[FEATURE REQUEST] Support LM Studio

[shouse]

This should be pretty easy. Using Ollama configuration I was able to get a response generated:

{ "id": "chatcmpl-fnorwbq9iodgef6p6glf29", "object": "chat.completion", "created": 1741797539, "model": "deepseek-coder-v2-lite-instruct", "choices": [ { "index": 0, "logprobs": null, "finish_reason": "stop", "message": { "role": "assistant", "content": " 4" } } ], "usage": { "prompt_tokens": 169, "completion_tokens": 2, "total_tokens": 171 }, "stats": {}, "system_fingerprint": "deepseek-coder-v2-lite-instruct" }

Looking at the Response schema for OpenAI it looks like the parsing should be the same for choices.