v2

latestOpenAPI 3.1.02026-08-05267431678.1 KB
inference-history

Submit Inference Feedback

Submit human feedback on a specific inference.

Marks an inference as correct or incorrect. When marking as incorrect, the expected output must be provided for downstream training data curation.

Args: inference_id: The inference record UUID to annotate. request: Feedback payload with verdict, optional corrected output, and notes. auth: Authentication result with user_id.

Returns: InferenceFeedbackResponse confirming the stored feedback.

Raises: HTTPException: 404 if inference not found or outside the caller's team.

post/inferences/{inference_id}/feedback

Path parameters

inference_idstring required

Request body

verdict'correct' | 'incorrect' required

Human judgment: 'correct' or 'incorrect'

{"stackTrail":"components:schemas:InferenceFeedbackRequest:properties:corrected_output:anyOf","oasType":"schema","type":"unknown","title":"Corrected Output","description":"Expected output for incorrect verdicts (JSON structure matching task output)","nullable":true}
notesstring nullable

Optional reviewer notes

Response

Successful Response

inference_idstring required

Inference ID that was annotated

human_verdictstring required

Stored human verdict

human_feedback_atstring date-time required

When the feedback was submitted