When using

baml-cli

to serve OpenAPI, is it possible to receive the original LLM response along with the type-parsed one? I'd like to send back the original response in subsequent requests as opposed to the parsed one in order to utilize the LLM cache in case of back-and-forth applications (chats or agents using many tools). If not possible, is it considered in the roadmap?

This is definitely on the roadmap. We are considering how to support this better without making th http response super noisy.

Elixir