.gitignore

# Python
__pycache__/
*.py[cod]
*.egg-info/
.venv*/
venv/

# Data & generated outputs (reproducible via scripts)
/data/
synth/out/
bem_out/
*_results.json

# Ultralytics base checkpoints auto-download into the CWD on first use
# (e.g. train.py --pretrained yolo11n.pt). Weights are never committed;
# real runs live under ~/workspace/datasets/detector_runs/ (see detector/QUICKSTART.md).
/yolo11*.pt
/detector/yolo11*.pt
# Weights are never committed, whatever the model name or directory.
*.pt
*.onnx
*.pth

# ML model artifacts (reproducible via review_classifier/train.py)
# -- triage.py loads these; npz is required, pkl raises ValueError --
/review_classifier/trained_model.pkl
/review_classifier/trained_model_extraction_type.pkl
/review_classifier/trained_model_needs_human.pkl
/review_classifier/trained_model_urgency.pkl
/review_classifier/trained_model_resolution.pkl
# -- laya_experiment.py loads these; npz is required, pkl raises ValueError --
/review_classifier/trained_model_route_to_review.pkl
/review_classifier/trained_model_urgency.pkl
/review_classifier/trained_model_schedule_match.pkl

# Agent runtimes & MCP server state
.agents/
.foreman/

# Planning docs (personal/dynamic — not project source)

# OS
.DS_Store

# wheel builds (scripts/check_release_licenses.py)
build/
dist/

# gh CLI state and tool caches (PR A of the #1093 top-level clean)
.local/
.hypothesis/
.pytest_cache/
.ruff_cache/

# Runtime feedback log (PR C of the #1093 top-level clean):
# regenerates via `matchline feedback --log` / demo runs; never commit working logs
/feedback.jsonl
.coverage