YITAM MCP Server

Your Intelligent Task Assistant Manager - Model Context Protocol Server

Description

YITAM MCP Server provides a vector database-backed retrieval system using the Model Context Protocol. It supports both Qdrant and Chroma vector databases for efficient semantic search capabilities.

Features

• Semantic search using vector embeddings
• Support for multiple vector databases (Qdrant, Chroma)
• MCP-compliant server implementation
• TypeScript/Node.js implementation
• Built-in FastEmbed support

Prerequisites

• Node.js (LTS version)
• npm
• Either Qdrant or Chroma vector database

Installation

1. Clone the repository:

git clone [repository-url]
cd yitam-mcp

2. Install dependencies and build:

# For clean/production install:
npm run install:clean

# For development install:
npm run install:dev

3. Copy the example environment file and configure it:

cp .env.example .env

Configuration

Edit the .env file with your settings:

• COLLECTION_NAME: Your vector database collection name
• DATABASE_TYPE: Choose between 'qdrant' or 'chroma'
• QDRANT_URL: Your Qdrant server URL (if using Qdrant)
• QDRANT_API_KEY: Your Qdrant API key (if using Qdrant)
• CHROMA_URL: Your Chroma server URL (if using Chroma)
• GEMINI_API_KEY: Your Google Gemini API key for embeddings
• GEMINI_MODEL: Embedding model to use (default: gemini-embedding-001)
• GEMINI_EMBEDDING_DIMENSIONS: Optional dimension reduction (default: 3072 for gemini-embedding-001)

Usage

Development

npm run dev

Production

npm start

License

MIT