Gemini模型连接服务

Gemini-MCP Bridge

An advanced Model Context Protocol (MCP) server that connects Google's Gemini models to your development environment, providing GitHub Copilot (and other MCP-compatible tools) with powerful repository exploration and AI-driven architectural capabilities.

🚀 Key Capabilities

This MCP server exposes several high-signal tools to your AI agent:

🧠 Gemini Intelligence

• plan_task: Generates a detailed, step-by-step implementation plan for complex coding tasks.
• review_patch: Performs a deep review of code diffs, identifying logic bugs, security risks, and maintainability issues.
• ask_gemini: A general-purpose technical assistant for any coding or architectural questions.

🔍 Repository Exploration & Semantic Search

• list_repo_files / read_repo_file: Safely explore the project structure and content with smart safeguards.
• search_repo_code: Fast, recursive grep-style search across your codebase.
• index_repository: Indexes your files using Gemini embeddings to enable semantic search.
• semantic_search_code: Uses a vector database (LanceDB) to find code conceptually related to your query.

📐 Architectural Mapping

• map_architecture: Builds a high-level graph of your project. It uses AI to generate "Functional Signatures" for each file and maps their structural dependencies (imports).
• get_graph_data: Exports the architectural graph as JSON for visualization in the built-in UI.

📊 Graph Visualization UI

The project includes a built-in web application to visualize your code's architecture.

1. Launch the UI:
   复制

   cd ui
   python main.py
   

2. Access the Dashboard: Open http://localhost:8000 in your browser.
3. Dynamic Loading: Use the path input to load the .mcp_data from any project on your system.

🛡️ Security & Observability

• Persistent Memory: A local SQLite database (.mcp_data/history.db) stores a detailed history of every tool call, request, and response for task continuity.
• Vector Store: Uses LanceDB for high-performance, local vector storage of your repository's code.
• Path Validation: Strict safeguards prevent path traversal attacks and access to sensitive system files.
• Structured Logging: All activities are logged to stderr, mcp_server.log, and the SQLite database.

⚙️ Setup

1. Install Dependencies:

   复制

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt
   

2. Configure Environment: Create a .env file in the root directory:

   复制

   GEMINI_API_KEY=your_api_key_here
   

🛠️ VS Code & GitHub Copilot Integration

To use this bridge with GitHub Copilot, you must configure it via an mcp.json file.

1. Open MCP Configuration:

   • Open the Command Palette (Cmd+Shift+P on macOS or Ctrl+Shift+P on Windows/Linux).
   • Type "MCP: Open User Configuration" and press Enter.

2. Add the Gemini Bridge: Add the following configuration (adjust the paths to match your local setup):

   复制

   {
     "servers": {
       "gemini-bridge": {
         "type": "stdio",
         "command": "/home/hazem/personal_dev/mcp-gemini-bridge/venv/bin/python",
         "args": ["/home/hazem/personal_dev/mcp-gemini-bridge/server.py"],
         "env": {
           "PYTHONPATH": "/home/hazem/personal_dev/mcp-gemini-bridge",
           "GEMINI_API_KEY": "your_api_key_here"
         }
       }
     }
   }
   

3. Use in Copilot:

   • Open GitHub Copilot Chat.
   • Switch to Agent Mode.
   • Click the Tools (wrench icon) to verify "Gemini Bridge" is connected and active.

🧪 Development & Testing

Run the server locally to verify the protocol:

PYTHONPATH=. ./venv/bin/python server.py

Logs will appear in stderr and mcp_server.log.