Datasets:

---

SlayerLab

/

polish-dynaword

Like 29

Follow

Fabryka AI 101

kultura_bez_barier_pl has release: None and is not in v0.2.6, but main's default config (data/*/*.parquet) still serves it through load_dataset and the viewer. The catalogue's CC BY-SA 3.0 statement may cover only the foundation's accessibility pictograms, not its publications, and that question is still open. Main is as public as a release tag, so the data comes off main until the scope is answered, as sejm_interpellations' did in #216.

• Removed: data/kultura_bez_barier_pl/ (shard, sample, attribution, decisions, review table, licence-basis evidence, QA, stats, datasheet and NOTICE). It comes back with the answer.
• Kept: the registry entry (release: None), src/build_kultura_bez_barier_pl.py, src/clean_kultura_bez_barier_pl.py, src/licence_basis_kultura_bez_barier_pl.py, their two tests and artifacts/kultura_bez_barier_pl_ontology_manifest.json. The tests skip while the folder is missing (src/conftest.py).
• src/sources.py: the legal note says the data is held off main, and why.
• README.md: main now has the same data as v0.2.6 (27 sources, 3,394,876 documents, 7.04B tokens); "On main, not in v0.2.6" becomes "Held off main", listing the three held sources.
• CHANGELOG.md: the Unreleased section and a "Removed from main" entry under v0.2.6.
• AGENTS.md (known drift), artifacts/source_findings.md (new entry), the src/conftest.py docstring and skip reason.

Older revisions of main still hold the files; removing them from the history is a separate step.

python3 -m pytest src/: 2152 passed, 50 skipped.

🤖 Generated with Claude Code