BabyNest is an Offline React Native Privacy-Focused mobile application for pregnancy tracking, health-record management, and AI-assisted support.

Medical disclaimer: BabyNest is an organizational and informational tool, not a substitute for professional medical advice, diagnosis, treatment, or emergency care.

• Offline AI-Powered Chat: Local GGUF LLM with user context integration & RAG support

• Task Management: View tasks & appointments by pregnancy week with priority sorting

• Health Tracking: Weight, symptoms, medications, blood pressure

• Manage pregnancy related tasks and appointments.

• Visualize pregnancy progress with a week-by-week timeline in graphical form.

• Store structured app data in a local SQLite database.

For Android development, install:

• Node.js and npm
• JDK 17
• Android Studio and Android SDK Platform 35
• Android SDK Build-Tools 35.0.0
• Android NDK 27.1.12297006
• An Android device with USB debugging enabled, or an Android emulator

The current Android configuration specifies minSdkVersion 24, compileSdkVersion 35, targetSdkVersion 34, and Kotlin 2.0.21. Check the Gradle files in your checkout if these values change.

See the official React Native 0.77 environment setup guide for environment configuration.

1. Clone the repository

   git clone https://github.com/AOSSIE-Org/BabyNest.git
   cd BabyNest

2. Install dependencies

   npm install
   # or
   pnpm install

3. Install React Native dependencies

   npx react-native link

Check the Setup.md file for additional setup instructions for Android and iOS.

The app uses a local GGUF model for chat responses. The model will be automatically downloaded on first use.

Model Details:

• Model: Qwen2-0.5B-Instruct
• File: qwen2.5-0.5b-instruct-q4_k_m.gguf

The model is configured in src/model/model.jsx:

• Context Window: 32,768 tokens
• Generation limit: 8,192 tokens
• Parameters: 0.49 billion
• Model file size: ~469 MiB

This is the current directory structure for the Frontend app:

BabyNest/
├── .bundle/
│   └── config
├── .eslintrc.js
├── .gitignore
├── .prettierrc.js
├── .watchmanconfig
├── __tests__/
├── android/
├── ios/
├── src/
│   ├── Components/
│   ├── Screens/
│   ├── assets/
│   ├── context/
│   ├── data/
│   ├── database/
│   ├── hooks/
│   ├── model/
│   ├── navigation/
│   ├── services/
│   ├── storage/
│   ├── theme/
│   └── utils/
├── App.jsx
├── Gemfile
├── Gemfile.lock
├── index.js
├── jest.config.js
├── metro.config.js
├── package-lock.json
├── package.json
├── pnpm-lock.yaml
├── react-native.config.js
├── README.md
├── RAG_COMMANDS.md
├── Setup.md
├── app.json
└──  tsconfig.json

// Context-aware conversation building
  const {context, refreshContext, initializeContext, isInitialized} = useAgentContext();

  const filteredTasks = tasks
    .filter(task => {
      // Filter by week
      const weekMatch =
        task.starting_week <= currentWeek && task.ending_week >= currentWeek;

      // Filter by status
      if (filter === 'pending')
        return weekMatch && task.task_status === 'pending';
      if (filter === 'completed')
        return weekMatch && task.task_status === 'completed';
      return weekMatch;
    })
    .sort((a, b) => {
      // Sort by priority first, then by starting week
      const priorityOrder = {high: 3, medium: 2, low: 1};
      const priorityDiff =
        priorityOrder[b.task_priority] - priorityOrder[a.task_priority];
      if (priorityDiff !== 0) return priorityDiff;
      return a.starting_week - b.starting_week;
    });

const textToSend = typeof content === 'string' ? content : userInput;

    if (!textToSend?.trim()) return;

    let finalMessage = textToSend.trim();

    if (isQuickReply && pendingQueryRef.current) {
      finalMessage = `${pendingQueryRef.current}. ${finalMessage}`;
    }

    // Store this query as context for a possible follow-up.
    pendingQueryRef.current = finalMessage;

    setUserInput('');

    // Message sending and core logic handled by hook
    const result = await sendMessage(textToSend, useRAGMode, initializeContext);

