JARVIS — AI Voice Assistant

A serverless, AI-powered voice assistant mobile app built with React Native (Expo) and AWS. Record voice memos to set alarms, create reminders, send texts, and manage tasks — all hands-free.

Features

• 🎙️ Voice Recording — Record voice memos that are automatically transcribed and analyzed
• ⏰ Smart Alarms — "Set an alarm for 7am" → schedules a device notification
• 📌 Reminders — "Remind me to finish the report by 5pm" → creates a timed reminder
• ✅ Task Lists — "Add buy groceries to my list" → creates and tracks tasks
• 🔔 Push Notifications — Remote notifications for reminders via Expo Push
• 🔐 Authentication — Secure sign-up/sign-in with AWS Cognito
• ☁️ Fully Serverless — Zero servers to manage, pay only for what you use

Architecture

Prerequisites

• Node.js 20+ and npm
• AWS CLI configured with credentials (aws configure)
• AWS CDK (npm install -g aws-cdk)
• Expo CLI (npm install -g expo-cli)
• iOS Simulator (macOS) or Android Emulator or a physical device
• Amazon Bedrock — Claude 3.5 Sonnet model access enabled in us-east-1

Quick Start

1. Configure AWS

# Configure AWS CLI with your credentials
aws configure
# Enter: Access Key ID, Secret Access Key, Region (us-east-1), Output (json)

# Bootstrap CDK (one-time setup)
cd infra && npx cdk bootstrap

2. Deploy Backend

cd infra
npx cdk deploy

# Note the outputs:
# ApiUrl, UserPoolId, UserPoolClientId, IdentityPoolId, AudioBucketName

3. Configure Mobile App

cd app
cp .env.example .env
# Edit .env with the CDK output values

4. Run the App

cd app
npx expo start

# Then press:
# i — open iOS Simulator
# a — open Android Emulator
# Scan QR code — open on physical device with Expo Go

Project Structure

JARVIS/
├── app/                    # React Native (Expo) mobile app
│   ├── app/               # Expo Router screens
│   │   ├── (auth)/        # Login, signup, verification
│   │   └── (tabs)/        # Main app tabs (Record, Tasks, Settings)
│   ├── components/        # Reusable UI components
│   ├── hooks/             # Custom React hooks
│   ├── services/          # API client, auth, SMS
│   ├── constants/         # Theme, config
│   └── types/             # TypeScript interfaces
├── lambdas/               # AWS Lambda function source code
│   ├── memo-upload/       # Presigned URL generation
│   ├── memo-processor/    # Transcribe + Bedrock pipeline
│   ├── task-manager/      # Task CRUD
│   ├── notification-sender/ # Push notification delivery
│   ├── push-token-manager/  # Device token storage
│   └── shared/            # Shared types, DB helpers, Bedrock client
├── infra/                 # AWS CDK infrastructure
│   ├── lib/infra-stack.ts # All AWS resources
│   └── bin/infra.ts       # CDK app entry point
└── README.md

How It Works

1. Record — User records a voice memo in the app
2. Upload — Audio is uploaded directly to S3 via presigned URL
3. Transcribe — S3 event triggers Lambda → Amazon Transcribe converts speech to text
4. Analyze — Transcript is sent to Bedrock (Claude 3.5 Sonnet) for intent extraction
5. Execute — For each detected intent:
   • Alarm: Schedules a local device notification
   • Reminder: Schedules EventBridge → Lambda → Expo Push notification
   • Message: Opens native SMS composer with pre-filled recipient and body
   • Task: Stored in DynamoDB and displayed in the Tasks tab
6. Notify — At scheduled times, EventBridge triggers Lambda to send push notifications

Cost Estimate (Low Volume)

License

MIT

Built With

• react-native