Currently, there are two options to get the decision tree representations: export_graphviz and export_text.

I would like to add export_dict, which will output the decision as a nested dictionary.

While you can convert the graphviz representation using cli tools, it's a bit unruly and is a weird workflow. When converting back, the nesting is also kind of weird, with /n chars floating around.

The main utility for json/dict representations are

1. That they're far easier to work with for any front-end work.
2. Working with a dict allows you to use the representation easily for any downstream tasks using the decision tree (that aren't raw inference).