https://github.com/whoshotu/-BUDDY Buddy - Voice-First Dementia Care Assistant
AI-powered voice assistant for dementia patients, built for the Amazon Nova AI Hackathon 2026.
๐ฏ MVP Scope
Alexa Custom Skill - Patient voice interface with Amazon Nova
Caregiver Dashboard - Web interface for managing patient knowledge base
3-Level Safety Escalation - Smart alerts via Amazon SNS SMS
Core Use Cases: Routine guidance, family recognition, medication reminders
๐๏ธ Architecture
๐ Quick Start
Prerequisiteshttps://github.com/whoshotu/-BUDDY Buddy - Voice-First Dementia Care Assistant
AI-powered voice assistant for dementia patients, built for the Amazon Nova AI Hackathon 2026.
๐ฏ MVP Scope
Alexa Custom Skill - Patient voice interface with Amazon Nova
Caregiver Dashboard - Web interface for managing patient knowledge base
3-Level Safety Escalation - Smart alerts via Amazon SNS SMS
Core Use Cases: Routine guidance, family recognition, medication reminders
๐๏ธ Architecture
Patient Device (Alexa)
โ
AWS Lambda (Alexa Skill Handler)
โ (bidirectional streaming)
Amazon Bedrock Nova 2 Sonic (speech-to-speech)
โ (tool calls)
Nova 2 Lite Agentic Workflow
โ
โโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโฌโโโโโโโโโโโโโโ
โ DynamoDB โ RAG Query โ Safety Gate โ
โ(Knowledge โ (Patient โ(Escalation โ
โ Base) โ Context) โ Logic) โ
โโโโโโโโโโโโโโโดโโโโโโโโโโโโโโดโโโโโโโโโโโโโโ
โ
Caregiver Alert (SMS/SNS) or Emergency Assist
๏ฟผ
TypeScript
๐ Quick Start
Prerequisites
AWS CLI configured (aws configure)
AWS Account with Nova access
Node.js 18+ (for Lambda deployment)
Python 3.9+ (for infrastructure scripts)
Phase 1: Deploy Infrastructure
Deploy DynamoDB tables and seed data
./deploy.sh --environment dev --region us-east-1
๏ฟผ
TypeScript
This creates:
BuddyCaregivers-dev - Login credentials and notification preferences
BuddyPatients-dev - Patient profiles, routines, medications, people
BuddyAssignments-dev - Caregiver-patient relationship mapping
BuddyConversationLogs-dev - Conversation history (90-day TTL)
Test Credentials:
Username: caregiver_test
Password: Demo2026!
Patient: John Doe
Phase 2: Deploy Alexa Skill
Configure Environment Copy environment template cp .env.example .env ๏ฟผ TypeScript Edit .env with your AWS values nano .env ๏ฟผ TypeScript
Deploy Lambda Function Deploy the Alexa skill Lambda
./deploy-lambda.sh --environment dev --region us-east-1
๏ฟผ
TypeScript
This creates:
Lambda function buddy-alexa-skill-dev
Alexa Skills Kit trigger configured
Environment variables for DynamoDB tables
Save the Lambda ARN - you'll need it for the Alexa Developer Console.
Configure Alexa Developer Console Go to Alexa Developer Console
Click "Create Skill"
Skill Name: "Buddy - Dementia Care Assistant"
Choose a model: Custom
Choose a method: Provision your own
Click "Create Skill"
Interaction Model:
Click on "JSON Editor" (left sidebar)
Delete the default content
Copy the contents of src/alexa-skill/models/en-US.json and paste it in
Click "Save Model" then "Build Model"
Endpoint Configuration:
Go to "Endpoint" (left sidebar)
Select "AWS Lambda ARN"
Paste your Lambda ARN from deployment
Click "Save Endpoints"
Account Linking (Optional):
Go to "Account Linking" (left sidebar)
Configure if you want user authentication
Test:
Go to "Test" tab
Enable testing for your skill
Test: "Alexa, open Buddy Assistant"
Test Your Skill
Test commands
"What do I do this morning?"
"Who is Sarah?"
"What do I take after breakfast?"
