Save a model and register a version
Save a completed training job as a library model (same as POST /v1/models with a training_job source). With model set, the saved model is also registered as the next version of that lineage (created on first use; its first version becomes production). Promote it with POST /v1/models//aliases/.
Authorizations
API key with the vk_ prefix. Create one from the dashboard.
Body
Body for POST /v1/models/register: save a completed job as a library model and (model versions Task 1) optionally register it into a lineage.
Library display name (slug).
Lineage name to register the saved model into (created on first use).
The saved model becomes the lineage's next version; the response
carries it under version.
"full" (default) or "adapter" (LoRA jobs; required for fleet flips).
Dataset snapshot (Task 2) or plain dataset id recorded on the version.
Free-form tags recorded on the version.
Response
The existing model (idempotent repeat save)
The base checkpoint this model derives from. For artifact_type 'adapter' this names the model the adapter must be served on top of.
'full' (standalone checkpoint) | 'adapter' (LoRA adapter; only attachable to a multi_adapter deployment).
Present when the save also registered the model into a version
lineage (POST /v1/models/register with model): the version it
became (e.g. acme-bot@v6).

