fraudtwin.persist_scale_records#

fraudtwin.persist_scale_records(records, run_id, *, dsn=None, batch_size=2_000)[source][source]#

Persist a partition-record stream without whole-run materialization.

Scale rows are kept in a narrow staging table so chunks can be consumed independently by downstream jobs. The established persist_run API remains unchanged for small, typed in-memory runs.

Return type:

PostgresPersistenceResult

Parameters:

  • records (Iterable[Mapping[str, Any]])

  • run_id (str)

  • dsn (str | None)

  • batch_size (int)