MCPServerDemo

# MCP Server A WebSocket-based MCP (Model Control Protocol) server for handling model requests and providing responses. ## Project Structure ``` MCPServerDemo/ ├── src/ # Source code directory │ └── mcp/ # MCP implementation │ ├── handlers/ # Request handlers │ ├── models/ # Data models │ ├── protocol/ # Protocol definitions │ └── server/ # Server implementation ├── client/ # Client example │ └── index.html # HTML client for testing ├── .env # Environment variable configuration (needs to be created from .env.example) ├── .env.example # Environment variable example ├── main.py # Application entry point └── README.md # Project description ``` ## Installation This project uses uv for dependency management. ```bash # Install dependencies uv pip install fastapi uvicorn pydantic pydantic-settings ``` ## Configuration 1. Copy the environment variable example file and configure it: ```bash cp .env.example .env ``` 2. Edit the `.env` file to set your configuration. ## Running ```bash # Start the server python main.py ``` The server will run at `http://localhost:8000`, with the WebSocket endpoint at `ws://localhost:8000/ws`. ## Testing Client Open the `client/index.html` file in a browser to test the MCP server. ## MCP Protocol The MCP protocol is a WebSocket-based protocol for controlling and managing AI models. ### Request Format ```json { "id": "Request ID", "type": "request", "action": "Action Name", "messages": [ { "role": "user", "content": "Message Content" } ], "parameters": { "Parameter Name": "Parameter Value" } } ``` ### Response Format ```json { "id": "Request ID", "type": "response", "action": "Action Name", "messages": [ { "role": "assistant", "content": "Response Content" } ], "metadata": { "Metadata Name": "Metadata Value" } } ``` ### Error Format ```json { "id": "Request ID", "type": "error", "action": "Action Name", "error": "Error Message", "details": { "Detail Name": "Detail Value" } } ``` ### Event Format ```json { "id": "Event ID", "type": "event", "event": "Event Name", "data": { "Data Name": "Data Value" } } ``` ## Supported Actions - `chat`: Chat request - `completion`: Text completion request ## Extension You can extend the MCP server in the following ways: 1. Create new handlers in the `src/mcp/handlers/` directory 2. Register new handlers in `main.py` 3. Customize protocol implementations in the `src/mcp/protocol/` directory