Oh good, nice to know that the OpenAI proxy approach will work. Looking at the trace, it looks like Tool Choice = required But it's probably because I was calling Claude which isn't supported yet. No worries on the timing, I'm still in pre-launch research mode and the OpenAI tools are working fine for now. Thanks!
Thanks for the super fast turnaround on adding gpt-4o-mini! One minor (but important!) thing to note about gpt-4o-mini is that it has a 16K output token window. That's 4x all other GPT-4 models and twice the max of Gemini and Sonnet.
