Command-line tools

pyLACE ships eleven console_scripts entry points covering the full calibrate → detect → clean → audit → review pipeline.

CLI	What it does
`pylace-annotate`	Annotate a reference frame; emit the arena sidecar JSON.
`pylace-tune`	Live preview to tune background + detector params.
`pylace-roi`	Standalone ROI editor for an existing sidecar.
`pylace-detect`	Run the detection pipeline; emit per-detection CSV.
`pylace-clean`	Detections CSV → cleaned per-frame trajectory CSV.
`pylace-fingerprint`	Build per-track appearance medians from confident frames.
`pylace-audit`	Cost + appearance + verdict-aware identity-swap audit.
`pylace-candidates`	Audit-independent candidate detector (contact, jump, NaN).
`pylace-explore`	Interactive inspector; `--review` opens the review dock.
`pylace-metrics`	Summarise an audited trajectory.
`pylace-inspect`	Lightweight viewer for raw detections.

Three-pass review workflow

# 1. find candidate frames where identity is at risk
pylace-candidates  video.pylace_trajectory.csv

# 2. run the cost-based audit
pylace-audit       video.pylace_trajectory.csv

# 3. open the GUI with the review dock
pylace-explore     video.mp4 video.pylace_audited.csv --review

# 4. re-run the audit; the verdicts sidecar is auto-detected
pylace-audit       video.pylace_trajectory.csv

Verdict keys in the GUI: a accept-swap (uses the spinbox permutation — pair-swap or 3-way cycle), r reject-swap, m mount (rows tagged event_type=mount in the audited CSV and dropped by downstream readers unless --include-mount), u unknown.

Per-CLI `--help`

Each entry point honours --help; the smoke-test suite asserts that main(["--help"]) exits with status 0 for every entry point.