get_run

Scope: read · Signature: get_run(experiment_id: str, view: {summary|stages|pipeline|events|log|notifications|settings|config|explanations|downloads} = summary, since: int | None = None, limit: int | None = None) -> Any

Use to observe a run: is it running, what failed and why, what config ran.

view="summary" (default): status, stage progress, the Run-all chain, explanation status, latest notifications, the error, and the plan's stop_rule with its verdict (stop, continue, or undecidable + reason) once the run finished. Other views: stages, pipeline (the stage graph), events (since=seq cursor; last limit, default 50), log (the console's last limit lines, default 80 — read it when a run failed), notifications, settings (labelled launch settings, and monotone_contract: the monotone directions a supervised fit bound, per feature — increasing/decreasing, derived from the data or pinned by an analyst, with the measured partial Spearman ρ), config (the resolved config the run used), explanations (the Explain step's status and stage graph), downloads (the files the run produced, by name — downloading them is the parleml UI's). For metrics use get_model_evaluation; to compare configs across runs use compare.

Powered by: the experiment runner's stage manifest, event log and console (backend/app/services/experiment_runner.py).