Skip to content

0.3.0 - #154

Merged
davidackerman merged 2 commits into
mainfrom
release-0.3.0
Oct 22, 2026
Merged

0.3.0#154
davidackerman merged 2 commits into
mainfrom
release-0.3.0

Conversation

@davidackerman

@davidackerman davidackerman commented Oct 22, 2026 •

Copy link
Copy Markdown
Collaborator

Bumps the version to 0.3.0 in the three places that carry it (cellmap_flow/__init__.py, pixi.toml, Fileglancer's runnables.yaml) and dates the changelog's Unreleased section, which lists everything since 0.2.3 (2026-03-19): the cleanup (#103), the browser app, finetuning in both modes, serving on chunkmirage, the worldarray reader, the consolidation (#150) and the paint logic written once (#152).

Publishing: merging this and pushing the tag 0.3.0 on the merge commit runs .github/workflows/publish.yaml, which builds and uploads to PyPI (the token secret, as the workflow is written). Checked locally on this commit: python -m build and twine check pass; the wheel installs only cellmap_flow; its 48 requirements carry no direct URL (the earlier chunkmirage @ git+… would have made PyPI refuse the upload; it is chunkmirage>=0.1.0a2,<0.2 since #152), worldarray>=0.1,<1 and upstream neuroglancer.

What a pip install gets vs pixi: viewing, inference and training work from PyPI; painting in server mode also needs the neuroglancer voxel-annotation fork (not on PyPI; pip install git+https://github.com/briossant/neuroglancer@fae54db8…, the commit pixi.toml pins — the pyproject comment now names it) and the MinIO binaries (conda-forge). pixi and the Fileglancer deployment get both from the lockfile as before.

Kept this release, deliberately: the pre-0.3.0 console scripts (cellmap_flow_yaml, _view, _blockwise, _blockwise_multiple, _app, _server) and the compatibility shims. salamander calls the scripts in 34 files and imports five private names; skoots-mito calls cellmap_flow_blockwise twice. They go in the release after this one, once those repositories have moved to cellmap_flow <subcommand>.

The neuroglancer fork, decided

The wheel depends on upstream neuroglancer from PyPI; the paint fork (google/neuroglancer#858, not on PyPI) stays pixi's. So pip install cellmap-flow serves models and runs the inference dashboard, and painting needs pixi or the fork's git install. Two fixes in the last commit make that visible instead of silent: the dashboard logs one warning at start when the installed neuroglancer cannot paint, with the install line, and cellmap_flow doctor's install line names the commit pixi pins (a test keeps the two together). Publishing the fork on PyPI under its own name was weighed and left for when pip users who paint show up.

https://claude.ai/code/session_01EzWMS4pG3tm1ZtvKULfSfW

The first release since 0.2.3 (2026-03-19): the cleanup (#103), the browser
app, finetuning in both modes, serving on chunkmirage, the worldarray
reader, the consolidation (#150) and the paint logic written once (#152),
as the changelog's Unreleased section listed them; it is dated here. The
version moves in the three places that carry it (the package, pixi.toml,
Fileglancer's runnables.yaml). The commands from before 0.3.0 keep working
as aliases this release and go in the next, with the shims, once
salamander, skoots-mito and cellmap-analyze have moved off them. The
pyproject comment on painting names the neuroglancer fork commit pixi
pins, not the moving branch.

Claude-Session: https://claude.ai/code/session_01EzWMS4pG3tm1ZtvKULfSfW
…e doctor's fork install to pixi's commit

The PyPI release depends on upstream neuroglancer, whose client has none of
the paint fork's tools; the fork is installed by pixi. A pip install now hears
why at start (cli.doctor.painting_unavailable, the doctor's check), with the
install line, and that line names the commit pixi pins instead of the fork's
branch; tests/web/test_pins.py keeps the two the same.

Claude-Session: https://claude.ai/code/session_01EzWMS4pG3tm1ZtvKULfSfW
@davidackerman davidackerman changed the title Cut 0.3.0 0.3.0 Oct 22, 2026
@davidackerman
davidackerman marked this pull request as ready for review October 22, 2026 13:38
@davidackerman
davidackerman merged commit fa8c77b into main Oct 22, 2026
7 checks passed
@davidackerman
davidackerman deleted the release-0.3.0 branch October 22, 2026 13:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant