/RAG-FirstAid-Chatbot

This chatbot uses a Retrieval-Augmented Generation (RAG) approach to provide first-aid support for Diabetes, Cardiac, and Renal emergencies.

Primary LanguagePython

🤖 RAG-Powered First-Aid Chatbot

A smart chatbot that provides short, medically-aware first-aid guidance for emergencies related to:

  • Diabetes
  • Cardiac issues
  • Renal conditions

It uses:

  • 🧠 Gemini API for condition detection and final answer generation
  • 📚 Local medical knowledge (from a curated Excel file)
  • 🌐 Live web search using Serper.dev (Google Search API)
  • ✅ Answers limited to ≤ 250 words with disclaimers and citations

⚠️ Disclaimer

This chatbot is for educational purposes only and is not a substitute for professional medical advice.

📁 Folder Structure

RAG-FirstAid-Chatbot/
│
├── src/
│   ├── chatbot.py                  # Main pipeline
│   ├── model.py                    # Checkout the model available through API
│   ├── triage.py                   # Gemini: symptom → condition
│   ├── retrieval.py                # Local snippet retriever
│   ├── web_search.py               # Serper.dev integration
│   └── generator.py                # Gemini: RAG response builder
│
├── tests/
│   ├── test_bot.py                 # Unit test for pipeline
│   |── test_queries.py             # Evaluates all 10 official test cases
│   └── test_queries_response.md    # Response of all 10 official test cases
│
├── Assignment Data Base.xlsx       # Local knowledge base
├── requirements.txt                # List of required libraries
├── .env                            # API keys
├── README.md
├── architecture.md                 # System flow
└── performance.md                  # Evaluation results

🛠️ Setup Instructions

git clone <your-repo-link>
cd RAG-FirstAid-Chatbot

python -m venv venv
venv\Scripts\activate           # (use source venv/bin/activate for Linux/Mac)

pip install -r requirements.txt

🔐 Environment Variables (.env)

Create a .env file and add:

GEMINI_API_KEY=your_google_api_key
SERPER_API_KEY=your_serper_api_key

▶️ Running the Chatbot

# From project root
python -m src.chatbot

Then modify the query variable inside chatbot.py for different test inputs.

🧪 Run Tests

# Unit test
set PYTHONPATH=.
pytest tests/

# 10 official test cases
python tests/test_queries.py

🧠 Design Trade-offs

Area Design Decision Trade-off
Triage Gemini API More accurate than keyword matching, but slower
Knowledge Base Excel file + embeddings (local) Simple setup; no DB required
Retrieval Top-3 local + top-3 web Balances trust (local) + freshness (web)
Answer Gen Gemini RAG prompt with strict template Keeps output short & informative
Testing Manual + Pytest + Query evaluation script No UI, but CLI is simple to demo

✅ Features Recap

  • ✅ Gemini-powered symptom triage
  • ✅ Combined local and web retrieval
  • ✅ Medically aware answers
  • ✅ Unit tests + test queries
  • ✅ Ready for demo and GitHub push

💡 Future Improvements

  • Add interactive chat UI (e.g., Streamlit or Gradio)
  • Deploy as web app with REST API
  • Add medical named entity recognition (NER)