v0.3.1 — metadata & docs polish

Richer PyPI metadata and project links (homepage, changelog, docs), clarified --fast scope in --help (applies to --from-list only), expanded README FAQ, new landing page at https://boogy-ro.github.io/framepin/

v0.3.0 — pinfiles: framepin pin + verify --against-file

What's new

• framepin pin DATASET — snapshot and write the version id to a committed one-line pinfile (default framepin.pin). Also --from-list, --split, --file, and --version <id> to pin an already-stored version without re-hashing.
• verify --against-file framepin.pin — CI reads the pinned version from the pinfile. Accepting an intentional data change = re-run framepin pin + commit. No more hardcoded hashes in CI config.
• GitHub Actions templates in examples/ci/: a verify gate, and a workflow that opens a baseline-update PR automatically when the data drifts.

95 tests, Python ≥3.9, no dependencies. pip install framepin

v0.2.0 — CI allow-lists, per-split manifests, MLflow/W&B export

What's new

• verify --allow GLOB — allow-list expected churn in the CI gate: if every drifted path matches an allow glob, the gate still passes. --json gains gate and allowed fields.
• snapshot --split NAME=GLOB — per-split version ids for stratified datasets (train/val/test). Pure metadata: the top-level version id is unchanged. Drift reports now say which split moved.
• framepin.integrations.to_mlflow(run) / to_wandb(run) — tag the active MLflow/W&B run with the pinned dataset version ids. Lazy imports; framepin stays zero-dependency.

79 tests, Python ≥3.9, no dependencies. pip install framepin