Alpaca's Scope

[Jeffser]

Hi everyone, I've been reading your submissions from the form I made a week ago, of course I'll wait a couple of months before publishing all the results but for now I wanted to follow the suggestion that someone made about being clear with Alpaca's scope, so that's what I'm going to do.

Some of these might be based on past discussions, issues and submissions from the form, thanks!

Future Features

• Latex rendering
• 🔧 Development: Categorization of models #420
• 🔧 Development: SQLite transition #422
• Better model creator
• Make and attach screenshots
• A nice wiki (in progress)
• Separate Flatpak and Snap packages with and without Ollama included (Since 3.0.0)
• Download code blocks as files
• Better integration with narrator (Very important!)

Ideas Worth Looking Into

• MacOS support (Experimental)
• Better website
• Streamline translations
• Plasma search plugin
• Attach photos directly from webcam
• Export chats to PDF
• API integration (ChatGPT, Gemini)

Probably Not Unless Someone Else Wants To Do It

• Windows support

Definitely Won't Include

• Image generation (really against that, sorry)
• Serious release names (What's the fun in naming an update just by it's number?)

[olumolu]

But image genaration could be fun or if you agree there are models which can upscale to remove background from image to image and this is something i really find useful as sometime you need to have a white background photo but with gimp you need huge time and Photoshop is not supported in linux so that thing works really good just in couple of sec you have background removed...
Other than that Categorize available models inside the model manager
Better model creator are good to have feature.

[Jeffser]

I get where you are comming from, but I won't add image generation, is weird for someone making an AI tool to draw the line there but that's what I'm going to do.

I simply do not vibe with image generation, there are other tools out there that can do the job but Alpaca won't be one of them and won't promote that kind of thing.

The only way I would add image generation would be if first Ollama adds that functionality and if it gets normalized which I hope it won't.

[Jeffser]

Update on Categorize available models inside the model manager

I reprogrammed my internal tooling for extracting model metadata from Ollama's website, it now gets categories like math, code and multilingual, I also manually wrote each models languages once it detected it was a multilingual model.

Anyways, it's going to be useful to make some filters and cool stuff in the model manager