Evals

Create a new eval using an eval type configuration.

Args: eval_request (Eval): Data for the eval to create, including eval type and configuration.

Returns: Eval: The created eval object.

Raises: AymaraAPIError: If the request is invalid.

Example: POST /api/evals { "eval_type": "...", "workspace_uuid": "...", ... }

post/v2/evals

Request

Request body

eval_uuidstring nullable

Unique identifier for the evaluation.

namestring nullable

Name of the evaluation.

created_bystring nullable

Name of the user who created the evaluation.

ai_descriptionstring required

Description of the AI under evaluation.

eval_typestring required

Type of the eval (safety, accuracy, etc.)

eval_instructionsstring nullable

Additional instructions for the eval, if any.

languagestring nullable

Language code for the eval (default: "en").

modality'text' | 'image' | 'video'

Content type for AI interactions.

num_promptsinteger nullable

Number of prompts/questions in the eval (default: 50).

is_jailbreakboolean

Indicates if the eval is a jailbreak test.

is_sandboxboolean

Indicates if the eval results are sandboxed.

workspace_uuidstring nullable

UUID of the associated workspace, if any.

status'created' | 'processing' | 'finished' | 'failed'

Resource status.

created_atstring date-time nullable

Timestamp when the eval was created.

updated_atstring date-time nullable

Timestamp when the eval was last updated.

Response

Created

eval_uuidstring nullable

Unique identifier for the evaluation.

namestring nullable

Name of the evaluation.

created_bystring nullable

Name of the user who created the evaluation.

ai_descriptionstring required

Description of the AI under evaluation.

eval_typestring required

Type of the eval (safety, accuracy, etc.)

eval_instructionsstring nullable

Additional instructions for the eval, if any.

languagestring nullable

Language code for the eval (default: "en").

modality'text' | 'image' | 'video'

Content type for AI interactions.

num_promptsinteger nullable

Number of prompts/questions in the eval (default: 50).

is_jailbreakboolean

Indicates if the eval is a jailbreak test.

is_sandboxboolean

Indicates if the eval results are sandboxed.

workspace_uuidstring nullable

UUID of the associated workspace, if any.

status'created' | 'processing' | 'finished' | 'failed'

Resource status.

created_atstring date-time nullable

Timestamp when the eval was created.

updated_atstring date-time nullable

Timestamp when the eval was last updated.

Changes

    • breaking

      removed the required property ////// from the response with the status

    • breaking

      removed the required property //////// from the response with the status

    • warning

      removed the request property //////

    • warning

      removed the request property ////////

    • breaking

      removed subschema #1 subschema #4 from the request property anyOf list

    • info

      added subschema #1 subschema #4 to the request property anyOf list

    • info

      added subschema #1 subschema #4 to the response property anyOf list for the response status

    • info

      removed subschema #1 subschema #4 from the response property anyOf list for the response status