AnnZarro 0.2.0

First release on PyPI, with desktop apps for macOS, Windows and Linux.

pip install annzarro            # or: pip install "annzarro[remote]" for s3://, gs://, https:// stores
annzarro start --data-dir /path/to/stores

Documentation: https://annzarro.readthedocs.io

Highlights

• Faster data transfer: vectors travel as binary float32, sparse rows are read in bounded chunks, and responses are cached (ETag) and gzipped.
• h5ad support alongside zarr, with the same answers from both readers.
• Stable cell subsets for large datasets: a default 100,000-cell subset above 200,000 cells, seeded and nested (a larger subset only adds cells), optionally balanced by an obs column and filtered, and recorded in share links.
• Interface: a Hover picker for the columns in the hover label, the header kept on one row, tables that follow the focused cell and gene, typeahead search for genes and cells, and clearer coverage notices.
• Hosting: security tiers for what the server exposes, login and session fixes, reverse-proxy path prefixes, and CORS off by default.
• Desktop apps: a bundled server that works fully offline, downloads of 122 to 132 MB, and a start-up that always ends in the app or an error page with Retry.
• Documentation: tutorials, a user guide, data preparation, deployment and a full reference.
• Lighter install: pandas and cryptography are no longer required.

Desktop apps

The builds are not code-signed yet; see first launch for the one extra step on macOS and Windows. Verify downloads against SHA256SUMS.txt. If you have the 2025 desktop app (v0.1.1) installed, delete it: it needs a separate Python, and on quit it can stop other Python processes.

v0.1.1

Bump version to 0.1.1

v0.1.0

update tag, original never finished