Properties
content
content: object
categories?
optionalcategories?:object[] |null
child_metrics?
optionalchild_metrics?:object[] |null
criteria
criteria: string
description?
optionaldescription?:string
Description
Free-form description of the metric. Defaults to an empty string.enabled_in_prod?
optionalenabled_in_prod?:boolean
Description
Whether this version should run against production traffic. Defaults tofalse for non-HUMAN metrics and true for HUMAN metrics.
Default
filters?
optionalfilters?:object
filters.filterArray
filterArray: object[]
model_name?
optionalmodel_name?:string|null
model_provider?
optionalmodel_provider?:string|null
name
name: string
needs_ground_truth?
optionalneeds_ground_truth?:boolean
Description
Whether this metric requires ground-truth labels to evaluate. Defaults tofalse.
Default
return_type?
optionalreturn_type?:"string"|"boolean"|"float"|"categorical"
Description
Return type of the metric. Defaults tofloat.
Default
sampling_percentage?
optionalsampling_percentage?:number
Description
Percentage of events the metric should run against, 0–100. Defaults to10.
Default
scale?
optionalscale?:number|null
threshold?
optionalthreshold?: {max?:number;min?:number;pass_when?:number|boolean;passing_categories?:string[]; } |null
type
type:"LLM"|"COMPOSITE"|"PYTHON"|"HUMAN"
deploy_immediately?
optionaldeploy_immediately?:boolean
message
message: string
metric_id
metric_id: string