The diagrams produced by model_to_graphviz() are not actually Kruschke diagrams. They are usually called Bayesian networks or graphical models. According to ChatGPT, they have their roots in a book by Judea Pearl called Probabilistic Reasoning in Intelligent Systems. The WinBUGs/OpenBugs sampling program that is the great grandfather of PyMC had a tool allowing you to draw them and could even do its sampling directly from the diagram. They were very important for those early samplers because they used Gibbs sampling which broke down the problem by conditional variances and so did a lot of symbolic work on the diagrams or the programmatic representation of them that it created from modeling code.
Kruschke introduced what are now known as Kruschke diagrams in his book Doing Bayesian Data Analysis. Actually, I’m not sure about that. He may have published them in an earlier journal article. You can get a short introduction here. They are graphical models or Bayesian networks made pretty and it would be very nice if PyMC could generate them.
Opher