`aip_sdk.TestPlanDetail`

aip_sdk.TestPlanDetail(id: str, workspace_id: str, name: str, version: int, plan: TestPlan, user_id: str, created_at: datetime | None, updated_at: datetime | None, applicable_quality_checks: list[str], required_columns: list[str], unresolved_metrics: list[str], metric_drift: list[MetricDrift], unresolved_metric_configs: list[str])

One version read back, resolved against the workspace's catalogue as it is now — what get_test_plan and get_test_plan_by_id return.

Everything below is derived at the moment of the read, never stored, so two reads of one immutable version can legitimately differ. That is the point: each field answers "does this methodology still work today?", which only a live catalogue can answer.

Attributes

required_columns list[str]: Every column a dataset needs for the plan's metrics to score now. A metric redeployed with a new required column lengthens this list; the field says what to curate, and a stale list would understate it. An empty list means the plan genuinely needs no columns.
unresolved_metrics list[str]: Plan metrics that no longer resolve to an op in the workspace, e.g. because one was deregistered after publish. They contribute nothing to required_columns and are named here instead of being silently dropped, so a shorter column list is never mistaken for a smaller methodology. Empty means every metric resolved.
metric_drift list[MetricDrift]: One MetricDrift per metric whose recorded publish-time version and required columns no longer match what it resolves to now. Advisory: nothing is blocked, and a run's own metric check remains the gate. What it gives you is the reason a dataset that used to satisfy this plan no longer does. A metric that resolves to nothing at all is in unresolved_metrics instead, since there is nothing left to compare it against, and a plan published before the platform recorded the version is not reported as drifted against a record it never had. Empty means nothing moved.
unresolved_metric_configs list[str]: Plan metrics whose config_ref no longer resolves to a metric config this metric can use — either nothing live resolves, or what resolves now configures a different metric. The pin was usable when the plan was published — publishing an unusable one fails — so an entry here means the config was retired or repointed at another metric afterwards, and every run created from this plan will fail when it dispatches. Empty means every pin still resolves.