video Video

Live-broadcast lifecycle state. Universal across livestream-capable platforms (YouTube, Twitch, TikTok Live, Instagram Live, …). `none` = not a livestream / livestream finished and was archived as VOD; `upcoming` = scheduled but not yet started; `live` = currently broadcasting.

Classification of a synced field's data sensitivity. Drives subject- rights workflows and cache-purge retention decisions.

Path-based cross-reference relative to .corpospec/ root. Pattern: `^[a-z0-9_-]+(/[a-z0-9_.-]+)+$`

Retention policy applied to a synced record. Tiers are named after the discipline they impose; the connector layer maps each upstream surface onto the appropriate tier. - `thirty_day_strict` — atomic records that auto-purge or refresh at the thirty-day boundary. The dominant tier on platforms whose Terms of Service impose a thirty-day cache discipline on derived data (e.g. YouTube Developer Policies §III.E.4 for Data API outputs). - `thirty_day_aggregate_with_lineage` — atomic record purges at thirty days; the aggregate computed from it may be retained when accompanied by explicit `derived_from` + `derivation_method` lineage. - `auth_lifetime` — atomic record retains against active OAuth authorisation; tears down on token revocation or tenant deletion. Used for analytics surfaces whose ToS explicitly exempt them from the thirty-day rule (e.g. YouTube Analytics API per §III.E.4). - `creator_authored` — data the tenant originates rather than fetches; no upstream-platform retention constraint applies. - `metadata_indefinite` — structural identifiers and immutable timestamps the upstream ToS explicitly permits to retain.

Source attribution for a synced field. Records the kind of upstream surface that produced the value; combined with the parent record's `platform`, this fully attributes any value back to its origin. Open-ended by design: new connector kinds (e.g. monetary or content- owner APIs) extend without breaking changes. The platform discriminator (`Channel.platform`) carries the brand identity; this enum carries only the kind of API surface the value came from.

Provenance metadata stamped on every synced atomic record across the creator pillar. Per-record granularity; the `retention_tier` reflects the strictest tier of any field on the record so the cache-purge job can decide conservatively. Field-level retention distinctions are documented in each entity type's struct doc-comments — the schema documents the static taxonomy, the data carries only the per-fetch state.

Resolution band for the primary streamable rendition. `hd` covers 720p and above; `sd` covers everything below.

Video subtype derived from the short-form / long-form / livestream classification heuristic: - `duration_seconds <= 60 && broadcast_state == "none"` → `short` - `livestream_started_at != null && broadcast_state == "none"` → `livestream_vod` - `broadcast_state == "live"` → `live` - `broadcast_state == "upcoming"` → `upcoming` - otherwise → `regular_vod` Where the host platform exposes an authoritative content-type signal (e.g. YouTube Analytics' `creatorContentType`), the connector is free to reconcile the heuristic against it on a periodic cadence.

Content-licence selection. `platform_default` covers any platform's own all-rights-reserved licence; `creative_commons` covers any CC licence the platform exposes. Connectors map native licence values onto the corresponding member.

Video-level privacy status. Public, unlisted, and private are the three states common across most video platforms; values map onto each platform's native enum at the connector boundary.

Aggregate counts attached to a single video. Snapshot semantics: the counts are valid as of `provenance.synced_at`. Comment text and commenter identities are intentionally not stored.

Upload-processing status surfaced by the host platform during and after upload. The five states cover the union surveyed across video platforms; connectors map their native states onto the closest member.