.gitignore

/target
*.gdb
!examples/*.gdb
# Vector-attribute sidecars (<db>.vec.<attr>). Generated by `vec_build`
# from a .gdb plus an embedding file; large, and only valid for the exact
# database they were built from (see the fingerprint in
# src/vector/sidecar.rs). Always regenerate rather than commit.
*.vec.*
__pycache__/
/bench/data/
# Bench artifacts (timestamped per run; regenerate via the bench
# scripts under bench/scripts/). Never useful to commit.
/bench/results/
# Cross-system bench artifacts: timestamped run outputs and external-
# system loaded data (each system imports LDBC CSVs into its native
# format under bench/data/cross-system/<system>/).
/bench/cross-system/results/
/bench/cross-system/.venv/
/bench/data/cross-system/
# Scaling sweep outputs (bench/scaling/run_scaling.sh), timestamped per run.
/bench/scaling/results/
# Internal bench (typecheck_bench) outputs: full.csv/txt + smoke.csv/txt
# captured per local run. Regeneration is fast; not useful in git.
/bench/internal-results/
/python/.venv/
/python/foo.py
/python/pypi_token
/python/.mypy_cache/
/python/dist/
*.swp
# Large generated databases (re-create with scripts/convert_dev_datasets.py)
examples/neoflix.gdb
examples/olympics.gdb
examples/address.gdb
examples/recommendations.gdb
examples/twitter.gdb
examples/video_games.gdb
examples/bluesky.gdb
examples/buzzoverflow.gdb
examples/books.gdb
examples/ldbc-sf01.gdb
# LDBC SNB benchmark inputs (~100 MB at SF=0.1; rebuild with `gqlite ... --import-ldbc-csv`)
/bench/social_network-*
latex
# Editor / OS / MCP junk
.DS_Store
**/.DS_Store
/.playwright-mcp/
/asd