when using groq as a custom key, the gptoss frequently gets this eror:

when using groq as a custom key, the gptoss frequently gets this eror: AI_APICallError responseBody: {"error":{"message":"'messages.2' : for 'role:assistant' the following must be satisfied[('messages.2' : property 'reasoning_content' is unsupported)]","type":"invalid_request_error"}}. add a option in api settings page to include or exclude reasoning content.also, add ability to add multiple models and switch betw them. that will be convenient

Start by reproducing the Groq custom-key request with a gptoss model and inspect the API settings page and model-selection entry points. Done means users can include or exclude reasoning content and configure multiple models that they can switch between without triggering the reported request error.

Written by the indexing model from the issue text.