Thanks! So the iOS/iPadOS is to select the whole Text, while the Mac can select a range of text. Get it, but it make the textSelection less useful as the Text is very long (output by Foundation Model).

It is not guardrail, just answers back from LLM. Actually the first question is about Apple II (but sometimes it answers).

Yes,it is solved in beta 4.

As in, you can't select the text?

It works on beta 3... then error some other days. No OS changes.

Thanks! I didn't see the code, only the portion on ignoring the visionOS.

The code works fine for visionOS, just some part of the code.

So far, the only solution I find is to rewrite the upper part of UI as a toolbar. Then it will go around the controls in toolbar. I am not aware of other ways.

Just normal using session.streamResponse(to: question, options: options), around the 4th session I got this error (typically I will go around 8/9th round before exceededContextWindowSize error).

Not using streaming, I also got this error: InferenceError::inferenceFailed::Failed to run inference: Context length of 4096 was exceeded during singleExtend.. catch LanguageModelSession.GenerationError.exceededContextWindowSize still does not work.

Thanks, after repeated testing, I can get Vision Pro simulator finally appear on my iMac 2017! However: first time I try, the UI is black and later appeared glitchy. 2nd time I try, the UI shows up but app takes a very long time to load. I have to give up doing Vision Pro development on Intel Mac for now.

I downloaded it the 2nd time zip came out. The rest not sure what's the need.

Unfortunately in beta 3 the issue is back.