Member Junction
    Preparing search index...

    The compact, payload-safe outcome the builder writes back so the agent can narrate the result.

    interface PredictiveStudioBuildOutcome {
        errorMessage: string;
        headline?: string;
        heldReason: string;
        modelId?: string;
        pipelineId?: string;
        published: boolean;
        readyToPublish: boolean;
        success: boolean;
        trustTitle?: OutcomeTitleText;
        warnings?: FeatureStepWarning[];
    }
    Index

    Properties

    errorMessage: string

    A clean error message when the build failed; else null.

    headline?: string

    The trust verdict's headline ("Beats the flat guess by X [lo, hi] …"); absent when the build trained no model.

    heldReason: string

    When not published, the plain reason (the stored gate decision's reason, leakage, a calibration hold, kept in Draft, or the promotion gate's refusal); else null.

    modelId?: string
    pipelineId?: string
    published: boolean
    readyToPublish: boolean

    True when the model stays in Draft only because this build does not publish: it clears the bar, so the user may ask for it to be published (Promote ML Model, Draft → Validated → Published).

    success: boolean
    trustTitle?: OutcomeTitleText

    The trust verdict's title (TrustVerdicts.OutcomeTitle); absent when the build trained no model.

    warnings?: FeatureStepWarning[]

    Structured warnings emitted during plan translation or training (e.g. dropped candidate features).