The cross-validation meta node can be easily used for building cross-validated models of many sort. But how can one do external predictions on a cross-validated model? This requires breaking up of the cross-validation loop. Is the final model still stored in the predictor node?
Is there a specific reason that you would want to do that ?
Cross-validation is just a validation technique which helps you obtain an estimate of the performance variance of your model - as opposed to "only" having a point estimate of the performance indicator in case of a simple hold-out strategy. Therefore, as such a cross-validated model will never be deployed nor used for making actual predictions. After having properly assessed the model performance with or without cross-validation, you should re-train the model on the whole training set (or a subset if such is deemed necessary) and consider it the production model.