1. Start Metro bundler

   npx react-native start

2. Run on Android

   npx react-native run-android

3. Run on iOS

   npx react-native run-ios

1. Build Android APK

   cd android
   ./gradlew assembleRelease

2. Build iOS

   cd ios
   xcodebuild -workspace BabyNest.xcworkspace -scheme BabyNest -configuration Release

flowchart TD
    U([User]) --> APP[BabyNest App]
    APP --> CTX[Build / update shared app context]
    CTX --> Q[Receive user query or UI action]

    Q --> MODE{How should this request be handled?}

    MODE -->|UI action / screen available| UI[App UI screen]
    UI --> FH[In-app function handler]
    FH --> ROUTE{Route to matching handler}
    ROUTE --> PH[profileHandler.js]
    ROUTE --> AH[appointmentHandler.js]
    ROUTE --> AGH[agentHandler.js]
    ROUTE --> BPH[BPHandler.js]

    PH --> DB[(Local database inside APK)]
    AH --> DB
    AGH --> DB
    BPH --> DB
    DB --> CTXUP[Update / share app context]
    CTXUP --> APP

    MODE -->|RAG mode enabled| RAG[RAG service]
    RAG --> MATCH{Query matches a defined intent?}
    MATCH -->|Yes| EXTRACT[extractData: extract intent fields]
    EXTRACT --> CHECK[checkMissingFields: validate required fields]
    CHECK --> MISSING{Required fields missing?}
    MISSING -->|Yes| FOLLOW[generateFollowUpQuestions]
    FOLLOW --> ASK[Ask user for missing information]
    ASK --> Q
    MISSING -->|No| EXEC[executeAction: run matched action]
    EXEC --> HANDLER[Call mapped in-app handler]
    HANDLER --> DB

    MATCH -->|No| FALLBACK[Fall back to AI mode]
    MODE -->|AI mode enabled| LOCAL[Local language model]
    FALLBACK --> LOCAL
    LOCAL --> TEXT[Generate text response]
    TEXT --> RESP[Return response to user]
    RESP --> APP

    style APP fill:#fff3d6,stroke:#d97706,color:#111827
    style RAG fill:#fff3d6,stroke:#d97706,color:#111827
    style LOCAL fill:#e0f2fe,stroke:#0284c7,color:#111827
    style DB fill:#f3f4f6,stroke:#6b7280,color:#111827
    style CTX fill:#dcfce7,stroke:#16a34a,color:#14532d
    style CTXUP fill:#dcfce7,stroke:#16a34a,color:#14532d
    style FALLBACK fill:#fee2e2,stroke:#dc2626,color:#7f1d1d

Build/share app context
        |
        v
Receive query or UI action
        |
        +-- User selects an available UI screen/action
        |       -> In-app function handler
        |       -> Matching feature handler
        |       -> Local database, if required
        |       -> Update/share context
        |
        +-- ragMode == true
        |       -> RAG service searches for a defined intent
        |       -> Intent found?
        |            -> Yes: extract data
        |            -> Check required fields
        |            -> Missing fields: ask follow-up question and wait
        |            -> All fields present: execute mapped action
        |            -> No intent match: fall back to local AI mode
        |
        +-- aiMode == true
                -> Local language model
                -> Generate text response
                -> Return response to the user

npm test

1. Model Not Loading

   • Check internet connection for download
   • Verify sufficient storage space
   • Check .env file configuration

2. Context Not Loading

   • Check context agent initialization
   • Verify database setup
   • Check user profile exists

# Check model status
adb logcat | grep "Model"

# Check API calls
adb logcat | grep "fetch"

# Check context loading
adb logcat | grep "Context"

• Local Model: GGUF model runs locally, no data sent to external APIs
• Health Data: Stored locally in SQLite database
• Context Privacy: User health data only used for local AI responses
• Offline Capability: Chat works without internet connection

• Model Caching: GGUF model loaded once per session
• Context Caching: User context cached with hash-based invalidation
• Lazy Loading: Model loads only when ChatScreen is accessed
• Memory Management: Automatic model cleanup on component unmount

• Multi-user support
• Real-time health monitoring
• Advanced analytics dashboard
• Integration with health devices
• Better RAG services
• Voice chat interface

Please read CONTRIBUTING.md before submitting a change. Include clear reproduction steps with bug reports, and never include passwords, tokens, or private health information in issues or logs.

See the repository's LICENSE.md for the applicable license.

© 2026 AOSSIE-Org. All rights reserved. See COPYRIGHT.md for details on usage and warranty disclaimers.