Look into exposing tasks as CLI tools

[jluethi]

When exploring how to call the Fractal tasks core tasks through Galaxy, the conda distribution has become relevant (so great that we kept maintaining this! :) ). What we would actually need though are entry points to the different task init/compute executables. Instead of doing:

/data/app/Tasks/1/fractal-tasks-core/1.5.0a2/venv/bin/python /data/app/Tasks/1/fractal-tasks-core/1.5.0a2/venv/lib/python3.12/site-packages/fractal_tasks_core/tasks/cellvoyager_to_ome_zarr_init.py

We'd want to do something like:

. /data/app/Tasks/1/fractal-tasks-core/1.5.0a2/venv/bin/activate 
cellvoyager_to_ome_zarr_init

We'd do this by exposing the tasks for the CLI by adding something to the pyproject.toml like:

[project.scripts]
cellvoyager-to-ome-zarr-init = "fractal_tasks_core.tasks.cellvoyager_to_ome_zarr_init"

This would greatly simplify how the tasks can be used from the CLI by users or other workflow managers. If it works well, we could look into whether that's a pattern we also establish in the template package and how we'd make it easy to maintain that.

(obviously not a 1.5.0 priority, just creating this issue to track this useful idea from the hackathon)

[tcompa]

the conda distribution has become relevant

Side remark: note that installing the conda package does not install any of the optional dependencies (like cellpose, scikit-image, or napari-workflows) - see https://github.com/conda-forge/fractal-tasks-core-feedstock/blob/main/recipe/meta.yaml#L36-L47. Ref #611.

[jluethi]

Thanks for the heads-up for that!