When a model is contained in a sklearn Pipeline, you currently have to extract transformer and WrapClassifier elements and call them separately. It would be useful to have a light wrapper Pipeline class that incorporates the WrapClassifier and WrapRegressor conformal predict_* functionality.