Which integrations are available for this server?

Uses OpenAI API compatible LLM server (such as vLLM) to serve the Lingshu medical AI model for medical image analysis, structured report generation, and medical Q\&A capabilities

How do I use Lingshu FastMCP Medical AI Service?

1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Lingshu FastMCP Medical AI Service analyze this chest X-ray for pneumonia signs" That's it! The server will respond to your query, and you can continue using it as needed. Here is a step-by-step guide with screenshots.

Lingshu FastMCP Medical AI Service

This project implements a FastMCP server for the Lingshu medical AI model and a corresponding client for testing and integration.

Components

mcp_server_lingshu.py: FastMCP server wrapping the Lingshu model
mcp_client_lingshu.py: Test client demonstrating interaction with the Lingshu FastMCP server

Server Features

Medical image analysis
Structured medical report generation
Medical Q&A

Prerequisites

FastMCP framework
OpenAI API compatible LLM server (e.g., vLLM)
Required Python packages (install via pip install -r requirements.txt)

Setup

Clone the repository
Install dependencies: pip install -r requirements.txt

Usage

Use vLLM to serve the Lingshu Model

vllm serve lingshu-medical-mllm/Lingshu-7B --dtype float16 --api_key api_key --port 8000 --max-model-len 32768

Wrap the server with FastMCP

export LINGSHU_SERVER_URL="http://localhost:8000/v1" export LINGSHU_SERVER_API="api_key" export LINGSHU_MODEL="lingshu-medical-mllm/Lingshu-7B" # the above config depends on your vllm server config python mcp_server_lingshu.py --host 127.0.0.1 --port 4200 --path /lingshu --log-level info

Try connecting Lingshu with MCP

export LLM_SERVER_URL="xxx" export LLM_SERVER_API="xxx" export LLM_MODEL="xxx" ## this is your own model python mcp_client_lingshu.py --mcp-url http://127.0.0.1:4200/lingshu # the mcp-url should depend on the mcp server you deployed in the last step