Safety tests
"I'm scared" (repeat 4 times for Level 1 alert)
"I fell down" (immediate Level 2 alert)
๏ฟผ
TypeScript
Phase 3: Refinement & Monitoring (Optional)
Deploy Nova Sonic WebSocket API for advanced speech-to-speech
./scripts/deploy-websocket.sh --environment dev --region us-east-1
๏ฟผ
TypeScript
Deploy CloudWatch monitoring dashboard
./scripts/deploy-monitoring.sh --environment dev --region us-east-1
๏ฟผ
TypeScript
Nova Sonic Features:
Bidirectional streaming with barge-in support
Natural turn-taking and adaptive speech
WebSocket API for real-time conversations
Hybrid mode (traditional skill + Nova Sonic)
Emergency Assist Features:
Level 2 emergency protocol with location sharing
30-minute emergency session persistence
Dual notification system (SNS + SMS)
Real-time caregiver updates
See Nova Sonic README and Emergency Documentation for details.
Phase 4: Caregiver Dashboard
Start the React dashboard
cd src/caregiver-dashboard
npm install
npm run dev
๏ฟผ
TypeScript
Open
http://localhost:3000
๏ฟผ
TypeScript
Dashboard Features:
Patient profile overview
Real-time conversation monitoring
Safety alert management
Emergency contact information
Phase 5: Testing Without Alexa Device
Test Lambda function directly
cd test
./test-lambda.sh
๏ฟผ
TypeScript
This simulates Alexa requests and tests emergency flow
See Testing Guide for comprehensive test scenarios.
๐ Project Structure
buddy/
โโโ docs/ # Documentation
โ โโโ README.md # Documentation index
โ โโโ EXECUTIVE_SUMMARY.md # Project overview
โ โโโ ARCHITECTURE.md # System design
โ โโโ API.md # API reference
โ โโโ DEMO_SCRIPT.md # Demo guide
โ โโโ PRD.md # Product requirements
โ โโโ SETUP.md # Installation guide
โ โโโ DEPLOYMENT.md # Deployment strategies
โ โโโ EMERGENCY.md # Emergency features
โ โโโ BETA_TESTING.md # Testing procedures
โ โโโ DYNAMODB_OPTIMIZATION.md # Performance guide
โ
โโโ infrastructure/ # CloudFormation templates
โ โโโ dynamodb.yaml # DynamoDB tables
โ โโโ websocket-api.yaml # WebSocket API
โ โโโ monitoring.yaml # CloudWatch dashboards
โ โโโ seed_data.py # Test data
โ
โโโ src/
โ โโโ alexa-skill/ # Lambda function
โ โ โโโ index.js
โ โ โโโ skill.json
โ โ โโโ models/
โ โ โโโ package.json
โ โ โโโ tests/ # Unit tests
โ โ
โ โโโ caregiver-dashboard/ # React web interface
โ โ โโโ src/
โ โ โโโ package.json
โ โ โโโ README.md
โ โ
โ โโโ nova-sonic-websocket/ # Advanced speech-to-speech
โ โโโ index.js
โ โโโ test-client.html
โ โโโ README.md
โ
โโโ scripts/ # Deployment scripts
โ โโโ deploy.sh
โ โโโ deploy-iam.sh
โ โโโ deploy-lambda.sh
โ โโโ deploy-websocket.sh
โ โโโ deploy-monitoring.sh
โ โโโ verify.sh
โ
โโโ test/ # Test suite
โ โโโ test-lambda.sh
โ โโโ test-lambda.js
โ
โโโ demo/ # Demo materials
โโโ README.md # Main project README
๏ฟผ
TypeScript
๐ง Environment Configuration
Copy .env.example to .env and configure:
Required variables (auto-populated by deployment scripts)
AWS_REGION=us-east-1
ENVIRONMENT=dev
ROLE_ARN=arn:aws:iam::YOUR_ACCOUNT_ID:role/buddy-lambda-role
SNS_TOPIC_ARN=arn:aws:sns:us-east-1:YOUR_ACCOUNT_ID:buddy-alerts-dev
LAMBDA_ARN=arn:aws:lambda:us-east-1:YOUR_ACCOUNT_ID:function:buddy-alexa-skill-dev
DynamoDB tables (auto-populated)
CAREGIVERS_TABLE=BuddyCaregivers-dev
PATIENTS_TABLE=BuddyPatients-dev
ASSIGNMENTS_TABLE=BuddyAssignments-dev
LOGS_TABLE=BuddyConversationLogs-dev
๏ฟผ
TypeScript
๐งช Testing
Local Testing
Navigate to skill directory
cd src/alexa-skill
๏ฟผ
TypeScript
Install dependencies
npm install
๏ฟผ
TypeScript
Create test event
cat > test-event.json << 'EOF'
{
"version": "1.0",
"session": {
"new": true,
"sessionId": "test-session-001",
"application": { "applicationId": "test-app-id" },
"user": { "userId": "amzn1.ask.account.test" }
},
"context": {
"System": { "user": { "userId": "amzn1.ask.account.test" } }
