Error:

"Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.”

Docs: https://platform.openai.com/docs/api-reference/chat/create

This value is now deprecated in favor of max_completion_tokens, and is not compatible with o-series models.

Renaming the field in CreateChatCompletionSettings would break a lot of clients I presume?

Another option is keeping CreateChatCompletionSettings intact, and changing only OpenAIChatCompletionService to send maxCompletionTokens instead. This way we wouldn't change the universal interface, only the OpenAI-specific implementation.

I'm opening a Draft PR to help the discussion!