Evaluation API

The Evaluation API executes project functions and exposes their outputs. It is the interface to the Diafunc computation layer and covers ad-hoc evaluations, scheduled runs through triggers, historical evaluation state, as well as the configuration and selection of the underlying computation backends.

All Evaluation API endpoints are served under /api/v1/evaluation.

Read operations such as inspecting evaluation state, outputs, and logs require the evaluation:read scope; starting a function or stopping an evaluation requires evaluation:execute.

See Authentication for the shared token and error model.

The Evaluation API is organized around the following resource groups:

Evaluations — list and inspect running and historical evaluations for a given function or project; fetch the most recent evaluation and its source code snapshot.

Evaluation calls — invoke a function by id, optionally with input arguments, and correlate the call with the resulting evaluation.

Outputs and logs — retrieve the structured outputs emitted during an evaluation and the associated execution log.

Triggers — create, update, and remove triggers that execute a function on a schedule or in response to platform events; enable and disable triggers without deleting them.

Stop — interrupt a running evaluation by id.

Running a function from an external pipeline — call the function with the input payload, then poll the returned evaluation for completion and fetch its structured output.

Operating scheduled jobs — define a trigger that runs a given function on a schedule, observe triggered evaluations through the evaluations endpoint, and disable or delete the trigger when the job is decommissioned.

Debugging a long-running evaluation — fetch the evaluation log and emitted outputs, and stop the evaluation if it misbehaves.