SDK: models

The models area has 4 methods on the client. Each takes one object of path, query and body arguments and an optional CallOptions, and returns a promise of the response body. A refused call rejects with OverlaneApiError (see the SDK index).

modelsComplete

One model call inside a run, held to the run’s token and dollar caps

• Request: POST on /v1/tenants/{tenantId}/models/{modelId}/complete

modelsComplete(args: { tenantId: string; modelId: string; body: ModelCompleteRequest }, options?: CallOptions): Promise<ModelCompletion>

modelsCreate

Register a model executor; the key is write-only and stored encrypted (secrets class)

• Request: POST on /v1/tenants/{tenantId}/models
• Needs an approval for agents: class secrets

modelsCreate(args: { tenantId: string; body: CreateModelExecutorRequest }, options?: CallOptions): Promise<Record<string, unknown>>

modelsList

Model executors by alias, kind, model, caps and key state; a key is never returned

• Request: GET on /v1/tenants/{tenantId}/models

modelsList(args: { tenantId: string }, options?: CallOptions): Promise<ModelExecutorList>

modelsRemove

Remove a model executor and destroy its stored key (secrets class)

• Request: DELETE on /v1/tenants/{tenantId}/models/{modelId}
• Needs an approval for agents: class secrets

modelsRemove(args: { tenantId: string; modelId: string }, options?: CallOptions): Promise<Record<string, unknown>>