Skip to content

Prepare Akiki 0.2.2 - #9

Merged
tobyrjackson merged 1 commit into
mainfrom
chore/prepare-0.2.2
Nov 17, 2026
Merged

tobyrjackson merged 1 commit into
mainfrom
chore/prepare-0.2.2

Conversation

@tobyrjackson

Copy link
Copy Markdown
Contributor

Prepare Akiki 0.2.2 from the reviewed main branch. Update the package version, runtime version, version assertion, release manifest, and model card. The model identity, weight checksums, and engine and training source pins are preserved.

Release notes:

  • Add Akiki engine wheels for Windows x64, macOS Intel, and macOS Apple Silicon. Each wheel supports VAD and the current MiniLM/BERT decision runtime.
  • Limit automatic native-library discovery to the package's native directory. Use library_path or the engine-specific environment variables for a separately built library.
  • Retry temporary Windows file-sharing errors during atomic cache updates.
  • Add Windows and macOS installation and inference examples. Describe VAD and decision models as the primary interfaces.
  • Keep the training implementation in the private akiki-training companion. The public wrapper supports authorized local installations. The training extra supplies runtime dependencies only.

Validation: 456 tests passed offline with both existing local native engines required. Eleven checks were skipped because ONNX Runtime or an explicit real-model path was unavailable. Ruff and formatting checks passed. The packaging tests built and verified distributions with the 0.2.2 metadata. The version is not yet present on PyPI.

After merge, wait for the full builds and private training checks on main. Publish a GitHub release tagged v0.2.2 at that tested commit to start Trusted Publishing.

@tobyrjackson
tobyrjackson merged commit 2fc5a82 into main Nov 17, 2026
4 checks passed
@tobyrjackson
tobyrjackson deleted the chore/prepare-0.2.2 branch November 17, 2026 20:32
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