generated: '2026-08-01'
method: searched
source: openapi/leia-immersity-cloud-api-openapi.yml
docs: https://docs-api.immersity.ai/docs/getting-started
api: Immersity Cloud API
authentication:
  style: OAuth 2.0 client-credentials bearer JWT
  header: 'Authorization: Bearer <access_token>'
  token_endpoint: https://auth.immersity.ai/auth/realms/immersity/protocol/openid-connect/token
  detail: authentication/leia-authentication.yml
transport:
  base_url: https://api.immersity.ai
  alternate_base_url: https://api.leiapix.com
  content_type: application/json
  style: RPC-flavored REST — one POST per transformation verb (/disparity, /animation, /sbs, /topBottom,
    /encode, /decode, /video, /spatial) plus two GET reads (/prices, /get-upload-url)
result_delivery:
  model: presigned-URL write-back
  description: >-
    The caller supplies `resultPresignedUrl`, a cloud-storage URL carrying HTTP PUT permission, and the
    service writes the generated artifact directly to it rather than returning binary in the response body.
    Inputs are likewise passed by URL (`inputImageUrl`, `inputDisparityUrl`, `inputVideoUrl`). The 200
    response echoes `correlationId` and `resultPresignedUrl`.
  provider_helper: GET /api/v1/get-upload-url (StorageController_getStoragePresignedUrl) mints a Leia Storage
    presigned URL valid for 24 hours, returned as {url, urlExpiresUTC}
  supported_storage: AWS S3 presigned URLs, GCP Cloud Storage signed URLs, Azure Blob SAS
request_tracing:
  field: correlationId
  location: request body (and echoed in the 200 response body)
  format: UUID
  required: false
  behavior: >-
    "UUID to correlate this request. This must be unique for each request. If not provided, the service will
    generate one." Used to correlate the request, the callback and the written artifact.
idempotency:
  supported: false
  note: >-
    Immersity publishes no idempotency key and no replay contract. `correlationId` is explicitly documented
    as a per-request unique correlation identifier, not a deduplication key — resubmitting the same
    correlationId is not documented as returning the first result, and every accepted call consumes credits.
    Recorded as absent rather than mapped onto correlationId.
asynchrony:
  synchronous: image operations (disparity, animation, sbs, topBottom, encode, decode) return within the
    request; the documented client timeout is 3-5 minutes
  callback: video operations (/api/v1/video, /api/v1/spatial) require a `callbackUrl` and notify it when
    processing completes or fails
  callback_catalog: asyncapi/leia-callbacks.yml
pagination:
  supported: false
  note: no collection/list endpoints in the published surface — every operation is a single transformation
    or a single read
filtering_and_expansion:
  supported: false
metering:
  model: prepaid credits
  unit: credit
  documented_rate: 10 credits per API call flat rate; 100 free credits granted to a new account
  free_credit_restriction: free credits may not be used for commercial purposes
  price_lookup: GET /api/v1/prices (ProductPricingController_getPrices) returns current disparity and
    animation prices
  pre_authorization: >-
    Video conversions require the caller to send `videoLength` and a pre-calculated `creditAmount`; both are
    verified server-side and a mismatch rejects the request (ERROR_VIDEO_LENGTH_MISMATCH is returned on the
    callback payload).
  insufficient_funds: HTTP 402 on every credit-consuming operation
versioning:
  scheme: uri-path
  current: v1
  spec_version: 3.1.1
  detail: lifecycle/leia-lifecycle.yml
error_envelope:
  format: undocumented
  note: >-
    The published OpenAPI declares 400/401/402 responses with prose descriptions but no response schema and
    no media type, so the error body shape is not specified. Errors are not RFC 9457 problem+json. Video
    conversion failures surface as string error codes (e.g. ERROR_VIDEO_LENGTH_MISMATCH) in the callback
    payload.
  detail: errors/leia-problem-types.yml
rate_limiting:
  documented: false
  note: no published rate-limit policy or rate-limit response headers; throughput is governed by the credit
    balance rather than a request-per-second limit
support:
  channel: https://discord.gg/immersityai
cross_links:
  authentication: authentication/leia-authentication.yml
  scopes: scopes/leia-scopes.yml
  errors: errors/leia-problem-types.yml
  lifecycle: lifecycle/leia-lifecycle.yml
  data_model: data-model/leia-data-model.yml
  callbacks: asyncapi/leia-callbacks.yml
x-evidence:
  fetched: '2026-08-01'
  urls:
  - https://docs-api.immersity.ai/docs/getting-started
  - https://docs-api.immersity.ai/llms.txt
  http_status: 200