`aip_sdk.PublishedTestPlan`

aip_sdk.PublishedTestPlan(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])

Identity, provenance and the stored methodology — what every published version carries.

Attributes

id str: Server-assigned identifier for this version.
workspace_id str: Workspace the plan belongs to.
name str: Plan name, unique per workspace across versions.
version int: Assigned at publish — 1 for a new name, otherwise one past the previous version. A published version is never modified; an edit publishes the next one.
plan TestPlan: The stored methodology. Its version field is reconciled to match version above rather than whatever was set on the plan passed to publish_test_plan — the platform ignores that value rather than rejecting it, so echoing it back verbatim would make plan.version misleading for every version after the first.

Each metric's spec also comes back carrying published_version — the version of the metric implementation, not of the plan — and published_required_columns: the platform's record of what that metric resolved to when this version was published, which nobody authored. They are assigned exactly like version — ignored if you set them, recomputed on every publish — so a plan you fetched can be changed and republished as-is. Their value is what the platform can say later: once a metric is redeployed with a different required column, the difference between this record and the live required_columns on a TestPlanDetail is reportable as the metric having moved rather than as your dataset being wrong.

user_id str: Id of the user who published this version. Each version carries its own — an edit records whoever made it, and does not inherit the original publisher — so a plan's version history reads as who changed the methodology and when. Always set: the platform refuses a publish it cannot attribute.
created_at datetime | None: When this version was published.
updated_at datetime | None: Equal to created_at; a published version's row never changes.
applicable_quality_checks list[str]: The data-quality checks that will run against a dataset of this plan's schema_name — every check the workspace can see that accepts that schema, named the way a dataset reports its own check results. A plan does not author this and cannot narrow it: which checks run follows from the schema and the workspace's catalogue, so registering or retiring one changes this list on the next read of the same immutable version, with no republish. It is what the promotion path will dispatch, so you can see it before curating. Empty means no visible check accepts the schema.

This is not the plan's required_quality_checks, which names the subset of this that must have produced a verdict before a dataset version can be promoted. This list is what runs; that one is what is required. Publishing refuses a required check outside this set, so a plan can require less than this, never more.