Member Junction
    Preparing search index...

    Interface MLAutoTrainConfig

    Opt-in train-if-untrained declaration carried in an 'ML Model' Record Process's Configuration ({ "modelId": "...", "autoTrain": { ... } }).

    Why it exists: a trained model artifact is per-host (its bytes live on the server that trained it) and training always creates a NEW immutable model row. So an app that ships a scoring process can't ship a usable model — on a fresh install the model has to be trained there. Declaring autoTrain lets the scoring run itself do that, once, with no hand steps. Processes that don't declare it are unaffected.

    interface MLAutoTrainConfig {
        binding?: { mode?: "OnDemand" | "Scheduled"; targetColumn?: string };
        pipelineId: string;
    }
    Index

    Properties

    binding? pipelineId

    Properties

    binding?: { mode?: "OnDemand" | "Scheduled"; targetColumn?: string }

    Optional scoring binding to ensure for the trained model when the process has none yet (bindings that already point at this process are always repointed). Lets an app declare where scores land without shipping a binding row that would have to reference a model id that doesn't exist on a fresh install.

    Type Declaration

    • Optionalmode?: "OnDemand" | "Scheduled"

      Binding mode; defaults to Scheduled when the process is scheduled, else OnDemand.

    • OptionaltargetColumn?: string

      The target-entity column the process writes the score to (shown by the prediction panel).

    pipelineId: string

    The MJ: ML Training Pipelines id to train when the configured model needs training.

    Settings

    Member Visibility

    On This Page

    Properties