events
Best-effort live event publishing from inside Kitaru checkpoints.
attributelogger= logging.getLogger(__name__)attributeCHECKPOINT_STARTED_KIND= 'kitaru.checkpoint.started'attributeCHECKPOINT_PROGRESS_KIND= 'kitaru.checkpoint.progress'attributeCHECKPOINT_RETURNED_KIND= 'kitaru.checkpoint.returned'attributeCHECKPOINT_FAILED_KIND= 'kitaru.checkpoint.failed'funcflush(timeout=2.0) -> boolFlush pending live events when ZenML streaming is available.
Flushing does not require checkpoint scope because it only drains the
process-local ZenML publisher queue. It returns False if flushing itself
fails or times out.
paramtimeoutfloat= 2.0Returns
boolfuncpublish(kind, payload=None, *, message=None, correlation_id=None, index=None, flush=False) -> NonePublish a custom live event from inside the current checkpoint.
Publishing is best effort: transport failures are logged and dropped, but
misuse such as calling this outside @checkpoint raises immediately.
paramkindstrparampayloadMapping[str, Any] | None= Noneparammessagestr | None= Noneparamcorrelation_idstr | None= Noneparamindexint | None= Noneparamflushbool= FalseReturns
Nonefuncprogress(message, *, percent=None, correlation_id=None, flush=False, **fields) -> NonePublish a standard checkpoint progress event.
parammessagestrparampercentfloat | None= Noneparamcorrelation_idstr | None= Noneparamflushbool= FalseparamfieldsAny= {}Returns
None