Get the evaluation metrics an agent has opted into.
Returns the resolved per-agent metric set with weights applied, split into the LLM-judge bucket and the system-metric bucket. Each bucket's weights sum to 100 when non-empty. To modify, send the full replacement set to `PUT AgentEvaluationMetrics` — there is no partial-patch endpoint.
Authorization
Bearer JWT Authorization header using the Bearer scheme. Enter 'Bearer' [space] and then your token.
In: header
Path Parameters
Agent ID.
The tenant identifier
Response Body
curl -X GET "https://api.fruxon.com/v1/tenants/string/agents/string/evaluationMetrics"[
{
"metricId": "00000000-0000-0000-0000-000000000000",
"key": "string",
"name": "string",
"description": "string",
"weight": 0,
"isDeprecated": false
}
]{
"type": "string",
"title": "string",
"status": 0,
"detail": "string",
"instance": "string",
"property1": null,
"property2": null
}Binds a declared slot to a participant, a roster role, or the operator queue
The body's `bindingType` chooses how the slot is fulfilled (HITL unification design §3.2): `PARTICIPANT` names one person, `ROLE` defers to whoever holds that roster role in this Application — the same selection `consult_for_role` uses, so notify and consult cannot name different humans — and `QUEUE` routes to the awaiting-human queue. Re-binding replaces the binding in place; the binding id is preserved so downstream audit references stay correlatable. A binding says who gets <em>delivered to</em>. It is never a grant to answer or act — that is checked separately, at answer time (design §5.4).
Replace the agent's full opted-in metric set
Full replace, not a patch — anything omitted is removed from the agent. Either bucket may be empty (skipping that class of evaluation entirely), but if non-empty its weights must sum to exactly 100 or the call returns 400. Takes effect on the next eval run; in-flight runs already scheduled are unaffected.