🧠 PubMed RAG - Medical Research Assistant

🩺 Business Problem

Clinicians, researchers, and data scientists often need up-to-date insights from PubMed but end up manually reading dozens of abstracts or forming complex search queries. This slows down literature review and clinical decision-making.

---

🎯 Objective

Provide a chat-style assistant that:

1. Pulls the most relevant PubMed abstracts for a given topic in real-time.
2. Leverages Retrieval-Augmented Generation (RAG) so GPT-4o mini can answer follow-up questions grounded in those abstracts.
3. Available in both terminal and modern web interface.

---

⚙️ How It Works (High-Level)

User ⇄ Chat Interface (Terminal or Web) -> PubMed fetch (Entrez) -> Text chunking + BioBERT embedding -> ChromaDB vector storage -> GPT-4o answer

---

🚀 Quick Start

1. Clone & Install

git clone https://github.com/Rithvik-katakamm/PubMedRAG.git
cd PubMedRAG

# Create and activate virtual environment
python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

---

2. Set Your API Keys

Create environment variables or a .env file:

OPENAI_API_KEY=sk-...
ENTREZ_EMAIL=your_email@example.com

---

3. Choose Your Interface

🌐 Web Interface (Recommended)

Beautiful Gemini-inspired dark theme with modern UI:

./run_web.sh
# Then open: http://localhost:8000

Features:

• Dark theme with blue accents (like Gemini)
• Real-time text streaming
• Collapsible session sidebar
• Floating action buttons for chunks/metrics
• Elegant popups for viewing retrieved chunks
• Session management and history

💻 Terminal Interface

Rich console-based interface:

./run.sh
# or
python src/app.py

Features:

• Beautiful terminal UI with Rich library
• Session persistence
• All slash commands (/chunks, /clear, /help)
• Progress indicators

---

🎨 Web Interface Preview

The web interface features:

• Centered chat like Gemini
• Collapsible sidebar for session management
• Topic indicator in header
• Floating action buttons for:
  • 📄 View retrieved chunks
  • 📊 Data loading metrics
  • 🗑️ Clear data
  • ➕ New session
• Elegant modals for chunks and metrics
• Real-time streaming responses

---

🛠️ Tech Stack

Core

• Python 3.8+
• OpenAI GPT-4o mini
• ChromaDB (vector database)
• BioBERT (medical embeddings)
• NCBI Entrez (PubMed API)

Web Version

• FastAPI (backend)
• WebSockets (real-time communication)
• Modern HTML/CSS/JavaScript
• Dark theme with responsive design

Terminal Version

• Rich (beautiful console output)
• Click (command-line interface)

---

📁 Project Structure

rag_pubmed/
├── src/
│   ├── web_app.py          # FastAPI web application
│   ├── app.py              # Terminal application  
│   ├── static/             # Web UI assets
│   │   ├── style.css       # Gemini-inspired styling
│   │   └── script.js       # Frontend JavaScript
│   ├── core/               # RAG engine
│   │   ├── rag_engine.py   # Main orchestrator
│   │   ├── vector_db.py    # ChromaDB manager
│   │   ├── retrieval.py    # PubMed fetching
│   │   ├── chunking.py     # Text processing
│   │   └── embeddings.py   # BioBERT embeddings
│   ├── chat/               # Session management
│   └── ui/                 # Terminal UI
├── run_web.sh              # Web app launcher
├── run.sh                  # Terminal app launcher
└── requirements.txt        # Dependencies

---

🔧 Configuration

Environment Variables

• OPENAI_API_KEY - Your OpenAI API key
• ENTREZ_EMAIL - Email for NCBI Entrez API

Performance Tuning

The system is optimized for M2 Macs with:

• Max 5000 vectors per topic (sub-100ms retrieval)
• HNSW indexing for fast similarity search
• Automatic FIFO cleanup

---

💡 Usage Examples

Starting a New Session

1. Web: Click ➕ button → Enter email and topic
2. Terminal: Run app → Enter email and topic

Asking Questions

• "What are the latest treatments for diabetes?"
• "How effective is immunotherapy for lung cancer?"
• "What are the side effects of metformin?"

Viewing Retrieved Sources

• Web: Click 📄 button for elegant popup
• Terminal: Type /chunks or c

Session Management

• Web: Use collapsible sidebar
• Terminal: Use /sessions command

---

🧼 Notes

• Sessions are persistent across restarts
• Vector data is stored in ./data/chroma_db/
• Each topic maintains its own collection
• FIFO cleanup prevents memory issues
• Works offline once data is loaded

---

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

---

📄 License

MIT License - see LICENSE file for details.

---

Switch to Ollama: You can replace OpenAI with Ollama for fully local operation!