# A typical large-data project commits the notebook + data seed + tearsheets,
# but leaves bulky generated artifacts out of version control. Anyone cloning
# re-runs `jellycell run` to reproduce them deterministically.
#
# The tearsheet itself IS committed (it's small markdown) — readers see the
# last-run summary on GitHub without needing the underlying parquet.

artifacts/large_data/sample_*.parquet
artifacts/large_data/features_*.parquet
