Procfile

# Render reads this file to know how to start the web service.
# --workers 1   : single worker keeps RAM under free-tier limit (~512 MB)
# --timeout 300 : prediction tasks can run up to 5 minutes before Gunicorn kills them
# --bind        : Render injects $PORT automatically
web: gunicorn app:app --bind 0.0.0.0:$PORT --timeout 300 --workers 1