Create a custom model
POST/v1/models
Create a new custom model for object, category, and attribute detection.
Async semantics. This endpoint returns immediately after queueing the generation job, with the model in a pending status. The model is not usable until a follow-up GET /v1/models/{uuid} reports status=completed. Failed generations report status=failed or status=reverted with a populated error.
Authentication. Requires a valid X-API-Key header.
Idempotency. Not idempotent — repeated calls create independent models. Names are unique per billing account; conflicting names return MODEL_NAME_CONFLICT.
Request
Responses
- 200
- 400
- 422
- 429
- 500
The created model in PENDING status. Generation runs asynchronously in a worker; poll GET /v1/models/{uuid} to observe progress.
MODEL_NAME_CONFLICT, INVALID_MODEL_STATE
INVALID_REQUEST_BODY
RATE_LIMITED
INTERNAL_ERROR