v25

latestOpenAPI 3.1.0raw.githubusercontent.com2025-11-204973358.2 KB
Eval Runs

Update Answer Score

Update the score, confidence, or explanation for a specific response.

Args: eval_run_uuid (str): UUID of the eval run. response_uuid (str): UUID of the response to update. update_request (AnswerUpdateRequest): The update data. workspace_uuid (str, optional): UUID of the workspace. Defaults to None.

Returns: EvalScoredResponse: The updated response.

Raises: AymaraAPIError: If the response is not found or the request is invalid.

Example: PATCH /api/v2/eval-runs/{eval_run_uuid}/responses/{response_uuid} { "is_passed": true, "confidence": 0.95, "explanation": "Updated explanation" }

patch/v2/eval-runs/{eval_run_uuid}/responses/{response_uuid}

Path parameters

eval_run_uuidstring required
response_uuidstring required

Query parameters

workspace_uuidstring

Request body

is_passedboolean nullable

Whether the response passed the evaluation.

confidencenumber nullable

Confidence score for the response.

explanationstring nullable

Explanation for the response.

Response

OK

prompt_uuidstring required

Unique identifier for the prompt.

thread_uuidstring nullable

Unique identifier for the thread, if any.

turn_numberinteger

Turn number in the conversation (default: 1).

continue_threadboolean

Whether to continue the thread after this response.

content_type'text' | 'image' | 'video'

Content type for AI interactions.

exclude_from_scoringboolean

Whether to exclude this response from scoring.

ai_refusedboolean

Whether the AI refused to answer the prompt.

response_uuidstring nullable

Unique identifier for the response, if any.

explanationstring nullable

Explanation for the response, if any.

confidencenumber nullable

Confidence score for the response, if any.

is_passedboolean nullable

Whether the response passed the evaluation, if any.

manually_updatedboolean

Whether this response was manually updated.

manually_updated_atstring date-time nullable

Timestamp when this response was last manually updated.