Any tools to turn a codebase into a fine tuning dataset? [D]

Excerpt

I have a few web projects with pretty good UI/UX and I’m wondering if there’s any tool or workflow that can turn an existing codebase into a dataset for fine tuning. For example, given a React/Next.js project with components, pages, styling, etc. or a static html site, I’d like to turn it into something like: instruction/prompt -> code or whatever format actually makes sense for training an instruct/thinking/diffusion coding model. Also curious how people handle things like: keeping the context