RENDER_QUICKSTART.md

🚀 QUICK DEPLOYMENT GUIDE

✅ Files Changed (Production-Ready)

1. ✅ requirements.txt - Added psycopg2-binary, gunicorn
2. ✅ app/database.py - Hybrid SQLite/PostgreSQL support with Render URL fix
3. ✅ app/config.py - Environment variable support for SECRET_KEY
4. ✅ main.py - Auto-migration, production settings

📋 Render Configuration

Web Service Settings

Name: vietnamese-rating-prediction
Runtime: Python 3
Build Command: pip install -r requirements.txt
Start Command: gunicorn main:app --workers 4 --worker-class uvicorn.workers.UvicornWorker --bind 0.0.0.0:$PORT

Environment Variables (Required)

SECRET_KEY = <generate-with-openssl-rand-hex-32>
PYTHON_VERSION = 3.11.0

PostgreSQL Database

Name: vietnamese-rating-db
PostgreSQL Version: 15
Instance Type: Free

Link database to web service - DATABASE_URL will be auto-populated.

---

🎯 Deployment Steps

1. Push to GitHub

git add .
git commit -m "Deploy to Render"
git push origin master

2. Create Render Web Service

• Go to https://dashboard.render.com/
• New → Web Service
• Connect GitHub repo
• Use settings above

3. Create PostgreSQL Database

• New → PostgreSQL
• Use free tier
• Link to web service

4. Deploy

• Click "Manual Deploy"
• Watch logs for success

5. Test

https://your-app.onrender.com/health
https://your-app.onrender.com/docs
https://your-app.onrender.com/dashboard

---

🔧 Local Testing Before Deploy

Test hybrid database locally:

# Test with SQLite (no DATABASE_URL)
python main.py

# Test with PostgreSQL (set DATABASE_URL)
export DATABASE_URL=postgresql://user:pass@localhost/dbname
python main.py

Expected output:

🔧 Development Mode: Using SQLite
# OR
🚀 Production Mode: Using PostgreSQL
🔄 Creating database tables...
✅ Database tables created successfully!

---

⚠️ Important Notes

1. Render Free Tier Limitations:

   • App sleeps after 15 minutes of inactivity (first request takes 30-60s)
   • 512MB RAM (may need optimization for ML model)
   • 1GB PostgreSQL storage

2. ML Model Optimization:

   • Consider lazy loading (load on first request)
   • Use CPU-optimized PyTorch
   • Cache predictions if possible

3. Static Files:

   • Uploads are ephemeral on Render Free Tier
   • WordClouds will be deleted on container restart
   • Use cloud storage (S3, Cloudinary) for production

4. Database:

   • SQLite NOT recommended for production (file locking issues)
   • PostgreSQL required for concurrent requests
   • Free tier: 1GB storage, 97 connections

---

🆘 Common Issues

"Module not found"

→ Run pip install -r requirements.txt locally first

"Port binding error"

→ Use $PORT in start command (auto-set by Render)

"Database connection failed"

→ Check DATABASE_URL in environment variables

"Model loading timeout"

→ Free tier has 512MB RAM limit, optimize model or upgrade

---

Read DEPLOYMENT.md for detailed guide!