Medical Agentic RAG System

A Simple web interface for the Medical Agentic RAG (Retrieval-Augmented Generation) system built with Flask, featuring an intelligent routing system that dynamically selects the best data source for medical queries.

🌟 Features

• Intelligent Routing: Automatically determines whether to search Medical Q&A database, Medical Device Manuals, or perform Web Search
• Real-time Workflow Visualization: See the decision-making process in action
• Context Relevance Check: Ensures retrieved information is relevant to the query
• Beautiful Modern UI: Dark-themed, responsive design with smooth animations
• Example Queries: Quick-start with pre-populated medical questions
• Metadata Display: Shows route decision, information source, and relevance status

🚀 Getting Started

Prerequisites

• Python 3.12+
• All dependencies from the original agentic_RAG.py project
• Flask web framework

Installation

1. Install dependencies:

pip install -r requirements.txt

2. Ensure your environment variables are set:

   • OpenAI API key
   • Google Serper API key (for web search)

3. Make sure ChromaDB collections are initialized:

   • The app requires collection1 (Medical Q&A) and collection2 (Medical Device Manuals) from chromaDB.py

Running the Application

1. Start the Flask server:

python app.py

2. Open your browser and navigate to:

http://localhost:5000

3. Start asking medical questions!

📁 Project Structure

rag_agentic/
├── app.py                          # Flask backend server
├── agentic_RAG.py                 # Core agentic RAG logic
├── chromaDB.py                    # Vector database collections
├── results_openai.py              # LLM integration
├── requirements.txt               # Python dependencies
├── templates/
│   └── index.html                 # Main web interface
└── static/
    ├── style.css                  # Beautiful styling
    └── script.js                  # Interactive functionality

🎯 How It Works

1. User Input: Enter a medical query in the text area
2. Router Decision: AI decides which data source to use:
   • Retrieve_QnA: General medical knowledge, symptoms, treatments
   • Retrieve_Device: Medical devices, manuals, instructions
   • Web_Search: Recent news, brand names, external data
3. Context Retrieval: Fetches relevant information from the chosen source
4. Relevance Check: Verifies if the retrieved context is relevant
5. Response Generation: Creates an informed answer using the context
6. Display Results: Shows the answer with full metadata and workflow visualization

🎨 UI Features

• Dark Modern Theme: Easy on the eyes for medical professionals
• Responsive Design: Works on desktop, tablet, and mobile
• Smooth Animations: Elegant transitions and loading states
• Keyboard Shortcuts: Press Ctrl/Cmd + Enter to submit queries
• Example Queries: Quick-start buttons for common questions
• Workflow Visualization: See each step of the agentic process

📝 Example Queries

Try these queries to see the system in action:

1. "What are the treatments for Kawasaki disease?"
2. "What are the usage of Dialysis Machine Device?"
3. "What are medicines/treatment for COVID?"
4. "What's the export duty on medical tablets on India by USA in 2025?"

🔧 Configuration

Modifying the Workflow

The agentic workflow is defined in agentic_RAG.py. You can customize:

• Routing logic in the router() function
• Relevance checking criteria in check_context_relevance()
• Number of results retrieved (n_results parameter)
• Response length limit in the prompt

Styling Customization

Edit static/style.css to customize:

• Color scheme (CSS variables in :root)
• Layout and spacing
• Animations and transitions
• Responsive breakpoints

🐛 Troubleshooting

Port Already in Use

If port 5000 is already in use, change it in app.py:

app.run(debug=True, host='0.0.0.0', port=8080)

ChromaDB Not Found

Ensure your ChromaDB collections are properly initialized:

python chromaDB.py

API Keys Not Set

Set your environment variables:

export OPENAI_API_KEY="your-key-here"
export SERPER_API_KEY="your-key-here"

🔒 Security Notes

• Always use environment variables for API keys (never hardcode)
• Consider adding rate limiting for production use
• Implement user authentication if deploying publicly
• Use HTTPS in production environments

🙏 Acknowledgments

• Built with LangGraph for workflow orchestration
• Powered by OpenAI for language understanding
• ChromaDB for efficient vector storage
• Google Serper for web search capabilities