Member Junction
    Preparing search index...

    Interface MJRubricEvaluationScoreEntity_IRubricEvaluatorSelection

    AIAgentRubric.EvaluatorConfig — which evaluator scores the link, and with which prompts.

    For the LLM evaluator three prompts compose one call: the evaluator prompt (the parent, which owns the JSON reply contract), the judge prompt rendered into its judgePrompt slot, and the criterion prompt that renders each criterion. PromptID/PromptName name the judge; for the Decision evaluator they name the decision prompt instead.

    interface MJRubricEvaluationScoreEntity_IRubricEvaluatorSelection {
        AgentID?: string;
        CriterionPromptID?: string;
        CriterionPromptName?: string;
        EvaluatorName?: string;
        EvaluatorType:
            | "Agent"
            | "Human"
            | "AIPrompt"
            | "External"
            | "Deterministic"
            | "Self";
        Extensions?: Record<string, MJRubricEvaluationScoreEntity_JsonValue>;
        Mode?: "SinglePass" | "PerCriterion";
        ModelID?: string;
        ModelSelection?: "System" | "Judge";
        PromptID?: string;
        PromptName?: string;
        Samples?: number;
        SystemPromptID?: string;
        SystemPromptName?: string;
    }
    Index

    Properties

    AgentID? CriterionPromptID? CriterionPromptName? EvaluatorName? EvaluatorType Extensions? Mode? ModelID? ModelSelection? PromptID? PromptName? Samples? SystemPromptID? SystemPromptName?

    Properties

    AgentID?: string
    CriterionPromptID?: string

    LLM and Decision: the prompt that renders one criterion. Default Rubric Criterion.

    CriterionPromptName?: string
    EvaluatorName?: string

    A registered evaluator name (LLM, Decision, Agent, Deterministic, or a custom one). Wins over EvaluatorType.

    EvaluatorType:
        | "Agent"
        | "Human"
        | "AIPrompt"
        | "External"
        | "Deterministic"
        | "Self"
    Extensions?: Record<string, MJRubricEvaluationScoreEntity_JsonValue>

    Settings for a custom evaluator, keyed by its evaluator name.

    Mode?: "SinglePass" | "PerCriterion"
    ModelID?: string

    Pins the model.

    ModelSelection?: "System" | "Judge"

    LLM: which prompt's model bindings choose the model. System (default) or Judge.

    PromptID?: string

    LLM: the judge prompt. Decision: the decision prompt.

    PromptName?: string

    The same prompt by name, when no PromptID is set. For example Rubric Judge - Sage.

    Samples?: number

    LLM: runs the rubric this many times and keeps each criterion's median level.

    SystemPromptID?: string

    LLM: the parent evaluator prompt. Default Rubric Evaluator. Must return the same JSON.

    SystemPromptName?: string

    Settings

    Member Visibility

    On This Page

    Properties