Models and nodes
A model is an available AI capability. ItsnodeType identifies the model family, while subModelId selects a specific variant when the family offers more than one.
Use the catalog and schema endpoints before rendering an input form or submitting a run:
GET /v1/modelslists available models.GET /v1/models/searchfilters the catalog.GET /v1/models/{modelId}/schemareturns required fields, defaults, options, and limits.GET /v1/models/{modelId}/pricingreturns pricing details.
Standalone model runs
A run is one asynchronous model execution. Start it withPOST /v1/nodes/{nodeType}/run, save the returned runId, and poll GET /v1/nodes/runs/{runId}.
States
RUNNING moves to a terminal state such as COMPLETED, FAILED, or
CANCELED.Result
Completed runs return model output plus status, error, and credit-usage
details.
202 Accepted as confirmation that the job was queued, not as final output. Poll with a bounded timeout and handle every terminal state explicitly.
Credits
Each run debits credits from the authenticated user or active organization. Costs depend on the model and submitted parameters.GET /v1/credits/balancereturns the current balance.GET /v1/models/catalogincludes catalog cost metadata.POST /v1/nodes/estimate-creditsestimates a standalone model payload.
Webhooks
Execution surfaces that support asynchronous callbacks sign webhook deliveries with HMAC-SHA256 and retry failures with exponential backoff. Validate signatures and make receivers idempotent. See Webhooks for the complete contract.How it fits together
1
Discover a model
List or search the catalog and choose a
nodeType and, when applicable, a
subModelId.2
Inspect schema and pricing
Validate required fields and show an estimate before accepting expensive
input.
3
Start and poll
Start a standalone run, persist its
runId, and poll until it reaches a
terminal state or your bounded timeout expires.Quickstart
Walk through the REST API end-to-end with code samples.
Choose an endpoint
Map an integration task to the supported REST route.