Flask Hosting with Gunicorn and Git Deploy

Flask is the app a VeloxaHost Python site expects out of the box. A new Python site starts with a placeholder Flask app, and its default start command is gunicorn -b 127.0.0.1:8000 app:app: Gunicorn serving the app object from app.py. If your project follows that layout, you deploy it without setting anything.

Each site is its own container with Python 3.11 and a virtual environment at /opt/app/venv. The app runs as a system service called veloxa-app. The site's web server answers on your domain with a free Let's Encrypt certificate and forwards requests to Gunicorn on 127.0.0.1:8000.

Deploy from Git

Prepare the repository

Put a requirements.txt at the root of the app with Flask and Gunicorn in it. The deploy installs from that file; it does not read pyproject.toml.

Flask>=3.1
gunicorn

Keep the app in app.py as app = Flask(__name__) and the default command works. With an app factory, use a start command such as gunicorn -b 127.0.0.1:8000 -w 2 'hello:create_app()', the form Flask's own Gunicorn guide shows.

Behind the site's proxy, wrap the app with Werkzeug's ProxyFix so request.remote_addr is the visitor's address and generated URLs use your domain and https:

from werkzeug.middleware.proxy_fix import ProxyFix

app.wsgi_app = ProxyFix(app.wsgi_app, x_for=1, x_proto=1, x_host=1)

One proxy sets these headers, so 1 is the right count.

Create and deploy the site

1. In the console, open Websites and click Add Website.
2. Enter your domain and pick Python under Hosting type.
3. Leave Start Command empty for the default, or enter your own. It runs from /opt/app with the virtual environment's programs, so it must start with something your requirements install, such as gunicorn, and listen on 127.0.0.1:8000.
4. Click Create Website, then add the DNS records shown under DNS — point your domain here. HTTPS is issued once they resolve.
5. In the Deploy from Git card, click Deploy and fill in the Repository URL (https), the Branch, an optional Subdirectory, and Auto-deploy: Manual / webhook only, Every hour or Every day.
6. Click Deploy Now. The deploy makes a shallow clone, replaces the contents of /opt/app with it (keeping venv), runs pip install -r requirements.txt and restarts the app. Refresh the page to see the result under Recent deploys.
7. For push-to-deploy, POST to the CI Webhook URL from your CI, for example curl -fsS -X POST "$VELOXAHOST_DEPLOY_HOOK". Only one deploy runs per site at a time.

Secrets and settings

Open the console (Advanced > Open Console) and add them to the app's service, where deploys do not touch them:

systemctl edit veloxa-app
# [Service]
# Environment=SECRET_KEY=change-me
# Environment=DATABASE_URL=postgresql://...
systemctl restart veloxa-app

Read them in Flask with os.environ. A .env file in the repository would be public, because the deploy only clones public repositories.

Supported versions

Component	Version
Python	3.11, from Debian 12's packages
Flask	3.1 (supports Python 3.9 and newer)
Gunicorn	Whatever your requirements.txt installs

Python 3.11 gets security fixes until October 2027. There is no version picker; for another Python, use a VPS.

Which plan fits

A Flask app is light. Launch ($6.99/month: 1 vCPU, 1 GB RAM, 3 websites) runs a small app with two or three Gunicorn workers. Grow ($12.99/month: 2 vCPU, 4 GB RAM, 10 websites) suits busier apps or several of them. Each site can use up to its plan's RAM and vCPU, and gets 5 GB of disk on Launch and Grow, 2 GB on Scale and Business. Yearly billing is ten times the monthly price. Compare them on pricing.

Limits and when to use a VPS

• No database server on a Python site. SQLite works if you keep the file outside /opt/app, because each deploy replaces that folder. For PostgreSQL, MySQL or Redis, run them on a VPS.
• One process. Gunicorn workers are fine; a Celery or RQ worker, or a scheduler, needs a process manager the site does not offer. A VPS with the Python (Docker) app runs those.
• No build step. A deploy installs requirements and restarts; that is all.
• Five minutes for the install. A failed pip install does not fail the deploy today, so check journalctl -u veloxa-app if the app stops answering after a deploy.

For a bigger framework with an admin and migrations, see Django hosting; for async APIs, see FastAPI hosting.

FAQ

Do I need to set a start command for Flask?

Not if your app object is app in app.py and Gunicorn is in requirements.txt. The default command is gunicorn -b 127.0.0.1:8000 app:app.

Which port should Flask listen on?

Gunicorn should bind 127.0.0.1:8000. The site's web server forwards your domain there and handles HTTPS, so the app speaks plain HTTP.

Can I use the Flask development server?

No. flask run is for development. Use Gunicorn, which the default start command already does.

How do I see errors from my app?

Gunicorn writes to the service's journal. In the console, run journalctl -u veloxa-app -n 100 to read the last hundred lines.

Can I deploy a private repository?

Not today. The Git deploy clones public repositories over https and stores no credentials or deploy keys.