# EN: python / PT: python
__pycache__/
*.py[cod]
.venv/
dist/
build/
*.egg-info/
# EN: tests/lint / PT: testes/lint
.coverage
htmlcov/
.pytest_cache/
.ruff_cache/
# EN: macOS
.DS_Store
# EN: local pre-publish checklist, not versioned / PT: checklist local pre-publicacao, fora do git
PENDING.md
# EN: NEVER version real data / PT: NUNCA versionar dado real
data/real/
*.real.jsonl
# EN: benchmark: data is regenerated from the seed, results/cache are local / PT: dados regerados pela seed, resultados/cache locais
bench/data/**/*.jsonl
bench/results/
bench/results-*/
bench/cache/
# EN: private notes / PT: notas privadas
docs/decisions.md
# EN: internal review material. The published repo carries promises to the USER only, so the rules about
#   this repository's own test bench, and the record of how the review rounds were run, stay out of it.
#   PT: material interno de revisao. O repo publico carrega promessa ao USUARIO so.
spec/guarantees-internal.yaml
docs/revisao.md
# EN: paper artefacts. Every file in bench/paper/ is GENERATED by bench/report.py from bench/results/,
#   which is itself gitignored. Versioning a figure means versioning a number that can go stale against the
#   data that produced it, which is the drift this project keeps finding in its own documents. Regenerate
#   with `python -m bench.report`. Only *.png was ignored before, so the .pdf, .csv and .tex were versioned.
#   PT: artefato de artigo. Tudo em bench/paper/ e GERADO pelo bench/report.py a partir do bench/results/,
#   que ja esta fora do git. Versionar figura e versionar numero q pode envelhecer contra o dado q o gerou.
bench/paper/
