🩺 VitaAI – AI Health Chatbot

Symptom-based disease prediction powered by DistilBERT embeddings, Logistic Regression, and a conversational RAG pipeline.

---

✨ Features

• 🔬 Top-3 Disease Prediction — Enter symptoms in plain English and get the top 3 most likely conditions with confidence scores
• 🧠 DistilBERT Embeddings — Semantic understanding of symptoms using a pre-trained transformer model
• 📚 RAG Pipeline — Retrieval-Augmented Generation for medical knowledge retrieval
• 🗂️ Intent Classification — Understands user intent before answering
• 💬 React Web App — Full-stack React + Express interface for a smooth chat experience
• 🗃️ SQLite Storage — Lightweight local chat history via better-sqlite3

---

🖼️ Demo

Symptom Input	Predicted Diseases
fever, cough, sore throat, body pain, fatigue	Flu, COVID-19, Asthma
frequent urination, increased thirst, fatigue	Diabetes, Anemia, Hypertension
severe headache, nausea, vomiting	Migraine, Hypertension, Food Poisoning

⚠️ Disclaimer: VitaAI is for educational purposes only. It is not a substitute for professional medical advice, diagnosis, or treatment.

---

🛠️ Tech Stack

Tool	Purpose
React 19 + Vite	Frontend SPA
Express	API server
LangChain + Google Gemini	RAG & intent classification
transformers (DistilBERT)	Sentence embeddings
scikit-learn	Logistic Regression classifier
better-sqlite3	Chat history storage
Tailwind CSS	Styling

---

📁 Project Structure

vitaai/
├── server.ts           # Express + Vite server entry point
├── server/
│   └── routes.ts       # API routes (RAG, chat, intent classification)
├── src/                # React frontend
├── index.html          # HTML entry point
├── FAQ.csv             # Symptom-disease training dataset
├── package.json        # Node dependencies
├── vite.config.ts      # Vite configuration
└── tsconfig.json       # TypeScript config

---

🧩 How It Works

User enters symptoms
        │
        ▼
DistilBERT encodes symptoms → sentence embedding
        │
        ▼
Logistic Regression classifies embedding
        │
        ▼
Top-3 diseases returned with confidence scores
        │
        ▼
LangChain RAG retrieves medical context
        │
        ▼
Gemini generates a natural-language response

---

📊 Dataset

The model is trained on FAQ.csv, a curated dataset of symptom–disease pairs covering conditions including:

• Flu, COVID-19, Asthma
• Diabetes, Hypertension, Heart Disease
• Migraine, Anemia, Kidney Stone
• Food Poisoning, and more

---

🤝 Contributing

Contributions, issues, and feature requests are welcome!

1. Fork the repo
2. Create a branch: git checkout -b feature/your-feature
3. Commit: git commit -m "Add your feature"
4. Push: git push origin feature/your-feature
5. Open a Pull Request

---

👤 Author

Harshit Ranbhare

---

📄 License

This project is open source and available under the MIT License.