Function BuildDiscoveryEvalTestRecord

Parameters

• testCase: LabelledDiscoveryRequest

  The labelled request.

• cell: {
      baseline?: "semantic-search";
      failover?: boolean;
      inputPricePer1M?: number;
      label: string;
      modelId?: string;
      outputPricePer1M?: number;
      promptName?: string;
      seed?: number;
      temperature?: number;
      vendorId?: string;
  }

  The cell.

  • Optionalbaseline?: "semantic-search"

    Runs this baseline instead of the decision. The model settings then do not apply.

  • Optionalfailover?: boolean

  • OptionalinputPricePer1M?: number

    USD per million input tokens, for the cost estimate only.

  • label: string

    Names the cell. It becomes the bracketed part of every record's name.

  • OptionalmodelId?: string

  • OptionaloutputPricePer1M?: number

    USD per million output tokens, for the cost estimate only.

  • OptionalpromptName?: string

  • Optionalseed?: number

  • Optionaltemperature?: number

  • OptionalvendorId?: string

• reps: number

  Repeats per decision record.

• labelSource: string

  Where the label came from.