damast.integrations.tracking

Backend-agnostic interface for reporting pipeline runs to an experiment tracker (e.g. MLflow, W&B): the “metadata contract” (units, value ranges, computed statistics) alongside per-step timing, for audit trails of scientific runs.

Functions

flatten_metadata(metadata)	Flatten a MetaData contract into tracker-agnostic params/metrics/tags.
flatten_step_params(processing_stats)	Flatten the non-numeric part of DataProcessingPipeline.processing_stats into parameters.
flatten_step_stats(processing_stats)	Flatten the numeric part of DataProcessingPipeline.processing_stats into metrics.

Classes

ExperimentTracker()

Minimal interface a pipeline run can be reported to.