MedChat AI is a privacy-focused medical chatbot that uses a locally hosted Mistral 7B LLM with Retrieval-Augmented Generation (RAG) from the Gale Encyclopedia of Medicine to deliver accurate, context-aware health information. It features a responsive React UI, contextual memory, and location-based doctor recommendations—all with a strong commitment to ethical AI use.
- 🔒 Local LLM: Runs Mistral 7B via Ollama, ensuring full privacy and offline support.
- 📚 RAG Knowledge Base: Uses Gale Encyclopedia of Medicine for accurate, medical-grade responses.
- 🧠 Contextual Memory: Maintains conversation history for natural multi-turn dialogue.
- 📍 Doctor Finder: Recommends nearby specialists using a location-aware MongoDB dataset.
- 💬 React + FastAPI Stack: Modern, mobile-friendly frontend with a robust Python backend.
⚠️ Medical Disclaimer: Each response includes a built-in notice that it's not a substitute for professional care.
- Frontend: React + TypeScript + Tailwind CSS
- Backend: FastAPI + Python
- LLM: Mistral 7B (via Ollama)
- RAG: Custom ingestion using Gale Encyclopedia
.txtfiles + vector store - Database: MongoDB (for doctor & hospital info)
- Create a Environment
python -m venv medbot
source medbot/bin/activate- Install required packages
pip install -r requirements.txt-
Start Local database with MongoDBCompass and select Database.json file from data directory.
-
Install ollama and download the LLM you want to use.
-
Start backend
uvicorn main:app --reload- Start Frontend
cd frontend
npm install
npm run dev- Open localhost with your browser to see the result.
http://localhost:5173