🏥 Clinical Summary RAG Application

An AI-powered medical documentation system that generates comprehensive discharge summaries using Retrieval-Augmented Generation (RAG) architecture. This application leverages advanced language models, medical domain embeddings, and vector databases to assist healthcare professionals in creating accurate, structured clinical documentation.

📋 Project Description

The Clinical Summary RAG Application is a sophisticated healthcare documentation tool designed to streamline the process of generating medical discharge summaries. The system combines:

• RAG Architecture: Retrieval-Augmented Generation for context-aware responses
• Medical Domain Embeddings: Bio ClinicalBERT for medical text understanding
• Fast LLM Inference: Groq API with LLaMA 4 Maverick model for rapid generation
• Vector Search: ChromaDB for semantic similarity search across patient records
• Modern Web Interface: Streamlit frontend with FastAPI backend for optimal performance

The application enables healthcare professionals to:

• Search and retrieve patient information from MongoDB
• Generate structured discharge summaries automatically
• Find similar patient cases using semantic search
• Interact with an AI assistant for clinical queries
• Export summaries in multiple formats (TXT, DOCX, PDF)
• Customize summary templates based on insurance requirements

🎯 Use Cases

1. Automated Discharge Summary Generation

• Problem: Manual creation of discharge summaries is time-consuming and prone to inconsistencies
• Solution: Automatically generate comprehensive, structured discharge summaries from patient data
• Benefit: Reduces documentation time by 60-80%, ensures consistency, and minimizes errors

2. Clinical Decision Support

• Problem: Healthcare providers need quick access to similar cases for reference
• Solution: Semantic search across historical patient records to find similar cases
• Benefit: Provides evidence-based references for treatment planning and decision-making

3. AI-Powered Clinical Assistant

• Problem: Healthcare professionals need quick answers to patient-specific questions
• Solution: Conversational AI agent that answers questions based on patient context
• Benefit: Instant access to patient information and clinical insights

4. Template-Based Documentation

• Problem: Different insurance providers require different summary formats
• Solution: Upload PDF templates to generate summaries matching specific requirements
• Benefit: Ensures compliance with various documentation standards

5. Knowledge Base Enhancement

• Problem: Generated summaries should improve future recommendations
• Solution: Feedback loop that adds generated summaries back to the knowledge base
• Benefit: Continuously improving system accuracy and relevance

🏗️ Architecture

┌─────────────┐         ┌──────────────┐         ┌─────────────┐
│  Streamlit  │ ──────> │   FastAPI    │ ──────> │   Groq API  │
│  Frontend   │  HTTP   │   Backend    │  HTTP   │   (LLM)     │
│  (Port 8501)│         │  (Port 8000) │         │  (Cloud)    │
└─────────────┘         └──────────────┘         └─────────────┘
                              │
                              ├──> MongoDB (Patient Records)
                              └──> ChromaDB (Vector Search)

🚀 Key Features

• 🤖 AI-Powered Summary Generation: Uses Groq API with LLaMA 4 Maverick for fast, accurate summaries
• 💬 Conversational AI Agent: Interactive chat interface for clinical queries
• 🔍 Semantic Case Search: Find similar patient cases using Bio ClinicalBERT embeddings
• 📊 Modern UI: Beautiful dark theme interface with smooth animations
• ⚡ High Performance: FastAPI async backend for 40-60% faster responses
• 📄 Multiple Export Formats: Download summaries as TXT, DOCX, or PDF
• 🎨 Template Support: Custom PDF templates for different documentation requirements
• 💾 Feedback Loop: Continuous learning from generated summaries

🛠️ Technology Stack

• Frontend: Streamlit with modern dark theme UI
• Backend: FastAPI with async/await for high performance
• LLM: Groq API with meta-llama/llama-4-maverick-17b-128e-instruct
• Embeddings: Bio ClinicalBERT (medical domain-specific)
• Vector DB: ChromaDB for similarity search
• Database: MongoDB for patient records
• AI Agent: AutoGen for conversational interface

📦 Installation

Prerequisites

• Python 3.8 or higher
• Groq API key (Get one here)
• MongoDB connection (cloud or local)

Quick Start

1. Clone the repository

   git clone https://github.com/rajiv-rane/clinical-summary-rag-v2.git
   cd clinical-summary-rag-v2

2. Install dependencies

   cd ingestion-phase
   pip install -r requirements.txt

3. Set up environment variables Create a .env file in the ingestion-phase directory:

   GROQ_API_KEY=your_groq_api_key_here

4. Start the FastAPI backend

   python start_api.py

5. Start the Streamlit frontend (in a new terminal)

   streamlit run app.py

6. Access the application Open your browser to http://localhost:8501

For detailed setup instructions, see ingestion-phase/README.md

📚 Documentation

• Main README - Complete setup and usage guide
• FastAPI Documentation - Backend API details
• Groq Setup Guide - API key configuration

🔒 Security Considerations

• Never commit .env files containing API keys
• Use environment variables for sensitive configuration
• Implement authentication/authorization for production use
• Restrict CORS origins in production
• Enable HTTPS for production deployments

📝 License

[Add your license information here]

👥 Authors

• Rajiv Rane - GitHub

🙏 Acknowledgments

• Bio ClinicalBERT model by Emily Alsentzer
• Groq API for fast LLM inference
• FastAPI by Sebastián Ramírez
• Streamlit team
• ChromaDB team

---

Version: 2.0.0
Last Updated: 2024