},
"request": {
"type": "IntentRequest",
"intent": {
"name": "WhoIsIntent",
"slots": { "person": { "name": "person", "value": "Sarah" } }
}
}
}
EOF
๏ฟผ
TypeScript
Run test
node -e "
const handler = require('./index.js').handler;
const event = require('./test-event.json');
handler(event, {}, (err, result) => {
console.log('Result:', JSON.stringify(result, null, 2));
});
"
๏ฟผ
TypeScript
CloudWatch Monitoring
View Lambda logs
aws logs tail /aws/lambda/buddy-alexa-skill-dev --follow
๏ฟผ
TypeScript
Check invocation metrics
aws cloudwatch get-metric-statistics
--namespace AWS/Lambda
--metric-name Invocations
--dimensions Name=FunctionName,Value=buddy-alexa-skill-dev
--start-time $(date -u -v-1H +%Y-%m-%dT%H:%M:%S)
--end-time $(date -u +%Y-%m-%dT%H:%M:%S)
--period 60
--statistics Sum
๏ฟผ
TypeScript
๐ Security
Encryption: DynamoDB SSE (AES-256) at rest, TLS 1.2+ in transit
Authentication: JWT tokens (1-hour expiry) for caregiver API
Authorization: Caregivers can only access their assigned patients
Data Minimization: No audio stored, only text transcripts
Retention: 90-day auto-deletion of conversation logs (DynamoDB TTL)
IAM: Least-privilege access with table-specific permissions
๐
Project Timeline
Phase 1 (Jan 20-26): Core Infrastructure โ
DynamoDB tables deployed
IAM roles created
Seed data inserted
Phase 2 (Jan 27-Feb 2): Alexa Skill Implementation โ
Lambda function with Nova integration
Safety escalation system
SNS SMS alerts
Intent handlers for routines, medications, family
Phase 3 (Feb 3-7): Refinement โ
Nova Sonic speech-to-speech integration with WebSocket API
Enhanced Emergency Assist with location sharing and 30-min session
CloudWatch monitoring dashboard and alarms
Conversation logging and analytics
Phase 4 (Feb 1-7): Caregiver Dashboard โ
React dashboard with patient monitoring
Web UI implementation
API documentation
Edge case handling
Phase 5 (Mar 6-16): Demo & Submission โ
Demo video production
Devpost submission
Documentation polish
Test credentials for judges
๐ Documentation
Product Requirements Document - Full feature specification
Database Schema - DynamoDB design and access patterns
Alexa Skill README - Skill-specific documentation
Monitoring Guide - CloudWatch and safety monitoring
๐ค Contributing
This is a hackathon project. For issues or improvements, please:
Check existing issues first
Create a new issue with detailed description
Submit pull requests with clear commit messages
๐ License
MIT License - Amazon Nova AI Hackathon 2026
Status: 100% COMPLETE โ
- Ready for Submission
๏ฟผ๏ฟผ
GitHub - whoshotu/-BUDDY: An AI-powered voice assistant built specifically for dementia patients using Amazon Nova AI.
An AI-powered voice assistant built specifically for dementia patients using Amazon Nova AI. - whoshotu/-BUDDY
https://github.com/whoshotu/-BUDDY
๏ฟผ
๏ฟผ
AWS CLI configured (
aws configure)AWS Account with Nova access
Node.js 18+ (for Lambda deployment)
Python 3.9+ (for infrastructure scripts)
Phase 1: Deploy Infrastructure
This creates:
BuddyCaregivers-dev - Login credentials and notification preferencesBuddyPatients-dev - Patient profiles, routines, medications, peopleBuddyAssignments-dev - Caregiver-patient relationship mappingBuddyConversationLogs-dev - Conversation history (90-day TTL)Test Credentials:
Username:
caregiver_testPassword:
Demo2026!Patient: John Doe
Phase 2: Deploy Alexa Skill
1. Configure Environment
Copy environment template
Edit .env with your AWS values
2. Deploy Lambda Function
Deploy the Alexa skill Lambda
This creates:
Lambda function
buddy-alexa-skill-devAlexa Skills Kit trigger configured
Environment variables for DynamoDB tables
Save the Lambda ARN - you'll need it for the Alexa Developer Console.
