# Python
__pycache__/
*.py[cod]
*.egg-info/
.eggs/
# Anchored: an unanchored "build/" also matches scripts/build/, which holds the
# dataset build pipeline and is tracked.
/build/
dist/
.pytest_cache/
.mypy_cache/
.ipynb_checkpoints/

# Local data. Subsets are built, not tracked: scripts/build_tutorial_subset.py
# reproduces subsets/tutorial from the pinned ids, and the layout itself runs to
# hundreds of MB.
subsets/
sporc_parquet/

# Credentials. Never commit these: the repo is public, and a token pushed even
# once is compromised whether or not the commit is later removed.
.env
.env.*
*.pem
.netrc

# Agent scratch file; see the note in the repo's contributing docs.
CLAUDE.md
