MCP Server Using Gemini

🚀 Gemini MCP Server - GeminiRouter Using Gemini Flash 1.5 API

This project is an implementation that combines the Model Context Protocol (MCP) with the Google Gemini Flash 1.5 API, aiming to demonstrate how modular AI services can collaborate through a centralized router.

🧐 What is MCP and How Does It Work Here?

The Model Context Protocol (MCP) is an architectural pattern designed to enable multiple specialized AI agents or services to communicate in a modular and context-aware manner through lightweight requests. This project applies MCP principles to build a scalable and intelligent AI system with dedicated microservices coordinated by a central router.

Each service (or "context") - whether for chat, web search, weather information, deep reasoning, or Retrieval Augmented Generation (RAG) - performs a specific function and communicates via lightweight requests, all intelligently routed by the router/client.

↺ Overview of the Workflow

Here's how the system works from input to output:

1. User Input: The user sends a query from the front-end UI.

2. Routing Logic:

   • The router receives the query and analyzes the intent.
   • Based on keywords, context, or past interactions, it selects the appropriate server (chat, search, RAG, thinking, etc.) to handle the query.

3. Service Processing:

   • Chat Server: If the query is conversational or informal, it is routed here for a quick response.
   • Search Server: For queries requiring real-time or factual data (e.g., "What is the capital of Sweden?"), the router calls the Gemini API through this module.
   • RAG Server: For complex questions requiring comprehensive answers based on documents or external knowledge, the RAG module uses a retrieval layer + Gemini for answer generation.
   • Thinking Server: For logical reasoning or multi-step problem-solving, the query is passed here for in-depth thinking processing.
   • Weather (via Search): The search server is also integrated with the OpenWeather API to handle natural language weather queries, such as "What's the weather like in Tokyo?"

4. Response Aggregation:

   • The selected module processes the request and sends its response back to the router.
   • The router can optionally combine responses from multiple modules (if the query requires it).

5. Front-end Output: The final response is returned to the user interface and presented in a clean, conversational format.

✨ Features

• Modular AI services
• Central context router
• Integration with Gemini Flash 1.5 API
• Easily scalable architecture
• High-performance processing

🔧 Prerequisites

To run this project, you need the following environment:

• Operating System: Linux/macOS/Windows
• Programming Language: Python 3.8+
• Dependencies:
  • python-dotenv
  • requests

📋 Environment Variables

Create a .env file and add the following content:

GEMINI_API_KEY=your_api_key_here
OPENWEATHER_API_KEY=your_api_key_here

🏷️ Project Structure

gemini-router/
├── router.py          # Main router logic
├── services/         # Individual service modules
│   ├── chat_service.py
│   ├── search_service.py
│   ├── rag_service.py
│   └── weather_service.py
└── config/           # Configuration files and environment variable handling
    ├── settings.py
    └── .env

📦 Installation

📥 Install Dependencies

pip install python-dotenv requests

🔧 Start the Server

python router.py

🚀 Future Plans

• Implement automatic scaling functionality
• Support more weather APIs
• Add logging and monitoring
• Develop a graphical management interface

📄 Notes

⚠️ Important Note

The API keys in the configuration file .env need to be replaced with actual API keys.

💡 Usage Tip

Use developer tools for debugging.