environment variable:

export GEMINI_API_KEY=your-api-key-here

You can obtain your Gemini API key on the Google AI Studio website. Add this key to your shell configuration file or the environment configuration of Claude Desktop.

Installation with Claude Desktop

To use the server with Claude Desktop, add the following configuration:

On MacOS: ~/Library/Application Support/Claude/claude_desktop_config.json On Windows: %APPDATA%/Claude/claude_desktop_config.json On Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "docs-rag": {
      "command": "npx",
      "args": ["-y", "@kazuph/mcp-docs-rag"],
      "env": {
        "DOCS_PATH": "/path/to/your/docs",
        "GEMINI_API_KEY": "your-api-key-here"
      }
    }
  }
}

💻 Usage Examples

Basic Usage

Install and Start the Server

npm install -g @kazuph/mcp-docs-rag
mcp-docs-rag

Query Documents

# Example: Query document content
mcp rag_query --id your-document-id --query "your-query-here"

Add New Documents or Repositories

# Example: Add a Git repository
mcp add_git_repository --url https://github.com/username/repository.git

# Example: Add a text file
mcp add_text_file --url https://example.com/document.txt

⚠️ Important Note

• Ensure you have the necessary permissions to access all resources and APIs.
• Regularly check for updates to maintain optimal server performance.
• Follow relevant security protocols when handling sensitive information.