3. Configure Alexa Developer Console
Go to Alexa Developer Console
Click "Create Skill"
Skill Name: "Buddy - Dementia Care Assistant"
Choose a model: Custom
Choose a method: Provision your own
Click "Create Skill"
Interaction Model:
Click on "JSON Editor" (left sidebar)
Delete the default content
Copy the contents of
src/alexa-skill/models/en-US.json and paste it inClick "Save Model" then "Build Model"
Endpoint Configuration:
Go to "Endpoint" (left sidebar)
Select "AWS Lambda ARN"
Paste your Lambda ARN from deployment
Click "Save Endpoints"
Account Linking (Optional):
Go to "Account Linking" (left sidebar)
Configure if you want user authentication
Test:
Go to "Test" tab
Enable testing for your skill
Test: "Alexa, open Buddy Assistant"
4. Test Your Skill
Phase 3: Refinement & Monitoring (Optional)
Deploy Nova Sonic WebSocket API for advanced speech-to-speech
Deploy CloudWatch monitoring dashboard
Nova Sonic Features:
Bidirectional streaming with barge-in support
Natural turn-taking and adaptive speech
WebSocket API for real-time conversations
Hybrid mode (traditional skill + Nova Sonic)
Emergency Assist Features:
Level 2 emergency protocol with location sharing
30-minute emergency session persistence
Dual notification system (SNS + SMS)
Real-time caregiver updates
See Nova Sonic README and Emergency Documentation for details.
Phase 4: Caregiver Dashboard
Start the React dashboard
Open
Dashboard Features:
Patient profile overview
Real-time conversation monitoring
Safety alert management
Emergency contact information
Phase 5: Testing Without Alexa Device
Test Lambda function directly
This simulates Alexa requests and tests emergency flow
See Testing Guide for comprehensive test scenarios.
๐ Project Structure
๐ง Environment Configuration
Copy
.env.example to .env and configure:๐งช Testing
Local Testing
Navigate to skill directory
Install dependencies
Create test event
Run test
CloudWatch Monitoring
View Lambda logs
Check invocation metrics
https://github.com/whoshotu/-BUDDY Buddy - Voice-First Dementia Care Assistant
AI-powered voice assistant for dementia patients, built for the Amazon Nova AI Hackathon 2026.
๐ฏ MVP Scope
Alexa Custom Skill - Patient voice interface with Amazon Nova
Caregiver Dashboard - Web interface for managing patient knowledge base
3-Level Safety Escalation - Smart alerts via Amazon SNS SMS
Core Use Cases: Routine guidance, family recognition, medication reminders
๐๏ธ Architecture
๐ Quick Start
Prerequisites
AWS CLI configured (
aws configure)AWS Account with Nova access
Node.js 18+ (for Lambda deployment)
Python 3.9+ (for infrastructure scripts)
Phase 1: Deploy Infrastructure
This creates:
BuddyCaregivers-dev - Login credentials and notification preferencesBuddyPatients-dev - Patient profiles, routines, medications, peopleBuddyAssignments-dev - Caregiver-patient relationship mappingBuddyConversationLogs-dev - Conversation history (90-day TTL)Test Credentials:
Username:
caregiver_testPassword:
Demo2026!Patient: John Doe
Phase 2: Deploy Alexa Skill
1. Configure Environment
Copy environment template
Edit .env with your AWS values
2. Deploy Lambda Function
Deploy the Alexa skill Lambda
This creates:
Lambda function
buddy-alexa-skill-devAlexa Skills Kit trigger configured
Environment variables for DynamoDB tables
Save the Lambda ARN - you'll need it for the Alexa Developer Console.
3. Configure Alexa Developer Console
Go to Alexa Developer Console
Click "Create Skill"
Skill Name: "Buddy - Dementia Care Assistant"
Choose a model: Custom
Choose a method: Provision your own
Click "Create Skill"
Interaction Model:
Click on "JSON Editor" (left sidebar)
Delete the default content
Copy the contents of
src/alexa-skill/models/en-US.json and paste it inClick "Save Model" then "Build Model"
Endpoint Configuration:
Go to "Endpoint" (left sidebar)
Select "AWS Lambda ARN"
Paste your Lambda ARN from deployment
Click "Save Endpoints"
Account Linking (Optional):
Go to "Account Linking" (left sidebar)
Configure if you want user authentication
Test:
Go to "Test" tab
Enable testing for your skill
Test: "Alexa, open Buddy Assistant"
4. Test Your Skill
Phase 3: Refinement & Monitoring (Optional)
Deploy Nova Sonic WebSocket API for advanced speech-to-speech
Deploy CloudWatch monitoring dashboard
Nova Sonic Features:
Bidirectional streaming with barge-in support
Natural turn-taking and adaptive speech
WebSocket API for real-time conversations
Hybrid mode (traditional skill + Nova Sonic)
Emergency Assist Features:
Level 2 emergency protocol with location sharing
30-minute emergency session persistence
Dual notification system (SNS + SMS)
Real-time caregiver updates
See Nova Sonic README and Emergency Documentation for details.
