I want to use the OpenAI compatible Gemini API endpoint but it errors because it defaults to GPT 3.5 as the model. Would be great to have the ability to set a custom model name as well as the endpoint url.