POST
/
clusters
/
{clusterId}
/
runs

A Run can be started with a new message (human) or from an existing template.

Structured output

By default, a Run result will be a JSON object returned by the model. A specific output structure can be enforced by providing a result.schema value in JSON Schema form.

The Run will be forced to return a result in the provided format or null if this is not possible.

Result handler

As an alternative to retrieving a Run result via the API, you can provide an Inferable function as a handler which will be called with the result of the Run.

Testing Runs

Providing a test object will mark run as a test, this is used for running evaluations. Test runs can include mock function results.

Headers

authorization
string
required

Path Parameters

clusterId
string
required

Body

application/json
message
string

The prompt message, do not provide if using a prompt template

name
string

The name of the run, if not provided it will be generated

model
enum<string>

The model identifier for the run

Available options:
claude-3-5-sonnet,
claude-3-5-sonnet:beta,
claude-3-haiku
result
object
attachedFunctions
string[]

An array of attached functions (Keys should be service in the format <SERVICE>_<FUNCTION>)

metadata
object

Run metadata which can be used to filter runs

test
object

When provided, the run will be marked as as a test / evaluation

template
object

A prompt template which the run should be created from

reasoningTraces
boolean
default: true

Enable reasoning traces

callSummarization
boolean
default: true

Enable summarization of oversized call results

Response

201 - application/json
id
string
required

The id of the newly created run