MCP Rag Search Server

🚀 Custom MCP Server with RAG and Multi-Search Functionality

A custom MCP (Model Call Protocol) server that offers Retrieval Augmented Generation (RAG) capabilities based on LlamaIndex and supports multi-network searches via Google's Gemini 2.0 API and Linkup.

✨ Features

• RAG workflow using local documents.
• Multiple network search capabilities:
  • Advanced AI-driven search using Google's Gemini 2.0 (preferred).
  • Traditional network search using Linkup.
• Built with FastMCP.

📦 Installation

Prerequisites

• Python 3.8 or higher.
• Ollama installed locally with the DeepSeek model (or modify to use your preferred model).
• Google Gemini API key (obtain at https://ai.google.dev/).
• Linkup API key (optional).

Installation Steps

1. Clone this repository:

   git clone <repository-url>
   cd own-mcp-server
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Set up environment variables (create a .env file):

   # Required API keys
   GEMINI_API_KEY=your_gemini_api_key_here
   LINKUP_API_KEY=your_linkup_api_key_here
   
   # Optional configuration
   OLLAMA_HOST=http://localhost:11434
   

4. Add documents to the data directory (it will be created automatically if it doesn't exist).

Start the Server

Run the following command to start the server:

python server.py

💻 Usage Examples

The server provides the following tools:

1. web_search: Use the best available search method (prefer Gemini 2.0 AI and fall back to Linkup if necessary).
2. gemini_search: Search using Google's Gemini 2.0 AI.
3. linkup_search: Search using Linkup.
4. rag: Query your local documents.

📚 Documentation

This project uses the following libraries:

• llama-index - Provides core RAG functionality.
• ollama - Local LLM integration.
• Google Generative AI SDK - Gemini 2.0 integration.
• Linkup SDK - Network search functionality.
• FastMCP - MCP server implementation.
• python-dotenv - Environment variable management.
• nest-asyncio - Asynchronous support.

🔧 Troubleshooting

When encountering issues:

1. Ensure that Ollama is installed and running correctly.
2. Pull the DeepSeek model: ollama pull deepseek-r1:1.5b.
3. If you encounter Python 3.13 compatibility issues, consider downgrading to Python 3.11 or 3.10.
4. Verify that your API keys are correct and have the necessary permissions.
5. For Gemini 2.0 related issues, ensure that your API key can access the latest models.