Health Chatbot: AI-Powered Personalized Health Assistant

Final Year Project Phase - 1

An AI-powered healthcare support system utilizing a three-model architecture for accurate, safe, and research-backed health assistance.

Features

• Intelligent health assistance with research-backed responses
• Multi-platform support (Web & WhatsApp)
• Real-time health tips and recommendations
• Secure human handoff capabilities
• Vector database for efficient knowledge management

Technical Architecture

AI Models

• Gemini Flash (Query Analysis)
  • Temperature: 0.3
  • Response time: 1-2s
  • Fast query decomposition
• Perplexity Sonar (Research)
  • Context window: 128k tokens
  • Real-time data retrieval
• Gemini Pro (Response)
  • Temperature: 0.7
  • Enhanced prompting

Tech Stack

• Frontend: Streamlit
• Backend: Flask with RESTful API
• Database: ChromaDB
• Integration: Twilio for WhatsApp
• AI: Google Gemini and Perplexity Sonar

System Components

Frontend Features

• Interactive chat interface
• Health tips display
• Assessment interface
• Feedback system

Backend Features

• Vector database integration
• Knowledge base management
• Chat history tracking
• Multi-layer safety system

Safety Features

• Content filtering
• Intent detection
• Source credibility checks
• Medical disclaimer injection
• Warning generation