Context
Episodes need reliable provenance to remain useful for downstream QA, training, debugging, and customer review. Keystone stores lineage and generic metadata today, but it does not yet enforce required metadata completeness or expose a clear provenance model for every episode.
Scope
- Define required episode metadata fields such as robot, operator, station, factory, organization, scene, subscene, SOP, task, batch, recorder version, transfer version, recording profile, sensor config, and calibration references.
- Validate metadata completeness when episodes are created or during QA.
- Mark incomplete metadata as warning, needs review, or QA failure according to configurable policy.
- Add provenance fields to episode detail and export APIs.
- Track metadata schema version and provenance completeness status.
- Make missing metadata visible to dashboard and QA workflows.
Acceptance Criteria
- Keystone can determine whether an episode has required provenance metadata.
- Missing required metadata is surfaced with stable reason codes and operator-facing messages.
- Episode detail includes provenance completeness status and schema version.
- QA or acceptance workflows can route incomplete metadata to review or failure.
- Tests cover complete metadata, missing required fields, optional fields, and schema version handling.
Context
Episodes need reliable provenance to remain useful for downstream QA, training, debugging, and customer review. Keystone stores lineage and generic metadata today, but it does not yet enforce required metadata completeness or expose a clear provenance model for every episode.
Scope
Acceptance Criteria