I've just started testing hamilton for my pipeline moving from dvc. Our pipeline processes raw files (pdfs) and generates some intermediate results (xml, json) that get saved and used downstream. Currently, the functions in the pipeline take in just the file paths. I'm curious what the recommended way to use hamilton in this context where I would like the file content to be cached

As long as whatever you want to cache is hashable, there shouldn't be a problem to use

with_cache()

on your builder