A comprehensive, modular healthcare chatbot system built using LangGraph, designed to deliver intelligent, context-aware, and collaborative medical assistance. This project leverages a supervisor-agent architecture, with each agent specializing in a key healthcare domain. The system supports both command-line and web-based interactions, robust state management, and detailed logging for traceability and debugging.

• Project Overview
• Project Structure
• Agents Description
• Setup Instructions
• Running the Application
  • Command-Line Interface
  • FastAPI Backend
  • Frontend
• Environment Variables
• Logging
• API Endpoints
• Demo

This chatbot system orchestrates multiple specialized AI agents to provide:

• Symptom checking and preliminary triage
• Drug research and interaction information
• Crisis detection and escalation
• General healthcare guidance and follow-up
• Centralized supervision and intelligent routing

The supervisor agent dynamically routes user queries to the most appropriate agent, ensuring responses are accurate, comprehensive, and contextually relevant. All interactions and agent handoffs are logged for transparency and debugging.

Multi-Agent-Medical-Chatbot/

├── Agents/
│   ├── __init__.py
│   ├── crisis_agent.py
│   ├── guidance_agent.py
│   ├── research_agent.py
│   ├── supervisor_agent.py
│   └── symptoms_agent.py
├── Graph/
│   ├── __init__.py
│   └── main_graph.py
├── Loggers/
│   ├── __init__.py
│   ├── debugging.log
│   └── logger_config.py
├── State/
│   ├── __init__.py
│   └── state.py
├── medi-stream-chat/
│   ├── next.js code

Agents communicate via the LangGraph workflow defined in `Graph/main_graph.py`, with all state and context managed in `State/state.py`.

1. Clone the Repository

   git clone <your-repo-url>
   cd MEDICAL\ MULTI\ AGENT

2. Create a virtual environment:

# Create and activate a virtual environment using uv
uv venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate.bat or .venv\Scripts\Activate.ps1

Run the chatbot in CLI mode:

python main.py

• Interact directly with the multi-agent system from your terminal.
• Ideal for testing, debugging, or headless deployments.

Start the FastAPI server:

python app.py

• Exposes RESTful endpoints for chat, state, and agent routing.
• Integrates with the frontend and supports external API calls.

Launch the web interface:

cd medi-stream-chat
npm install
npm audit fix          # Optional: fix vulnerabilities
npm run build          # Optional: build for production
npm run dev            # Start the development server (default: http://localhost:3000)

• Next.js/React-based UI for user-friendly chat experience.
• Connects to the FastAPI backend for real-time multi-agent responses.

Define the following in your `.env` file:

SERPER_API_KEY = ""
GROQ_API_KEY = ""
MODEL_NAME = "llama-3.3-70b-versatile"
TAVILY_API_KEY = ""
GOOGLE_API_KEY = ""

• Required for LLM access and medical data APIs.

• All agent interactions, routing decisions, and errors are logged to `Loggers/debugging.log`.
• Logging configuration is managed in `Loggers/logger_config.py`.
• Use logs to trace conversation flows, debug issues, and audit system behavior.

• See `app.py` for detailed endpoint implementations and usage examples.

You can upload or link a demo video here to showcase the chatbot in action:

For support or feature requests, please open an issue in the repository.