Flask — Deployment Walk-Through

• Prepare a Flask app for production — gunicorn, requirements.txt, env vars.
• Push to GitHub.
• Deploy to one free host (Render, Fly.io, or PythonAnywhere).
• Debug the inevitable "500 in prod but works locally".

• ✅ requirements.txt — every pip dependency pinned.
• ✅ app.run(debug=True) only inside if __name__ == "__main__".
• ✅ SECRET_KEY, DB path read from environment.
• ✅ A WSGI server (gunicorn) instead of the dev server.
• ✅ Static files served by Flask is fine for small apps — large traffic moves to a CDN.

requirements.txt

pip freeze > requirements.txt

# typical contents:
Flask==3.0.3
gunicorn==22.0.0
bcrypt==4.2.0

gunicorn

Flask's built-in dev server is single-threaded and warns it's not for production. gunicorn is a real WSGI server:

pip install gunicorn

# Locally test it
gunicorn -w 2 -b 0.0.0.0:8000 app:app
# -w 2  = 2 worker processes
# app:app = filename:Flask-instance

Pick env vars carefully

import os
app.config["SECRET_KEY"] = os.environ["FLASK_SECRET"]
DB_PATH = os.environ.get("DB_PATH", "blog.db")

Three friendly hosts

Render (render.com)
  - Free tier with cold starts
  - "Web Service" → connect GitHub → Python → start: gunicorn app:app
  - Set env vars in dashboard
  - Persistent disk (paid) for SQLite

Fly.io (fly.io)
  - Free allowance; CLI-driven (fly launch + fly deploy)
  - flyctl detects Flask, writes a Dockerfile
  - Volume mount for SQLite persistence

PythonAnywhere (pythonanywhere.com)
  - Easiest for absolute beginners; web UI, no Docker
  - "Web app" → Flask → point to wsgi.py
  - SQLite lives on their disk by default

The deployment shape

1. push code to GitHub
2. create new service on host
3. point host at your repo + branch
4. set env vars (FLASK_SECRET at minimum)
5. set start command (gunicorn app:app)
6. wait 60 s for first build
7. visit the URL
8. read logs when something is wrong

1. Push to GitHub.

   git init
   git add app.py blog_db.py templates/ static/ requirements.txt
   git commit -m "blog v1"
   git branch -M main
   gh repo create my-flask-blog --public --source=. --push   # or use the web UI

2. Sign in to render.com → New → Web Service → Connect repo.
3. Settings:

   Environment        : Python 3
   Build command      : pip install -r requirements.txt
   Start command      : gunicorn app:app
   Instance type      : Free
   
   Environment variables
     FLASK_SECRET = <paste a 64-hex secret>

4. Click Deploy. Watch the build log. Two minutes later you have a URL like https://my-flask-blog.onrender.com.
5. Visit the URL on your phone 🎉

Common 500-in-prod culprits

• FLASK_SECRET not set → KeyError on import.
• SQLite file lost after each deploy on ephemeral disks. Use a volume (paid on Render) or migrate to Postgres.
• Static URLs hardcoded as /static/... work, but url_for("static", ...) is safer.
• Logs are gold — every host shows them; read them first.

Production = gunicorn + env vars + a host. Push, configure, deploy, read logs. Once. Future apps follow the same shape. Tomorrow we step back into the "why" — data ethics.