Put on napari hub

[constantinpape]

https://www.napari-hub.org/

[GenevieveBuckley]

Discussed previously here: https://forum.image.sc/t/publish-napari-plugin-with-conda-dependencies/65653

Summarizing:

• We need to release micro-sam on PyPI. Then napari-hub will automatically scrape PyPI and will find micro-sam, because we already have the Framework :: napari classifier.
• Because micro-sam requires some dependencies only available on conda (not pypi), we will...
  • Not worry about the fact it will not install everything correctly if users try pip install micro-sam (this is not great, but there is no good alternative if we want micro-sam discoverable on the napari-hub)
  • Add a line to the documentation (maybe around here) to tell users please do not use pip to install micro-sam. Hopefully anyone with installation problems will see this page and then try again with conda or from source.

[constantinpape]

Thanks for looking into this @GenevieveBuckley. Let's go ahead and put a "dummy" micro_sam package on pip then so that it can be installed via napari.

Would you have time to look into this? (There's no hurry!)

[GenevieveBuckley]

Thanks for looking into this @GenevieveBuckley. Let's go ahead and put a "dummy" micro_sam package on pip then so that it can be installed via napari.

Would you have time to look into this? (There's no hurry!)

I've made a PR here: https://github.com/computational-cell-analytics/micro-sam/pull/606

[constantinpape]

I have made a new release now, but the deploy part of the action that would push the package to PyPI was skipped, see https://github.com/computational-cell-analytics/micro-sam/actions/runs/9222705341.

I can't see why in the log; any idea why this may happen @GenevieveBuckley?

[GenevieveBuckley]

I have made a new release now, but the deploy part of the action that would push the package to PyPI was skipped, see https://github.com/computational-cell-analytics/micro-sam/actions/runs/9222705341.

I can't see why in the log; any idea why this may happen @GenevieveBuckley?

There are two conditions for the deploy job:

1. The previous update_release_draft job in the release_drafter workflow completed successfully, and
2. If there is a tag if: contains(github.ref, 'tags')

I was assuming that the deploy job would see that the first job created a new tag, but perhaps this is not how things work. Maybe because that tag is being created from inside the update_release_draft job?

[GenevieveBuckley]

I guess the main question is: how does the current release drafter work? Should the deploy job be turned into a step that runs after the others (possibly conditional on the results from a previous step?) Is there ever a case where a release could be drafted, but we don't want anything uploaded to pypi?