This repository contains code to make easier the ensembling of mlflow models. You can make a custom mlflow.Pyfunc meta-model from your already trained models. You just have to implement the way it stacks models scores and the final decision method if the default one is not suitable for your problem.