[feat]: track entire raw LLM response for proxy use cases #10
Description
going to hold this pr until we see user need. dont want to add support for something that might take a significant amount of work to support in the long term and has unclear interactions with reask, fix, filter and streaming behavior
Originally posted by @dtam in #7 (comment)