惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
V
Visual Studio Blog
IT之家
IT之家
博客园 - 聂微东
The Cloudflare Blog
月光博客
月光博客
阮一峰的网络日志
阮一峰的网络日志
S
SegmentFault 最新的问题
Apple Machine Learning Research
Apple Machine Learning Research
酷 壳 – CoolShell
酷 壳 – CoolShell
爱范儿
爱范儿
H
Help Net Security
博客园 - 叶小钗
V
V2EX
WordPress大学
WordPress大学
J
Java Code Geeks
Hugging Face - Blog
Hugging Face - Blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
博客园_首页
C
Check Point Blog
B
Blog
D
DataBreaches.Net
美团技术团队
罗磊的独立博客

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
AI CareCompanion - Offline Health Assistant
NARENDER KES · 2026-05-24 · via DEV Community

NARENDER KESWANI

Gemma 4 Challenge: Build With Gemma 4 Submission

Offline-first AI health companion powered by Google's Gemma 4 4B running entirely on-device. Multi-agent specialist routing, drug lookup, symptom triage, nearby doctor search, and Firebase-backed cloud sync — all in a single React Native app for Android and iOS.


YouTube Video

AI CareCompanion Demo


Screenshots

Splash Screen

Splash Screen


Sign In

Sign In


Model Download

Model Download


Home Dashboard

Home


Lab Report Analysis

Lab Report Analysis


AI Response

AI Response


Skin Lesion Analysis

Skin Lesion Analysis


Hindi Language Support

Hindi Language


Chat Reports

Chat Reports


Generated Report

Generated Report


PDF Patient Summary

PDF Patient Summary


PDF Clinical Report

PDF Clinical Report


Health Insights

Health Insights


Insight Detail

Insight Detail


Navigation Menu

Navigation Menu


Drug Search

Drug Search


Drug Detail

Drug Detail


Affordable Alternatives

Affordable Alternatives


Nearby Doctors

Find Nearby


On-Device AI

On-Device AI


System Architecture

High-Level Design (HLD)

High Level Design

Data flow between the Mobile Client, AI Orchestrator, Inference Engines, and SQLite/Firebase Data Layer.


Low-Level Design (LLD)

Low Level Design

The 5-step sequence flow of a user query being routed and processed by the AI Orchestrator and Specialists.

Features

On-Device AI — No Internet Required

  • Gemma 4 4B (IQ2_M) runs fully on-device via llama.rn (GGUF format)
  • ~1.5 GB model, ~2 GB RAM — works offline after one-time download
  • Zero data sent to external servers during inference
  • Cloud AI (Gemini) used as fallback when online

Multi-Agent Specialist System

  • 46 specialist agents — each with domain-specific system prompts and tool access
  • Central orchestrator auto-detects intent and routes to the right specialist(s)
  • Up to 3 agents collaborate on complex multi-system queries
  • Real-time routing info (agent name + response time) shown in every message
# Agent Specialty Image Support
1 Triage Assistant Orchestrator — intent detection & routing
2 General Health Specialist General Practice
3 Child Health Specialist Paediatrics
4 Chronic Care Specialist Internal Medicine
5 Heart Health Specialist Cardiology
6 Skin Care Specialist Dermatology
7 Brain & Nerve Specialist Neurology
8 Digestive Health Specialist Gastroenterology
9 Lung & Breathing Specialist Pulmonology
10 Kidney Specialist Nephrology
11 Hormone Specialist Endocrinology
12 Infection Specialist Infectious Disease
13 Women's Health Specialist Obstetrics & Gynecology
14 Bone & Muscle Specialist Orthopaedics
15 Eye Care Specialist Ophthalmology
16 Ear, Nose & Throat Specialist ENT
17 Mental Health Specialist Psychiatry
18 Dental Care Specialist Dentistry
19 Urinary Tract Specialist Urology
20 Blood Specialist Haematology
21 Joint & Immune Specialist Rheumatology
22 Allergy Specialist Allergy & Immunology
23 Newborn Care Specialist Neonatology
24 Senior Health Specialist Geriatrics
25 Emergency Care Specialist Emergency Medicine
26 Nutrition & Diet Specialist Nutrition & Dietetics
27 Medication Specialist Pharmacy
28 Radiology & Imaging Specialist Radiology
29 Sports Medicine Specialist Sports Medicine
30 Sleep Medicine Specialist Sleep Medicine
31 Palliative & Hospice Care Specialist Palliative Care
32 Rehabilitation Specialist Physiatry / Rehab Medicine
33 Genetics & Genomics Specialist Genetics
34 Pain Management Specialist Pain Management
35 Toxicology & Poisoning Specialist Toxicology
36 Travel Medicine Specialist Travel Medicine
37 Adolescent Medicine Specialist Adolescent Medicine
38 Reproductive & Fertility Specialist Fertility
39 Vascular Medicine Specialist Vascular Medicine
40 Transplant Medicine Specialist Transplant Medicine
41 Clinical Pathology Specialist Pathology
42 Integrative Medicine Specialist Integrative Medicine
43 Addiction Medicine Specialist Addiction Medicine
44 Occupational Medicine Specialist Occupational Medicine
45 Lifestyle Medicine Specialist Lifestyle Medicine
46 Preventive Medicine Specialist Preventive Medicine

