@lmstudio/sdk: method to declare a generator's capabilities

[cc1160]

I built a generator based on the openai-compat-endpoint with file‑handling, vision and image support. Text attachments work as expected, but image attachments do not; the UI refuses to process them.

Maybe this is more of a missing feature in the @lmstudio/sdk library than a bug in the LM Studio client.