Phase 4: Caregiver Dashboard
Start the React dashboard
Open
Dashboard Features:
Patient profile overview
Real-time conversation monitoring
Safety alert management
Emergency contact information
Phase 5: Testing Without Alexa Device
Test Lambda function directly
This simulates Alexa requests and tests emergency flow
See Testing Guide for comprehensive test scenarios.
๐ Project Structure
๐ง Environment Configuration
Copy
.env.example to .env and configure:๐งช Testing
Local Testing
Navigate to skill directory
Install dependencies
Create test event
Run test
CloudWatch Monitoring
View Lambda logs
Check invocation metrics
๐ Security
Encryption: DynamoDB SSE (AES-256) at rest, TLS 1.2+ in transit
Authentication: JWT tokens (1-hour expiry) for caregiver API
Authorization: Caregivers can only access their assigned patients
Data Minimization: No audio stored, only text transcripts
Retention: 90-day auto-deletion of conversation logs (DynamoDB TTL)
IAM: Least-privilege access with table-specific permissions
๐
Project Timeline
Phase 1 (Jan 20-26): Core Infrastructure โ
DynamoDB tables deployed
IAM roles created
Seed data inserted
Phase 2 (Jan 27-Feb 2): Alexa Skill Implementation โ
Lambda function with Nova integration
Safety escalation system
SNS SMS alerts
Intent handlers for routines, medications, family
Phase 3 (Feb 3-7): Refinement โ
Nova Sonic speech-to-speech integration with WebSocket API
Enhanced Emergency Assist with location sharing and 30-min session
CloudWatch monitoring dashboard and alarms
Conversation logging and analytics
Phase 4 (Feb 1-7): Caregiver Dashboard โ
React dashboard with patient monitoring
Web UI implementation
API documentation
Edge case handling
Phase 5 (Mar 6-16): Demo & Submission โ
Demo video production
Devpost submission
Documentation polish
Test credentials for judges
๐ Documentation
Product Requirements Document - Full feature specification
Database Schema - DynamoDB design and access patterns
Alexa Skill README - Skill-specific documentation
Monitoring Guide - CloudWatch and safety monitoring
๐ค Contributing
This is a hackathon project. For issues or improvements, please:
Check existing issues first
Create a new issue with detailed description
Submit pull requests with clear commit messages
๐ License
MIT License - Amazon Nova AI Hackathon 2026
Status: 100% COMPLETE โ
- Ready for Submission
๐ Security
Encryption: DynamoDB SSE (AES-256) at rest, TLS 1.2+ in transit
Authentication: JWT tokens (1-hour expiry) for caregiver API
Authorization: Caregivers can only access their assigned patients
Data Minimization: No audio stored, only text transcripts
Retention: 90-day auto-deletion of conversation logs (DynamoDB TTL)
IAM: Least-privilege access with table-specific permissions
๐
Project Timeline
Phase 1 (Jan 20-26): Core Infrastructure โ
DynamoDB tables deployed
IAM roles created
Seed data inserted
Phase 2 (Jan 27-Feb 2): Alexa Skill Implementation โ
Lambda function with Nova integration
Safety escalation system
SNS SMS alerts
Intent handlers for routines, medications, family
Phase 3 (Feb 3-7): Refinement โ
Nova Sonic speech-to-speech integration with WebSocket API
Enhanced Emergency Assist with location sharing and 30-min session
CloudWatch monitoring dashboard and alarms
Conversation logging and analytics
Phase 4 (Feb 1-7): Caregiver Dashboard โ
React dashboard with patient monitoring
Web UI implementation
API documentation
Edge case handling
Phase 5 (Mar 6-16): Demo & Submission โ
Demo video production
Devpost submission
Documentation polish
Test credentials for judges
๐ Documentation
Product Requirements Document - Full feature specification
Database Schema - DynamoDB design and access patterns
Alexa Skill README - Skill-specific documentation
Monitoring Guide - CloudWatch and safety monitoring
๐ค Contributing
This is a hackathon project. For issues or improvements, please:
Check existing issues first
Create a new issue with detailed description
Submit pull requests with clear commit messages
๐ License
MIT License - Amazon Nova AI Hackathon 2026
Status: 100% COMPLETE โ
- Ready for Submission
1 like ยท 8 views ยท Jun 17, 2026
Developed Buddy, a voice-first dementia care assistant for Amazon hackathon.
Timeline
Jan 19, 2026 - Mar 15, 2026