Problem

Decide on a deployment strategy. Options:

1. poetry, dockerize, PaaS (DigitalOcean Apps Platform)
2. DigitalOcean Functions (see https://www.digitalocean.com/community/conceptual_articles/how-to-write-a-serverless-function). This approach does not seem compatible with using FastAPI.
3. poetry, non-dockerize, IaaS (digital ocean). Choose this for simplicity. (without docker, the deployment seems fragile)
4. poetry, heroku (not sure if this will work)
5. deta (can't specify poetry as run command, might need to turn the maze portion into a package that the fastapi app depends on)

Then, execute this deployment strategy.

Acceptance Criteria

• Maze can be deployed to production following some repeatable steps.
• The repeatable steps are documented.

Solution Approach

Use DigitalOcean's App Platform given we already have a working Dockerfile. App Platform is optimized for ease of code deployment rather than deep customization of the underlying infrastructure. This approach relies on DO to analyze and build code from Github (the current chosen source control repo), and sidestep various DevOps complexities. If we want more control to customize deployment, we can revisit. For now, this seems like a good way to deploy to production.

Steps to deploy to production

1. Do a local build from Dockerfile. More details to follow.
2. Use DO's 'App Spec' to define the repo/branch to pull the code from to do the docker build. (For now, choose github and main branch)
3. In DO's UI, kick off a deploy (in case there is no changes to the source control) Rely on Apps Platform for TLS connection, and other security features.

cjolowicz/cookiecutter-hypermodern-python#1244 explains this pre-commit check fail when attempting to upgrade flake8 from 4.x to 5.0.

Consider removing flake8-bandit and run bandit directly instead.