Containerization with Docker

If dockerfile is set to "y", a simple Dockerfile is added to the repository. The Dockerfile installs pixi, sets up the environment and runs foo.py when run.

The container image now bakes only the runtime dependencies and expects project files (src, tests, notebooks, input) to be mounted as a volume at runtime. This better simulates running the same code/data snapshot in another environment (for example, HPC execution contexts).

The generated pixi.toml also includes container workflow tasks to simulate running model workflows in a reproducible environment (for example, to mimic HPC-style execution):

pixi run img
pixi run drun
pixi run dnb
pixi run dtest

dnb executes notebooks/workflow.ipynb with jupyter nbconvert --execute and writes workflow-executed.ipynb. Adjust the notebook path to match your project notebooks.

The docker image can be built with

docker build . -t my-docker-image

It can then be run against your project directory with a mounted volume, for example:

docker run --rm -v "$PWD":/workspace -w /workspace my-docker-image

Or run it in interactive mode with

docker run --rm -it -v "$PWD":/workspace -w /workspace --entrypoint bash my-docker-image