There is a lot of data generation code in the Shiloh libraries that can be put in their own github project and separated from Fluency itself.