/v1, and a
workspace-scoped bearer key. This reference defines the beta contract; endpoint availability
is confirmed during onboarding. Download OpenAPI.
Requests
Use anIdempotency-Key on every job-creating request. Retrying the same key and body within
24 hours returns the original job; a changed body returns 409. Keys are scoped by workspace
and endpoint. Inference requests are stateless.
List endpoints accept limit (1–100, default 20) and an opaque cursor.
Continue using next_cursor; its absence marks the last page.
Every response includes request_id. Preserve it when reporting a problem.
Errors
Jobs
Dataset imports, exports, tokenizer fits, adaptation, and simulation return202 with a job.
Poll Get job: queued → running → succeeded | failed.
A successful job includes a typed result; a failed job includes an error.
Poll every two seconds or follow Retry-After. Jobs expire after seven days; durable resources
remain addressable by ID. Download URLs have their own expiry.
Versions and access
Model, tokenizer, and dataset IDs identify immutable versions. A mismatch returns409;
unknown fields and items return 422. Requested capabilities must be enabled.
Beta retention, training scope, deployment location, and rate limits are agreed during onboarding.
Do not send production personal data before that scope is established.