v0.2.0 — API v1 namespace consolidation

The public API under /open/v1/* has been reorganised into clearer
namespaces. This version is the first stable shape we plan to keep for
external integrations — if you wrote code against earlier paths,
you'll need to update endpoints once; after this release the
namespaces are committed.

Breaking changes. All legacy ad / brand / image / video paths
have been removed and replaced. See Removed for the
rename table.

Added

Ad Library — global ad catalog

The /ads* endpoints moved under a single /ad-library namespace so
that "the global library" is a first-class noun, distinct from an
ad-account's own ads.

• GET /open/v1/ad-library/search — full-text + filter search across
  the global ad library.
• GET /open/v1/ad-library/saved — ads the calling workspace has saved.
• GET /open/v1/ad-library/{ad_id} — single ad detail.

Brand Library — global brand catalog

Mirror of Ad Library, for brands. Replaces /brands*.

• GET /open/v1/brand-library/search — search the global brand
  catalog by keyword.
• GET /open/v1/brand-library/followed — brands the workspace tracks.
• GET /open/v1/brand-library/{brand_id} — single brand detail.
• GET /open/v1/brand-library/{brand_id}/ads — ads belonging to a
  catalog brand.

Owned Brands — workspace-owned brand profiles

Promoted from /my-brands so the distinction from catalog brands is
visible in the URL. These are the brand profiles you use as brand_id
when calling the image-generation endpoints.

• GET /open/v1/owned-brands — list the workspace's own brand
  profiles.

Image Generations — image-gen lifecycle as a resource

The previous trio (POST /images/generate, GET /images/jobs/{id},
GET /images/{id}) is now a single image-generations resource with
nested generated-image lookups.

• POST /open/v1/image-generations — submit a creative brief; returns
  generation_id.
• GET /open/v1/image-generations/{generation_id} — poll status;
  image_ids populates once individual images are scheduled.
• GET /open/v1/image-generations/{generation_id}/images/{image_id}
  — fetch a single generated image's detail and URL.

Response field renames are part of the shape change:
project_id → generation_id, image_task_ids → image_ids,
concept_task_ids → concept_ids.

Video transcripts — POST + GET on the same path

Cached video transcripts now live under the owning ad-account, with
POST as the trigger and GET as the read.

• POST /open/v1/ad-accounts/{account_id}/videos/{video_id}/transcript
  — trigger transcription (synchronous; consumes credits on cache
  miss).
• GET /open/v1/ad-accounts/{account_id}/videos/{video_id}/transcript
  — cache-only lookup; never triggers, never charges credits. Returns
  404 when no transcript is cached yet.

Pick the verb by intent — same path, different cost.

Changed

Unchanged namespaces

These endpoints were already well-named and stayed put across the
0.2.0 rename pass — no client changes required:

• /open/v1/boards* — board list / detail / contained ads
• /open/v1/ad-accounts* — list / summary / per-ad performance

Removed

Legacy paths removed in 0.2.0. Each row tells you what to call instead.

Migration notes

1. Search-and-replace the path prefixes from the Removed
   table.
2. If you parse image-generation responses, rename:
   • project_id → generation_id
   • image_task_ids → image_ids
3. The single-image fetch is now keyed on (generation_id, image_id)
   instead of a flat image_task_id — update any client that stored
   only the flat id; it's still present as the id field on each
   generated-image response.
4. For video transcripts: pass account_id as a path parameter (not a
   query parameter or body field) and pick POST vs GET deliberately —
   POST charges on cache miss, GET never does.