Wow, this looks really cool.

What is the current support for OpenAI proxy or non-GPT models?

For example, using locally deployed Qwen models or LLaMA models.