mcp-docs-rag
The mcp-docs-rag project is an MCP server for managing and querying locally stored documents using Retrieval-Augmented Generation. It supports various features like querying documents, listing available files, and adding new documents from Git repositories or plain text. It utilizes LLMs and Google's Gemini API for document indexing and querying.
mcp-docs-rag MCP Server
RAG (Retrieval-Augmented Generation) for documents in a local directory
This TypeScript-based MCP server implements a RAG system for querying documents stored locally. Key features include document queries using LLMs, accessing documents via docs:// URIs, and supporting Git repositories or text files. Users can list, query, and add documents with the server automatically handling indexing using Google's Gemini API.
Features
- Resources: Access documents through specific URIs; supports Git repositories and text files.
- Tools:
list_documents: Lists all documents in the directory.rag_query: Queries documents using RAG.add_git_repository: Clones or updates Git repositories.add_text_file: Downloads text files to the directory.
- Prompts: Guides users on document and RAG functionality usage.
Setup
- Local directory setup for document storage with optional configuration.
- Uses Google's Gemini API; requires API key setup.
Usage
- Supports document addition from GitHub or URLs and querying stored documents. Automates document indexing.