Member Junction
    Preparing search index...

    Interface MLScoringConfiguration

    The shape a MJ: Record Processes.Configuration JSON carries for an 'ML Model' work type — the per-Record-Process scoring inputs the factory reads to construct the MLModelInferenceProcessor. The injected runtime seams (MLInferenceDeps) are supplied at registration time (closed over by the factory), NOT in this config, so the config stays declarative + storable on the Record Process row.

    interface MLScoringConfiguration {
        autoTrain?: MLAutoTrainConfig;
        datedSources?: DatedSourceSpec[];
        modelId?: string;
        primaryKeyField?: string;
    }
    Index

    Properties

    autoTrain? datedSources? modelId? primaryKeyField?

    Properties

    autoTrain?: MLAutoTrainConfig

    Opt-in train-if-untrained (see MLAutoTrainConfig). Off unless declared.

    datedSources?: DatedSourceSpec[]

    Optional dated sources supplying point-in-time ("as-of") features.

    modelId?: string

    The MJ: ML Models id this Record Process scores with. May be omitted only when autoTrain is declared — the first run then trains the pipeline and writes the resulting model id back here.

    primaryKeyField?: string

    Primary-key field on the target entity (defaults to ID).

    Settings

    Member Visibility

    On This Page

    Properties