Overview
Use this endpoint to evaluate a response in the context of a conversation. The system uses a judge model (specified byjudge_id
) to assess the quality or appropriateness of the given response
based on the prior conversation_prefix
.
This is useful for scoring, grading, or analyzing the helpfulness, accuracy, or tone of a response generated by an assistant.
The endpoint is designed to work with a conversation format where the
conversation_prefix
contains prior turns and response
is the assistant’s
reply. The current deployment returns:
score
: numeric grade (higher means more concerning for safety judges).category
: short label for the detected issue (may benull
).rationale
: free-form explanation from the judge model.
Authorizations
Bearer authentication header of the form Bearer <token>
, where <token>
is your auth token.
Body
application/json
Response
Successful Response
The response is of type any
.