Skip to content

DEPLOYMENT_GUIDE.md

Latest commit

 

History

History
379 lines (292 loc) · 8.79 KB

DEPLOYMENT_GUIDE.md

File metadata and controls

379 lines (292 loc) · 8.79 KB

🚀 Deployment Guide - Therapeutic AI Assistant

This guide will walk you through deploying your Therapeutic AI Assistant to production.

📋 Prerequisites

  • Node.js 18+ installed
  • npm or yarn package manager
  • Docker Desktop (for TTS service)
  • Firebase account
  • Stripe account
  • Upstash Redis account (for rate limiting)
  • Domain name purchased

🔧 Quick Start

  1. Run the automated setup:

    chmod +x scripts/setup-production.sh
./scripts/setup-production.sh

  2. Validate your setup:

    chmod +x scripts/deploy-check.sh
./scripts/deploy-check.sh

  3. If all checks pass, proceed to deployment section below.

🔐 Security Features Implemented

✅ Rate Limiting

  • Chat API: 20 requests/minute per user
  • Voice API: 10 requests/minute per user
  • Auth endpoints: 5 requests/5 minutes per IP
  • Payment API: 3 requests/minute per user

✅ Security Headers

  • Content Security Policy (CSP)
  • X-Frame-Options: DENY
  • X-Content-Type-Options: nosniff
  • Strict-Transport-Security
  • X-XSS-Protection

✅ Input Sanitization

  • DOMPurify integration
  • Zod validation with sanitization
  • XSS prevention
  • SQL injection protection

✅ CORS Configuration

  • Configurable allowed origins
  • Proper preflight handling
  • Credential support

🛠️ Manual Setup Steps

Step 1: Environment Configuration

  1. Copy environment template:

    cp .env.production.template .env.local

  2. Fill in your API keys in .env.local:

    Firebase (Required):

    NEXT_PUBLIC_FIREBASE_API_KEY=your_api_key
NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN=your_project.firebaseapp.com
NEXT_PUBLIC_FIREBASE_PROJECT_ID=your_project_id
# ... other Firebase config

    Stripe (Required):

    NEXT_PUBLIC_STRIPE_PUBLISHABLE_KEY=pk_live_xxxxx  # Use LIVE keys for production
STRIPE_SECRET_KEY=sk_live_xxxxx
STRIPE_WEBHOOK_SECRET=whsec_xxxxx

    Upstash Redis (Required for rate limiting):

    UPSTASH_REDIS_REST_URL=https://xxx.upstash.io
UPSTASH_REDIS_REST_TOKEN=xxxxx

    Domain Configuration:

    NEXT_PUBLIC_APP_URL=https://yourdomain.com
ALLOWED_ORIGINS=https://yourdomain.com,https://www.yourdomain.com

Step 2: Firebase Setup

  1. Update Firebase authorized domains:

    • Go to Firebase Console → Authentication → Settings
    • Add your domain to "Authorized domains"

  2. Configure Firestore security rules:

    rules_version = '2';
service cloud.firestore {
  match /databases/{database}/documents {
    match /users/{userId} {
      allow read, write: if request.auth != null && request.auth.uid == userId;
    }
  }
}

Step 3: Stripe Setup

  1. Create products and prices in Stripe Dashboard
  2. Configure webhook endpoint:
    • URL: https://yourdomain.com/api/payment/webhook
    • Events: checkout.session.completed, customer.subscription.*

Step 4: TTS Service Deployment

Option A: Co-located with app (simplest)

export COQUI_TOS_AGREED=1
docker-compose -f docker-compose.tts.yml up -d

Option B: Separate server

  • Deploy TTS service to dedicated server
  • Update XTTS_API_URL in environment variables

Step 5: Build and Test

npm install
npm run build
npm run test  # Optional: run tests

🌐 Deployment Options

Option 1: Vercel (Recommended)

  1. Connect repository:

  2. Configure environment variables:

    • Copy all variables from .env.local to Vercel dashboard
    • Go to Project Settings → Environment Variables

  3. Configure custom domain:

    • Go to Project Settings → Domains
    • Add your custom domain
    • Update DNS records as instructed

