generated: '2026-08-14'
method: searched
source: >-
  https://vidlab7-d7584a5d.mintlify.app/api-reference/introduction plus
  openapi/moonscale-openapi-original.json and live unauthenticated probes of
  https://api-prd.moonscale.com
authentication:
  style: api-key-header
  header: x-api-key
  see: authentication/moonscale-authentication.yml
base_url:
  production: https://api-prd.moonscale.com
  legacy_alias: https://api-prd.vidlab7.com
  note: >-
    Both hosts are live and answer with the same JSON gateway envelope. api-prd.vidlab7.com
    is the pre-rename host (Moonscale was VidLab7 until the 2026 rebrand) and is still the
    host hard-coded in the third-party Activepieces connector; the current documentation
    uses api-prd.moonscale.com throughout. The published OpenAPI's servers[] block names
    neither — it carries the Mintlify template default http://sandbox.mintlify.com, so the
    spec alone cannot tell a client where to send a request.
versioning:
  scheme: uri-path
  current: v1
  note: >-
    The live-avatar surface is versioned at /api/v1/. The video-generation surface is NOT
    versioned — it sits at /api/studio-avatar/ with no version segment, so the two halves
    of the same API follow different versioning conventions.
  see: lifecycle/moonscale-lifecycle.yml
idempotency:
  supported: false
  note: >-
    No idempotency key header or parameter is documented or present in the OpenAPI. The
    closest thing is VideoGenerationRequest.correlationId — a REQUIRED caller-supplied
    external reference that is echoed on the response and used to fetch the result later
    via GET /api/studio-avatar/generated-video/{correlationId}. It is a correlation handle
    for retrieval, not a de-duplication key: nothing in the docs states that replaying a
    POST with the same correlationId returns the original result instead of starting a
    second generation. Recorded as absent rather than inferred, so NO Idempotency pointer
    is emitted.
pagination:
  supported: false
  style: none
  note: >-
    No cursor, offset, page or limit parameter appears anywhere in the spec or the docs.
    The list-shaped operation (GET /api/v1/conversations) is bounded by a required time
    window instead — dateStart and dateEnd Unix timestamps whose range must not exceed 24
    hours — and returns the full conversations[] array for that window with no continuation
    token. A caller with more than one window of history must page by walking time.
filtering:
  - {operation: 'GET /api/v1/conversations', param: dateStart, type: integer, required: true, note: Unix timestamp, seconds since epoch}
  - {operation: 'GET /api/v1/conversations', param: dateEnd, type: integer, required: true, note: must be after dateStart and within 24 hours}
  - {operation: 'GET /api/v1/conversations', param: liveAvatarId, type: string, required: false, note: optional UUID filter to a single avatar}
field_expansion:
  supported: false
metadata:
  supported: true
  note: >-
    VideoGenerationRequest.metadata is a free-form object for custom caller metadata on a
    video. Live-avatar conversations additionally echo utm_id and utm_source captured from
    the conversation URL, which is how a caller ties a conversation back to its own contact
    record.
request_tracing:
  supported: true
  header: x-request-id
  direction: response
  note: >-
    Observed live on every unauthenticated response from api-prd.moonscale.com, including
    401 and 403 (for example x-request-id 3805d2a1-1018-492c-92fa-788be44cd5e1). The header
    is not documented, so a caller has no published guarantee about it; it is recorded here
    because it was directly observed and is the only correlation handle available for support.
error_envelope:
  format: proprietary
  shape: '{"message": "<human-readable string>"}'
  rfc9457: false
  content_type: application/json
  note: >-
    Not RFC 9457 problem+json. A flat single-key object with no type, title, status,
    detail, instance or machine-readable code, so an agent cannot branch on error class —
    only on HTTP status plus prose matching.
  see: errors/moonscale-problem-types.yml
rate_limit_signaling:
  documented: false
  headers_observed: []
  note: >-
    No RateLimit-*, X-RateLimit-* or Retry-After header was returned on any observed
    response, and no limits are published. See rate-limits/moonscale-rate-limits.yml.
async_pattern:
  style: submit-then-poll-or-webhook
  note: >-
    Video generation is asynchronous. POST /api/studio-avatar/generate-video returns 201
    with a status of PENDING; the caller then either polls GET
    /api/studio-avatar/generated-video/{correlationId} until status is COMPLETED or FAILED,
    or supplies webhookUrl on the request and is called back with the same payload. The
    documented status enum is PENDING, PROCESSING, COMPLETED, FAILED. The third-party
    Activepieces connector polls this at a 5 second interval and gives up after 300 seconds,
    which is the only published indication of expected generation time.
  see: asyncapi/moonscale-webhooks.yml
spec_docs_divergence:
  note: >-
    The published OpenAPI and the published reference do not describe the same API. This is
    a real interoperability hazard for any client generated from the spec, so it is recorded
    rather than smoothed over.
  divergences:
  - The reference documents GET /api/v1/conversations (conversations by date range) and GET /api/v1/avatars (in the API-key example request); neither path exists in the OpenAPI.
  - The OpenAPI keys a single conversation by {conversationShortCode}; the reference keys the same operation by {conversationId} (UUID).
  - The OpenAPI's ApiLiveAvatarCreateRoomRequest requires name and liveAvatarShortCode; the reference documents liveAvatarId, conversationShortCode, contactEmail, privacy, enableChat, roomSettings and context.
  - The OpenAPI's create-room response returns conversationShortCode; the reference returns conversationId.
  - The OpenAPI declares no operationIds on any of its four operations, so generated clients get synthesized method names.
  - The OpenAPI's servers[] is http://sandbox.mintlify.com, a Mintlify template default over plain HTTP, not a Moonscale host.