HealthGuard AI

---

About

HealthGuard AI is a comprehensive health and fitness Android application powered by multiple AI models. It provides intelligent health chat, fitness tracking, medicine price comparison, drug interaction checking, and smart health monitoring - all in one app.

---

Download

Android APK (v1.0.0)

Download the latest release: app-release.apk

---

Features

• AI Health Chat - Multiple AI modes (Fast, Standard, Deep Think, Max Deep Think, Agent)
• Fitness Hub - 2000+ exercises with GIF animations and AI workout generation
• Medicine Price Comparison - Compare prices across Indian e-commerce platforms
• Health Dashboard - Track vitals like BP, blood sugar, weight, temperature, heart rate
• Drug Interaction Checker - Check drug interactions safely
• WhatsApp Notifications - Get medicine reminders via WhatsApp
• Voice Input - Talk to AI using microphone
• Nearby Pharmacy Map - Find pharmacies near you using Google Maps
• PDF Reports - Generate and download health reports
• Chat History - Save and manage AI conversations
• Firebase Authentication - Secure Google Sign-In

---

Current Android Build Updates

This build includes the latest Android-first fixes and feature additions:

• Native Google Sign-In: Android Google login now uses Capacitor Firebase Authentication with the native Google account picker, Firebase SHA fingerprints, and google-services.json support.
• Voice Input Fix: The chat mic records supported mobile audio formats, sends audio to /api/transcribe, and falls back to Groq Whisper in-app if the backend transcription route fails.
• Improved Health Chat Output: AI answers now render richer mobile-friendly Markdown with care-flow cards, simple explanations, do-first actions, and seek-help guidance.
• Smart Health Features: Added Smart Medicine Cabinet, Health Memory Timeline, AI Visit Prep PDF, Report Scanner, Family/Caregiver mode, and offline local reminders.
• Fitness Hub Reliability: Exercise search in generated plan/chat flows can render relevant local ExerciseDB results from the bundled exercise dataset.
• Reminder & Email Backend: Reminder email support is wired for Resend on the backend. Use environment variables for all email/API credentials.
• Location Backend Ready: Google Places support exists behind backend configuration, but billing-dependent Google Maps/Places features should remain disabled until billing is available.

Sensitive files such as .env, .env.local, and backend/.env are intentionally ignored and must not be committed.

---

Tech Stack

Frontend

Mobile (Android)

AI & Machine Learning

AI Architecture (How It Works)

┌─────────────────────────────────────────────────────────┐
│                    HealthGuard AI                        │
├─────────────────────────────────────────────────────────┤
│                                                          │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐ │
│  │   Fast Mode │    │Standard Mode│    │Deep Think   │ │
│  │   (Groq)    │    │  (Gemini)   │    │  (Gemini)   │ │
│  └──────┬──────┘    └──────┬──────┘    └──────┬──────┘ │
│         │                  │                   │        │
│         └──────────────────┼───────────────────┘        │
│                            │                            │
│                   ┌────────▼────────┐                   │
│                   │  LangGraph      │                   │
│                   │  (State Machine)│                   │
│                   └────────┬────────┘                   │
│                            │                            │
│         ┌──────────────────┼──────────────────┐         │
│         │                  │                  │         │
│  ┌──────▼──────┐    ┌─────▼─────┐    ┌──────▼──────┐  │
│  │  Clinical   │    │  Vitals   │    │  Medicine   │  │
│  │  Graph      │    │  RAG      │    │  Search     │  │
│  │  Agent      │    │  System   │    │  Agent      │  │
│  └─────────────┘    └───────────┘    └─────────────┘  │
│                                                          │
└─────────────────────────────────────────────────────────┘

LangGraph Clinical Agent

• Uses LangGraph StateGraph for clinical conversation flow
• Implements SOCRATES framework for medical history taking
• Stateful conversations with MemorySaver checkpointer
• Multi-node workflow: Intent Extraction → Information Gathering → Clinical Abstraction → Medical Search → Diagnosis

