I created a custom step recipe that does feature selection based on causal modeling. Basically, it runs the algorithm which generates a graph (network) and only keeps the variables that are connected to the output variable I'm trying to predict. It works well and I no major issues with it . However, it would be nice to save the entire graph it generates as it runs through fit_resamples() or tune_grid(). I'm currently just writing the graph to a text file which isn't ideal. Any recommendations?