With our architecture it will be trivial to add runtime inference and casting. And we will eventually provide a simple API so you don’t have to fiddle with the underlying modules.
But feel free to play around with it, any kind of feedback is appreciated!
