🔬 MedTech AI - Intelligent Biomedical Assistant

A production-ready, RAG-powered chatbot specialized in biomedical engineering topics. Built with LangChain, FAISS, and Groq LLM for intelligent document-based question answering.

📋 Table of Contents

• 🌟 Key Features
• 🏗️ System Architecture
• 🚀 Quick Start
• 📁 Project Structure
• 🐳 Docker Deployment
• 🧪 Testing
• 📝 Configuration
• 🎯 Usage Examples
• 🔧 Troubleshooting
• 📊 Performance Metrics
• 🛠️ Technology Stack

🌟 Key Features

• 🧠 RAG Architecture: Retrieval-Augmented Generation for accurate, context-based responses
• 📚 Document Processing: Automatic PDF processing with OCR fallback
• 💬 Dual Mode Operation: Technical Q&A + casual conversation support
• 🚀 Production Ready: Proper error handling, logging, and monitoring
• 🐳 Docker Support: One-command deployment with Docker Compose
• ✅ Tested: Unit tests with pytest
• 📊 Session Management: Track queries and conversation history

🏗️ Architecture

┌─────────────┐
│   User UI   │  (Streamlit)
└──────┬──────┘
│
┌──────▼──────────────┐
│   QA Chain Manager  │
│  (LangChain + Groq) │
└──────┬──────────────┘
│
┌──────▼──────────────┐
│ Vector Store (FAISS)│
│   + Embeddings      │
└──────┬──────────────┘
│
┌──────▼──────────────┐
│  PDF Documents      │
│  (Technical Docs)   │
└─────────────────────┘

📁 Project Structure

MedTech AI/
├── app.py                      # Main Streamlit application
├── main.py                     # Alternative application entry point
├── config.py                   # Centralized configuration
├── requirements.txt            # Python dependencies
├── Dockerfile                  # Docker container setup
├── docker-compose.yml          # Docker Compose configuration
├── uv.lock                     # Dependency lock file
├── .python-version             # Python version specification
├── view.png                    # Project view image
│
├── core/                       # Core functionality modules
│   ├── __init__.py
│   ├── vectorstore.py         # FAISS vector store management
│   ├── qa_chain.py            # QA chain operations
│   └── document_processor.py   # PDF processing & chunking
│
├── utils/                      # Utility modules
│   ├── __init__.py
│   └── logger.py              # Logging configuration
│
├── scripts/                    # Utility scripts
│   ├── __init__.py
│   └── build_vectorstore.py   # Index documents script
│
├── tests/                      # Unit tests
│   ├── __init__.py
│   └── test_qa.py             # Test suite
│
├── data/                       # PDF documents (add your files here)
├── vectorstore/                # FAISS index storage
│   └── db_faiss/
├── logs/                       # Application logs
│
├── .env.example               # Environment variables template
├── .gitignore                 # Git ignore rules
├── .gitattributes             # Git attributes
├── LICENSE                    # MIT License
└── README.md                  # This file

🚀 Quick Start

Prerequisites

• Python 3.13+
• Groq API Key (Get one free)
• PDF documents for your knowledge base

Installation

1. Clone the repository

git clone https://github.com/beastNico/MedTech-AI.git
cd MedTech-AI

2. Create virtual environment

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Set up environment variables

copy .env.example .env
# Edit .env and add your GROQ_API_KEY

5. Add your PDF documents

# Place PDF files in the data/ directory
copy your_documents.pdf data\

6. Build vector store

python scripts\build_vectorstore.py

7. Run the application

streamlit run app.py

Visit http://localhost:8501 in your browser!

🐳 Docker Deployment

Using Docker Compose (Recommended)

# Build and run
docker-compose up -d

# View logs
docker-compose logs -f

# Stop
docker-compose down

Using Docker directly

# Build image
docker build -t medtech-ai .

# Run container
docker run -p 8501:8501 ^
  -v %cd%/data:/app/data ^
  -v %cd%/vectorstore:/app/vectorstore ^
  -e GROQ_API_KEY=your_key_here ^
  medtech-ai

🧪 Testing

Run tests with pytest:

# Run all tests
pytest

# Run with coverage
pytest --cov=core --cov=utils

# Run specific test file
pytest tests\test_qa.py -v

📝 Configuration

Edit config.py to customize:

# Model settings
LLM_MODEL = "deepseek-r1-distill-llama-70b"  # Change model
LLM_TEMPERATURE = 0.0                         # Adjust creativity

# Retrieval settings
RETRIEVAL_K = 6                               # Number of docs to retrieve
CHUNK_SIZE = 500                              # Text chunk size
CHUNK_OVERLAP = 50                            # Chunk overlap

🎯 Usage Examples

Casual Chat

User: Hi? Who Are You?
Bot: Hello. I'm a biomedical engineering assistant. It's nice to meet you. How can I assist you today?

Troubleshooting

User: Compare MRI and CT imaging techniques - what are their advantages and limitations?
Bot: Based on the provided context, here is a comparison of MRI and CT imaging techniques:

Advantages of MRI:

Non-invasive procedure
Does not require injecting a contrast medium
Greater sensitivity for detecting disk problems and spinal cord involvement...

API Key Errors

• Verify your .env file exists
• Check that GROQ_API_KEY is set correctly
• Ensure no quotes around the key value

📊 Performance Metrics

• Response Time: ~2-4 seconds per query
• Accuracy: Depends on document quality
• Uptime: 99%+ with proper deployment
• Concurrent Users: Supports multiple users (Streamlit limitation)

🛠️ Technology Stack

Component	Technology
Frontend	Streamlit
LLM	Groq (DeepSeek R1 Distill)
Embeddings	HuggingFace (MiniLM)
Vector Store	FAISS
Framework	LangChain
Document Processing	PyPDF + Unstructured
Logging	Python logging
Testing	Pytest
Containerization	Docker