Feedback and QoL suggestions

[JaneSmith]

Hi there! First off, thank you for creating this program. I've been using it a while and really enjoy it. It's nice to have a native GNOME-y interface, and I find it better overall than Newelle (which is very messy and confusing, although perhaps more powerful).

I have a list of suggestions and observations below. This is just my personal feedback, if it's of any use. There's no need to listen to it or change anything if you disagree with it.

General chat things:

• I'd like to have an option to disable auto-scroll during AI generation. The forced scroll makes it difficult to read the message as it's being written. If I attempt to scroll up with the mouse wheel, it just gets overridden by the auto-scroll and forced back down until the message is complete. The only workaround I can find is to grab the scrollbar and drag it up, but I have to do this every time a message begins.
• I would like the ability to toggle thinking on both a per-message and per-model basis, instead of it only being a toggle for the entire Ollama instance buried in the settings menu.
• When thinking is enabled, I would like an option to hide (minimise) the thinking output by default.
• For the "Preferred Activity Mode" setting, I would like a "Hide" option. For example, when using web search, the sidebar popping out to show websites can be distracting for me. I’d prefer the web search to just happen in the background.

Quick Ask:

• I'd like the app to remember the Quick Ask window size. Currently, it opens too short to display the suggested prompts, which is a bit annoying. And I'd generally just prefer it to be a bit taller.
• Quick Ask always seems to default the model to whatever model is at the top of the list. I would prefer it to default to the selected default model, just like with regular chats.
• Is it possible to enable a global keyboard shortcut for Quick Ask, so that even when the Alpaca window is closed (running in the background), a Quick Ask window could be quickly opened?

Chat management / sidebar:

• There doesn't seem to be any way to re-order chats or folders by directly dragging and dropping. As a workaround, I've found I can drag something into a folder and then drag it back out to the parent location, but this is a bit awkward.
• Mouse navigation of folders is a bit tedious, because folders are placed at the top, while the "Back" button is all the way at the bottom, so I have to keep moving my mouse all the way up and down while navigating folders.
• Keyboard navigation with the Up/Down keys only seems to work for chats, but not folders. I can't seem to select folders by the keyboard at all.

Bugs:

• The "Characters" functionality doesn't seem to work at all. I can set them up and then select the character from the initial chat screen, and it appears to upload a file when I do so, but it doesn't affect the LLM at all. The LLM just works as usual.
• In the Ollama manager, my status reads "Incompatible GPU, Vulkan might work," yet it still works fine and utilises my discrete NVIDIA GPU even with Vulkan disabled.
• When clicking a button like "Thought" or "Metadata", the button seems to do nothing at all if the information is already open in a background tab. The expected behaviour would be to bring that tab to the front.
• When I attach an image, it appears cut-off in the Alpaca chat history, with a horizontal scroll bar visible. The expected behaviour would be for it to fit the entire image. I wonder if this might be DPI related?
• If I try using an OpenAI-compatible API backend instead of Ollama (local llama-server), tool calling seems completely disabled? Is this not supported?

UI confusion:

• As someone using only Ollama, the UI feels a bit fragmented. Some settings are in "Preferences," while others are hidden under "Manage Instances -> Edit". I know this would make more sense when using lots of different backends, but in my case it's confusing. But then again, it's mainly the "thinking" toggle that I need to keep going into the Ollama settings for, so if that were made available elsewhere then I wouldn't need to keep diving into the Ollama settings...
• Whenever I go to edit the Ollama settings, I always find myself hesitating in the "Manage Instances" window, feeling confused between the "Open Ollama Manager" and "Edit Instance" buttons for a few seconds. It isn't immediately clear which one leads to the settings.
• When clicking a button like "Thought", it opens the thinking output in the sidebar. If I then click the "Thought" button again, nothing happens. I would prefer that it close the thinking output tab, and toggle the sidebar back off. Also, if the sidebar is already closed it, it should open it instead of doing nothing.
• When the model makes a lot of steps in its processing, it becomes hard to navigate/understand the buttons in Alpaca. For example, at the top of the AI response I see "Web Search" "Web Search" "Web Search" "Web Search" "Web Search" "Web Search" "Web Search", where the area has a horizontal scrollbar when I mouse-over. That horizontal scrollbar partially overlaps the buttons, making for a difficult click-surface. And after I click one and view it, I find that I very easily lose track of which one I'd just clicked and looked at, because the buttons all look the same. There's no indicator of which button maps to which thing is currently being displayed in the sidebar.
• If I try using an OpenAI-compatible API backend instead of Ollama (local llama-server), Alpaca doesn't seem to be able to let me automatically pick any of its models for usage. Instead of being able to use them straight away, I have to go to "Manage Models", then add/register them, and then they show up as usable. I don't understand why this step is necessary. I feel it adds confusion and extra friction.

Thanks.