Rohan7530/AI-Medical-Triage-System

title: Medical Triage System emoji: 🩺 colorFrom: blue colorTo: green sdk: docker pinned: false

Medical Triage MCP Server + LangGraph Integration

Summary

The Medical Triage System is an intelligent platform designed to streamline patient assessment and healthcare resource allocation. By integrating a Model Context Protocol (MCP) server with advanced LangGraph workflows, the application automates the triage process from symptom intake to appointment booking. Patients can input their symptoms, which are then analyzed using a medical knowledge base and AI-powered reasoning. The system determines the urgency of each case, matches patients with suitable doctors or specialists, and manages appointment scheduling and notifications. Its modular architecture ensures scalability, maintainability, and easy integration with external large language models (LLMs) for enhanced natural language understanding. The platform is suitable for clinics, hospitals, and telemedicine providers seeking to optimize patient flow, reduce manual workload, and improve healthcare outcomes through intelligent automation.

Features

• MCP Server: Exposes medical resources and tools for AI agents
• LangGraph Workflow: Orchestrates the complete triage process
• Medical Knowledge Base: Symptoms, conditions, and triage protocols
• Doctor Database: Available doctors and specialists
• Appointment System: Booking and availability management
• Notification System: Patient communication

Quick start (local)

python3.12 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Set env for LLM
export USE_LLM=true
export LLM_MODEL=qwen2.5:7b
export LLM_BASE_URL=https://your-public-llm.example.com
# Run frontend
streamlit run frontend.py

Deploy on Hugging Face Spaces (Docker SDK)

1. Create a Space (SDK: Docker, Public).
2. Push this repo (must include Dockerfile, requirements.txt, frontend.py).
3. In Space → Settings → Variables and secrets, set:
   • USE_LLM=true
   • LLM_MODEL=qwen2.5:7b
   • LLM_BASE_URL=https://your-public-llm.example.com
4. Spaces will build and run the container. Open the Space URL.

Environment variables

• USE_LLM (default true): set to false to force fallback parsing
• LLM_MODEL (default qwen2.5:7b)
• LLM_BASE_URL (required when USE_LLM=true)

Project Structure

Medical_mcp/
├── src/
│   ├── mcp_server/
│   │   ├── __init__.py
│   │   ├── server.py
│   │   ├── resources.py
│   │   └── tools.py
│   ├── langgraph_workflow/
│   │   ├── __init__.py
│   │   ├── workflow.py
│   │   ├── nodes.py
│   │   └── state.py
│   ├── models/
│   │   ├── __init__.py
│   │   └── data_models.py
│   └── utils/
│       ├── __init__.py
│       └── helpers.py
├── tests/
├── examples/
├── frontend.py
├── main.py
├── requirements.txt
├── Dockerfile
└── README.md

Installation

1. Create a Python 3.12 virtual environment:

python3.12 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Make sure Ollama is running with the llama3:8b model:

ollama pull llama3:8b
ollama serve

Usage

Run the main application:

python main.py

Architecture

The system follows a modular architecture:

1. MCP Server (src/mcp_server/): Handles medical resources and tools
2. LangGraph Workflow (src/langgraph_workflow/): Orchestrates the triage process
3. Data Models (src/models/): Defines data structures
4. Utilities (src/utils/): Helper functions

Workflow

1. Parse patient symptoms
2. Assess triage urgency using medical knowledge
3. Find appropriate doctors
4. Book appointments
5. Send notifications