Bug 1. Attempting to upload an image to the chat with Claude leads to a popup saying it has switched to GPT-Vision and a weird combo name for the model (see screenshot).

Bug 2. Similarly although you can get Claude models with vision to work, if you swap between a GPT-4 and Claude based model using images the image shows correctly in the chat but doesn't seem to actually be sent to the model.

Request 1. Can there be a button on the prompt/model panel settings to allow us to quickly save system instructions and settings to the new assistants menu?

Request 2. Could there be a way to have a favorited list of apis to use. I basically switch back and forth between Claude and GPT-4, which involves a lot of clicking

Request 3. For the sake of clarity, can you make it say "custom" or something similar on the assistants selection pane if we have overwritten the system prompt. This confused me a lot for a bit because the assistant pane would say a certain thing but my prompt would be different. Then, if you click on the pane and select the prompt, it changes it. A simple ui change would make it much more clear. (Sorry if this is a bad explanation)

Request 4. Would it be possible to remember the tab we are on when opening up the settings bar? A very common workflow for me is to go to the advanced menu and mess with temperature slightly. This is mostly because if you do this the providers don't cache your result, so you get a more unique output. (It would be great if you added an advanced setting to do this automatically but I understand that would be a little annoying).

Request 5. Is there any way you could make the sliders be able to take values between say decimal integers? I understand why the slider having defined points is nice, but just having the ability to type into the box would be great.

I really appreciate the hard work you've done thus far. Hopefully, the product is doing well enough that you will be able to continue your development of it. Let me know if you have any questions.