Custom model options & logging

[archasek]

Hi,

Q1: is there a way to pass custom options directly to the llm? For example, I'd like to pass verbosity and reasoning_effort for got-5 family via azure openai. Couldn't make it work, and setting reasoningEffort in forwardOptions.modelConfig didn't appear in request body logger output.

Q2: Do you support openai's responses api? If not, do you plan to maintain various provider integrations or there's perhaps an easier way to get around different quirks?

Q3: Is it possible to track/log cached_tokens, reasoning_tokens in responses or only standard usage?

Q4: Why is stream field exposed in both modelConfig and forwardOptions directly whilst also possible to use it via streamingForward?

Q5: How can I log specific params per http request, such as top_p, etc? Simple logger shows request body with prompt messages only, i don't see params there.

Thanks

[archasek]

@dosco possible to address at least point 1 (with Jules or somethin)?
btw there was a discussion on that on the discord channel as well in that regard