Shrinking ML Package Build | Generated by AI
Done! Here’s what happened:
RELEASE LIVE: https://github.com/lzwjava/ww/releases/tag/v0.1.0
What was done:
- Excluded ww/ml/ from build artifacts
- ww/ml/ had 133MB of ML datasets/models (.h5, .pkl.gz, .csv, .wav)
- Added hatch build config in pyproject.toml to exclude it
- Wheel: 87MB → 713KB, sdist: 87MB → 687KB
- Created GitHub Actions release workflow
- File: .github/workflows/release.yml
- Triggers on tag push (v*)
- Builds wheel + sdist with uv build
- Creates GitHub release with auto-generated release notes
- Attaches .whl and .tar.gz as downloadable assets
- Published v0.1.0 release
- Tag v0.1.0 created and pushed
- Release created by github-actions[bot] with both artifacts
How users install:
From GitHub release (download .whl)
pip install ww-0.1.0-py3-none-any.whl
Or directly from the repo
pip install git+https://github.com/lzwjava/ww.git@v0.1.0
Future releases — just bump version in pyproject.toml and tag:
bump version in pyproject.toml to 0.2.0, commit, then
git tag v0.2.0 git push origin main –tags
The workflow auto-builds and publishes the release.