Quick Prompt Frozen after initial AI Command run and focus is returned to the target application.

[stevenmoto]

Hello,

I am running local ollama models only with no internet access and have only Ollama defined in settings.
The system is currently Windows 10 - LTSC.
Chat in Witsy is working fine; I would like to use other applications such as notepad, word, etc.
To do this I am using the Run AI Command Shortcut after highlighting text in the application.

This returns uses the local ollama model correctly and returns results that I can insert and manipulate. Cursor and clicking are working fine at this point.

Now, I return to the application which returns focus to it. At this point, I run commands again however the quick prompt is frozen and I unable to have cursors, change text, clear, etc with it.
Same with just returning to the quick prompt.

At this point I have close Witsy application, relaunch and I can run the commands again but the issue is reproducible 100% for me at this time.

Please see my screenshot below, quick prompt frozen and I can't do anything with it.

To Reproduce
Steps to reproduce the behavior:

1. Only define a local ollama model in settings
2. Open Notepad for a new document
3. Enter Text to highlight and run an AI command against it
4. Get Results, notice the cursor is working and can be clicked
5. Return back to notepad to have quick prompt out of focus
6. Run AI command again, no new text is generated, and unable to manipulate anything in the quick prompt of Witsy

Expected behavior
Be able to run Commands in Quick Prompt for Application, Return to Application, and run quick prompt again without having to close the Witsy application and relaunching

Desktop (please complete the following information):

• OS: [Windows 10 LTSC 2019]
• Witsy Version 2.14.0

[stevenmoto]

In addition, I tried with Windows 11 24H2 and the frozen quick prompt is also reproducible.

[stevenmoto]

An update to this and maybe help pinpoint the issue, I was not able to reproduce this issue on my Physical Linux Desktop nor even my Physical Window's Desktop (with using RDP).

Still an issue with my Virtual Desktops for Windows 10 and 11 with compiled later source code. My Virtual Windows Infrastructure does not have Hardware 3D acceleration (all software based), could it be a possible issue with that?

I will attach a log snip-it of when I ask the LLM and insert from quick prompt into my application. In this example, I have an Custom OpenAI API defined (Local Network OpenWeb) for the model.

log_snippit.txt