Health Tools

  • Drug Info — Search 1mg medicine database (name, dosage, warnings, interactions)
  • Affordable Alternatives — Find cheaper generic substitutes with price comparison
  • First Aid Guide — Step-by-step emergency guidance
  • Find Nearby Doctors — Location-based doctor and hospital search
  • Image Analysis — Analyze wound, rash, or medical images via camera/gallery
  • Health Metrics — Track vitals (BP, glucose, weight) over time
  • Medication Reminders — Local notifications for medication schedules
  • Chat Reports — Export conversation summaries as PDF

Privacy & Sync

  • All AI inference runs on-device — no health data leaves your phone
  • SQLite local database for all chats, profiles, and history
  • Firebase Firestore optional cloud sync with incremental pull + outbox pattern
  • Biometric authentication (fingerprint/face) for app lock

UI/UX

  • Light and Dark mode
  • Markdown rendering in AI responses
  • Animated splash screen
  • Search history across drug lookups and AI queries

Tech Stack

Layer Technology
Framework React Native 0.76 (TypeScript)
On-Device AI llama.rn — Gemma 4 4B IQ2_M GGUF
Cloud AI Google Gemini (online fallback)
Auth Firebase Auth (Email/Password + email verification)
Local DB SQLite via react-native-sqlite-storage
Cloud Sync Firebase Firestore
Navigation React Navigation v7
Biometrics react-native-biometrics

Setup

Prerequisites

  • Node.js >= 18
  • React Native CLI
  • Android Studio (API 33+) or Xcode 15+
  • Firebase project with Auth and Firestore enabled

1. Clone and install

git clone https://github.com/narender-rk10/AI-CareCompanion-Offline-Health-By-Gemma.git
cd AI-CareCompanion-Offline-Health-By-Gemma
npm install

Enter fullscreen mode Exit fullscreen mode

2. iOS pods

cd ios && pod install && cd ..

Enter fullscreen mode Exit fullscreen mode

3. Firebase configuration

  1. Create a project at console.firebase.google.com
  2. Enable Authentication → Email/Password
  3. Enable Firestore Database
  4. Download google-services.json → place in android/app/
  5. Download GoogleService-Info.plist → add to Xcode project

4. Environment variables

cp .env.example .env

Enter fullscreen mode Exit fullscreen mode

Then fill in .env:

# Gemini API key — https://aistudio.google.com/app/apikey
GEMINI_API_KEY=your_gemini_api_key_here

# Google Sign-In Web Client ID (Firebase Console → Auth → Google → Web SDK config)
GOOGLE_WEB_CLIENT_ID=your_google_web_client_id.apps.googleusercontent.com

# Direct GGUF download URL for Gemma 4 4B (~1.5 GB)
MODEL_DOWNLOAD_URL=https://huggingface.co/your-org/model/resolve/main/model.gguf

Enter fullscreen mode Exit fullscreen mode

Firebase credentials go in android/app/google-services.json and ios/GoogleService-Info.plist — not in .env.

5. Run

# Android
npm run android

# iOS
npm run ios

Enter fullscreen mode Exit fullscreen mode

6. Download the AI model

On first launch, the app will prompt you to download Gemma 4 4B (~1.5 GB). Requires a stable Wi-Fi connection and ~3 GB free storage. After download the model runs fully offline.

Project Structure

src/
├── components/          # Reusable UI components
│   ├── ui/              # AppButton, AppInput, AppCard, ScreenHeader
│   ├── ChatBubble.tsx
│   ├── BiometricGate.tsx
│   ├── MarkdownText.tsx
│   └── SyncStatusBar.tsx
├── screens/             # All app screens
│   ├── ChatScreen.tsx               # Main AI chat
│   ├── HomeScreen.tsx               # Dashboard
│   ├── DrugInfoScreen.tsx           # Medicine search
│   ├── AffordableAlternativesScreen.tsx
│   ├── FirstAidGuideScreen.tsx
│   ├── FindNearbyDoctorScreen.tsx
│   ├── ImageAnalysisScreen.tsx
│   ├── AgentSelectorScreen.tsx
│   ├── ChatReportsScreen.tsx
│   ├── SettingsScreen.tsx
│   └── ...
├── services/            # Business logic
│   ├── LlmService.ts              # Gemini cloud AI
│   ├── LocalLlmService.ts         # llama.rn on-device AI
│   ├── AgentService.ts            # Multi-agent orchestration
│   ├── ModelManager.ts            # Model download & lifecycle
│   ├── DatabaseService.ts         # SQLite wrapper
│   ├── FirebaseService.ts         # Auth + Firestore sync
│   ├── DrugInfoService.ts         # 1mg medicine DB
│   └── PlaceSearchService.ts      # Nearby doctor search
├── context/
│   └── HealthProfileContext.tsx   # Global health profile state
├── navigation/
│   └── AppNavigator.tsx           # Auth flow + tab navigation
└── theme/                         # Colors, typography, dark mode

Enter fullscreen mode Exit fullscreen mode


Demo Prompts

"I have chest pain and shortness of breath"
"What is the dosage for metformin 500mg?"
"My child has had a fever for 3 days"
"Find a cardiologist near me"
"What are affordable alternatives to Crocin?"
"Analyze this rash on my arm" (attach photo)

Enter fullscreen mode Exit fullscreen mode


Privacy

  • AI inference is 100% on-device when using the offline model
  • Health profile and chat history stored only in local SQLite by default
  • Firebase sync is opt-in — no data uploaded without user action
  • Biometric lock prevents unauthorized access

License

Copyright (c) 2026 Narender Keswani. All Rights Reserved.

This project is proprietary software. You may view the source code for personal reference, but you may not copy, distribute, modify, or use it commercially without explicit written permission.

See LICENSE for full terms.


Built for people who need reliable health guidance anywhere — with or without internet.