Excellent   

Gonna check this puppy out !!

Seems neat, here are some issues I found so far:

• The keyboard shortcut input form seems a bit dodgy, not recognizing key presses sometimes, as well as some key combos not working.
• F13 to F24 appear not to be usable right now, which is unfortunate, since I would have liked to make a conflict-free mapping to one of my macro keys. Double unfortunate since otherwise, Left Shift + Windows key + F23 could be mapped to it, which is actually the key combo used by the copilot button on some new Windows keyboards.
  
• Text appears not to render correctly. It looks like one token is displayed, but then it immediately disappears.
• Prompt input is extremely sluggish, it feels like about .5 seconds of latency on some key presses.
• I had some kind of dotnet exception window being displayed after closing the popup with Escape, then trying to show it again from the system tray. I could not replicate this behavior, however.
• The transparency slider in the settings is super-duper laggy, sometimes not responding at all.
• My available tokens are displayed as 00 (double zero)

Additionally, here are some things I would like to see:

• Local AI, both for privacy reasons and because it means that I won't have to keep buying tokens to use it. This could be implemented with something like llama.cpp or Ollama (which is based on the former)
• Dark mode. I'm guessing this is on the list already, but just wanted to add my +1

Finally, my system info:

• Windows version: Windows 11 Pro 23H2 (Build 22631.3527)
• CPU: AMD Ryzen 7 3700X
• GPU: AMD Radeon RX 7900 XTX
• RAM: 32 GB DDR4, 3200 MHz
• Mainboard: B550 GAMING X

Even just the ability to specify an OpenAI-compatible endpoint would be very useful for those like me who are already running AI models locally. I'd like to try DesktopGPT out, but I'm not willing to send any personal info to OpenAI.

All I get is this and pressing OK close the app:

As someone not really familiar with how these LLMs work, I can't help but notice the email I received mentioned the use of 75000 tokens, implying some sort of limitation on the software. Clicking the link to "learn more about DesktopGPT" also mentions receiving 75000 tokens, but neither explains what these tokens are, how they're used, or how to get more once you run out. 

I suppose the product marketing assumes you already know these things?

Agree. I might install it later to mess around, but "now available for you as part of your Object Desktop subscription at no additional cost" and then immediately following with "We are also giving you 75,000 tokens to use with the application" doesn't paint the clearest picture of how this works. It's available at no additional cost, unless it does cost additional?

I think I've managed to figure it out, but a disclaimer on the product page is probably a good idea.

This is just a window to use ChatGPT in. It does not run the AI or pay for that AI service, outside of the starting 75000 tokens. You have to pay for the ChatGPT tokens yourself.

A token is roughly worth 0.75 words. So 1000 tokens is 750 words. 75000 is 56250 words. This includes both input (what you send GPT) and output (what it sends you back).

ChatGPT token prices are found here, it differs based on the model used.

I would like to see this GPT included with DeskScapes at no extra cost ( no tokens )

If not I probably won't use it much ( if at all ) .

Like mentioned above you still have to have AI programs which the good ones cost money.

I've been using  AI programs for over a year now so I'm pretty good with writing promts

which varies with the different AI programs  .

So more money for more tokens doesn't work for me , when I tried it out earlier I typed in

3 or 4 word combos and it cost me 13000 tokens at the rate 75000 won't last long .

After paying for Object Desktop already , I can't afford more than that .

Jan Oscar

reply 4

All I get is this and pressing OK close the app:

Hello,
Sorry to hear you are having issues. For those who also might have similar issue. Please re-download the installer and reinstall. I believe the connect server issue was resolved internally. And the current installer should no longer have this issue.

Thank you.
Basj,
Stardock Community Assistant.

I feel the app should run in the background and have a hot key to bring it to the front of the screen. As it currently is, it is just an open task taking up space on the taskbar.

Agree.

SimplexDesigns

reply 10

I feel the app should run in the background and have a hot key to bring it to the front of the screen. As it currently is, it is just an open task taking up space on the taskbar.

Hello,
Sorry to hear you are having issues. But, you can setup hotkey to make the app appear.

Thanks
Basj,
Stardock Community Assistant.

basj

reply 12

Quoting SimplexDesigns,

reply 10

I feel the app should run in the background and have a hot key to bring it to the front of the screen. As it currently is, it is just an open task taking up space on the taskbar.

Hello,
Sorry to hear you are having issues. But, you can setup hotkey to make the app appear.



Thanks
Basj,
Stardock Community Assistant.

Very finicky still. Takes up taskbar space and does not obey the locked window position I had set it to.

Looks really very promising. But in practice, so far, I more often get a 'Failed to retrieve data' response than a meaningful response.

When will it be able to render code accurately? It doesn't parse markdown so now it just shows ```csharp for instance

Will this ever be a standalone purchase or always part of Object Desktop?

I will be impressed if it does things like 'Create gadget showing system ressources with 6 cores and 4 harddisks using skin XY'.

As of right now, the current plan is that it will stay exclusively part of Object Desktop - plans may could change in the future though.