Maybe LLM Automation Report, it can chew up all the tokens in a category (or a large amount of them) and compress.
It would not be recursive/chunked though, so it would be a bit limited.
Anyway, probably worth starting there as a proof of concept.
2 Likes