  4. Deploy:

    vercel --prod

Option 2: Self-Hosted

  1. Build for production:

    npm run build
npm start

  2. Use PM2 for process management:

    npm install -g pm2
pm2 start npm -- start
pm2 startup
pm2 save

  3. Configure reverse proxy (Nginx example):

    server {
    listen 80;
    server_name yourdomain.com;
    return 301 https://$server_name$request_uri;
}

server {
    listen 443 ssl;
    server_name yourdomain.com;
    
    ssl_certificate /path/to/cert.pem;
    ssl_certificate_key /path/to/key.pem;
    
    location / {
        proxy_pass http://localhost:3000;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection 'upgrade';
        proxy_set_header Host $host;
        proxy_cache_bypass $http_upgrade;
    }
}

🔍 Post-Deployment Testing

Critical Tests

  1. Authentication Flow:

    • Sign up with email/password
    • Login with Google
    • Password reset

  2. Payment Flow:

    • Subscribe to premium plan
    • Payment success/failure handling
    • Webhook delivery

  3. Chat Functionality:

    • Send messages
    • Receive AI responses
    • Rate limiting kicks in after limits

  4. Voice Synthesis:

    • Generate voice from text
    • Different voice options work
    • Audio playback in browser

  5. Security Headers:

Testing Commands

# Test rate limiting
curl -X POST https://yourdomain.com/api/chat \
  -H "Content-Type: application/json" \
  -d '{"message":"test"}' \
  --rate 30/m  # Should trigger rate limit

# Test security headers
curl -I https://yourdomain.com/

# Test TTS service
curl -X POST https://yourdomain.com/api/voice \
  -H "Content-Type: application/json" \
  -d '{"text":"Hello world","voiceId":"female-1"}'

📊 Monitoring Setup

Error Tracking (Recommended)

  1. Sentry:

    SENTRY_DSN=https://xxxxx@sentry.io/xxxxx

  2. Vercel Analytics:

    • Automatically enabled on Vercel Pro plan

Uptime Monitoring

  1. UptimeRobot (free): https://uptimerobot.com/
  2. Pingdom: https://pingdom.com/

Performance Monitoring

  1. Vercel Speed Insights
  2. Google PageSpeed Insights
  3. Web.dev Measure

🚨 Troubleshooting

Common Issues

1. Build Fails:

# Clear Next.js cache
rm -rf .next
npm run build

2. CORS Errors:

  • Check ALLOWED_ORIGINS in .env.local
  • Verify middleware configuration
  • Check browser developer tools

3. Rate Limiting Not Working:

  • Verify Upstash Redis credentials
  • Check Redis connection in logs
  • Ensure middleware is running

4. TTS Service Not Responding:

# Check TTS service status
docker ps | grep xtts

# Check TTS logs
docker-compose -f docker-compose.tts.yml logs -f

# Restart TTS service
docker-compose -f docker-compose.tts.yml restart

5. Firebase Auth Issues:

  • Check authorized domains in Firebase Console
  • Verify API keys are correct
  • Check browser network tab for errors

Performance Issues

Large Bundle Size:

# Analyze bundle
npm install -g @next/bundle-analyzer
ANALYZE=true npm run build

Slow API Responses:

  • Check database query performance
  • Implement caching for frequent requests
  • Optimize AI model inference

📈 Scaling Considerations

Traffic Growth

  • Vercel: Automatically scales serverless functions
  • Self-hosted: Use load balancer + multiple instances
  • Database: Consider migrating from Firebase to PostgreSQL
  • TTS: Deploy multiple TTS instances with load balancer

Cost Optimization

  • Monitor Vercel/Firebase usage
  • Implement response caching
  • Optimize bundle size
  • Use CDN for static assets

🔒 Security Maintenance

Regular Tasks

  • Update dependencies monthly: npm audit fix
  • Rotate API keys quarterly
  • Review security headers annually
  • Monitor error logs weekly
  • Check uptime monitoring alerts

Security Checklist

  • All dependencies up to date
  • No critical security vulnerabilities
  • Rate limiting functioning
  • HTTPS enforced
  • Security headers present
  • Input sanitization working
  • Error messages don't leak info

📞 Support

If you encounter issues:

  1. Check logs: Vercel dashboard or server logs
  2. Run deployment check: ./scripts/deploy-check.sh
  3. Review environment variables
  4. Test individual components

🎉 Success Criteria

Your deployment is successful when:

  • ✅ All API endpoints respond correctly
  • ✅ Authentication flow works end-to-end
  • ✅ Payment processing completes successfully
  • ✅ Voice synthesis generates audio
  • ✅ Rate limiting prevents abuse
  • ✅ Security headers are present
  • ✅ No critical errors in logs
  • ✅ Performance metrics are acceptable

Congratulations! Your Therapeutic AI Assistant is now live! 🚀