.gitignore

__pycache__/
*.py[cod]
.venv/
.coverage
.hypothesis/
.mypy_cache/
.pytest_cache/
.ruff_cache/
*.egg-info/
build/
dist/

# Built output. The site is regenerated from data/report.json on every run, so committing it
# would put two copies of the same claim in the repo and let them disagree.
site/

# The IPEDS archives are NOT ignored. data/HD*.zip and data/IC*.zip are 4.1 MB of public federal
# data with no key, no quota and no terms, and committing them is what makes data/national.json
# and data/snapshots/ipeds/*.json checkable by a reader rather than only by their author. Before
# they were committed, every input to the national artifact was ignored, so nobody who cloned this
# repository could regenerate the file that carries every national figure the site publishes.
#
# The full crosscheck payload stays ignored: it is 2.9 MB, it is byte-reproducible from the two
# archives above in under a second, and it is an intermediate rather than a claim.
data/crosscheck.json

# The Scorecard page cache `disclosed fetch --cache-dir` writes. Per-page bodies and their
# provenance, so a rerun needs no network; the committed capture under data/census/ is the
# durable record and carries the same provenance inside it.
.cache/