Vitals RAG System

• Retrieval-Augmented Generation for health context
• Stores user vitals in localStorage
• Analyzes trends and abnormal values
• Injects health context into every AI model's system prompt
• Ensures all AI modes are aware of user's health condition

Agentic AI Tools

• Function Calling with Google Gemini
• searchMedicine - Search medicine prices across platforms
• setHealthAlert - Set health reminders and alerts
• Workout Generation - AI-powered personalized workout plans

Backend

Backend Services

APIs & External Services

Maps & Location

PDF Generation

Markdown & Rich Text

Other Libraries

---

Deployment

Backend

• Platform: Red Hat OpenShift
• Runtime: Python Flask with Gunicorn
• Database: MongoDB Atlas (Cloud)
• Container: Docker

Frontend/Android

• Build Tool: Vite
• Mobile Wrapper: Capacitor
• Platform: Android (APK)

---

Architecture

HealthGuard-AI/
├── android/                    # Android native code (Capacitor)
│   ├── app/
│   │   ├── src/main/res/      # App icons, resources
│   │   └── build.gradle       # Android build config
├── backend/                    # Python Flask backend
│   ├── server.py              # Main API server
│   ├── agent_browser.py       # Medicine search agent
│   ├── playwright_order_agent.py  # Auto-order agent
│   └── data/exercises.json    # 2000+ exercises database
├── components/                 # React UI components
│   ├── TextChatInterface.tsx  # AI chat interface
│   ├── FitnessPanel.tsx       # Fitness hub
│   ├── HealthDashboard.tsx    # Health tracking
│   ├── DrugInteractionChecker.tsx  # Drug checker
│   ├── MedicinePriceCard.tsx  # Medicine prices
│   ├── NearbyPharmacyMap.tsx  # Pharmacy map
│   └── ...
├── src/
│   ├── agents/                # LangGraph agents
│   │   ├── clinicalGraph.ts   # Clinical conversation graph
│   │   ├── medicalKnowledgePrompts.ts  # Medical prompts
│   │   └── patientSession.ts  # Session management
│   ├── services/              # Frontend services
│   │   ├── firebaseAuth.ts    # Authentication
│   │   └── serverHealth.ts    # Server monitoring
│   ├── pages/                 # App pages
│   │   ├── Dashboard.tsx      # Main dashboard
│   │   ├── AuthPage.tsx       # Login/Signup
│   │   └── LandingPage.tsx    # Landing page
│   └── lib/                   # Utilities
├── services/                   # Shared services
│   ├── geminiService.ts       # Gemini AI integration
│   ├── groqService.ts         # Groq AI integration
│   └── vitalsRAG.ts           # RAG system for vitals
└── package.json               # Frontend dependencies

---

How RAG Works in HealthGuard AI

┌──────────────────────────────────────────────────────┐
│                  RAG Pipeline                         │
├──────────────────────────────────────────────────────┤
│                                                       │
│  1. User enters vitals (BP, sugar, weight, etc.)     │
│                    │                                  │
│                    ▼                                  │
│  2. Vitals stored in localStorage                    │
│                    │                                  │
│                    ▼                                  │
│  3. VitalsRAG analyzes:                              │
│     - Trend analysis (improving/worsening)           │
│     - Abnormal value detection                       │
│     - Severity classification                        │
│                    │                                  │
│                    ▼                                  │
│  4. Context string generated:                        │
│     "User has elevated BP (150/95), trending worse"  │
│                    │                                  │
│                    ▼                                  │
│  5. Context injected into ALL AI model prompts       │
│                    │                                  │
│                    ▼                                  │
│  6. AI responds with health-aware context            │
│                                                       │
└──────────────────────────────────────────────────────┘

---

App Screenshots

---

License

MIT License

---

Developer

Utkarsh Rana

• GitHub: @Rana3112

---