Get Harbor Status

Returns raw Polytomic refresh evidence for datasets written to a Harbor. Each pipeline corresponds to a bulk sync or model sync that writes at least one dataset to the Harbor's backing Connection. Tables populated outside Polytomic are not included, even when they are queryable through a customer-managed backing Connection. The response groups shared pipeline evidence so schedules and configuration are not repeated for every dataset: - Each entry in `pipelines` identifies the producer through `type` and `id`. Separate pipelines targeting the same physical dataset remain separate entries. - `datasets` is keyed by the effective destination dataset name. `last_success_at` is the start time of the most recent execution in which that dataset completed successfully, including a successful dataset within a bulk execution that completes with errors. The start time is a conservative upper bound because source reads and destination writes happen afterward. - `latest_status` preserves the latest Polytomic execution status. Never-run datasets omit this field. - Pipeline-level `schedules` preserves schedule parameters and selectors. A schedule can be manual, event-driven, advanced, selective, or limited to named source schemas, so callers should not reduce the list to one inferred cadence. - Continuous schedules use the scheduler's persisted next firing. If scheduler state is unavailable, `next_run_at` is omitted rather than recalculated with new jitter. - Paused pipelines remain present with `refresh_enabled` set to `false` and no `next_run_at`. Use absolute timestamps and the raw statuses to apply the maximum acceptable staleness for your task. The endpoint does not classify datasets or the Harbor as healthy, stale, or unhealthy.