Been working on a #opensource GUI that allows you to connect to any #AI provider (even local like #ollama ) and have a unified chat history, characters, tools, documents etc.

A bit like OpenWebUI but aims to be more user friendly and lighter (but hopefully just as powerful!)

It’s called #CompassAI - built with React Native so it’s possible to compile for pretty much any platform including as mobile Android and iOS app

Please feel free to check the GitHub at https://github.com/compass-ai-chat/compass#user-content-welcome-to-compass-

@opensource @privacy @artificialintelligencenews.in @programming

  • bagelberger@lemmy.world
    3·
    2 days ago

    This reminds me of another open source AI agent called Goose that connects to multiple models/providers.

    Being built by the folks at Square/Cash App

    • nordwestt@techhub.socialOP
      101·
      2 days ago

      @semperverus Damn, that’s crazy! Different technology but very similar aesthetic. I’m a bit jealous that they already have Windows and MacOS builds working - I’m struggling a bit with that but I know I’ll get there sooner or later ;)

      • Semperverus@lemmy.worldEnglish
        9·
        2 days ago

        Looking forward to seeing your work - it’s always good to have competitors, and gpt4all is also very crashy. If you have a lead in stability, I’d definitely use yours over theirs.

        Some other areas you could probably look into if you want to differentiate are:

        • Getting Started experience - recommend some high quality models and update the list as time goes on. Maybe include a good default one as part of the package.

        • Convenience - include a way to do what the modern chat interfaces do where asking it to do something other than text will call a different AI model built for that purpose and return the result (image generation, etc)

        • Voice conversations - Can we actually talk to the dang thing?

        • Assistant module - piggybacking off of the last one, can we invoke it with a wake-word or a button press and have it “always available” (similar to HomeAssistant with a Whisper plugin, but on-device).

        Anyway, I wish you well in your endeavor and will keep an eye out.

        EDIT: looks like the conversational bits are on your roadmap, and you do have some basic suggestions on startup.

        As for voice, the OpenWhisper module might fit your project’s theme a bit closer than elevenlabs.

        • nordwestt@techhub.socialOP
          21·
          2 days ago

          @semperverus Really appreciate the feedback - and I definitely agree that we need some competition

          Yeah, overall I’m putting user experience high - it should be snappy, intuitive and robust

          Thanks for the encouragement, and definitely keep an eye out for progress as I’m working on it every day

          Hopefully I get Windows build working this week

          • Semperverus@lemmy.worldEnglish
            2·
            2 days ago

            Are you using Pipelines in either github or another upstream source management platform? I don’t know if they have a free build plan or not, but you can have an entire pipeline that just spits out an executable for one or more platforms every time you commit to your main branch, depending on how you have compilation set up (you can have it use both a Linux and a Windows VM for different steps in the pipeline too). It can even handle publishing them to a website if youre handy with bash and/or powershell scripts (or python or JS or whatever you can call from the pipeline).

            I use the Azure DevOps version at work and its amazingly useful, but very confusing to learn at first.

      • hayk@lemmy.ml
        2·
        2 days ago

        i’d also recommend adding a way to pull the models (say for ollama this is trivial). that’s one of the things I liked about owui

        • nordwestt@techhub.socialOP
          21·
          2 days ago

          @hayk Already there - models can be pulled inside the “Providers” section if you have Ollama added

          But I’d like to make it more apparent and easier (perhaps accessible straight from the front page)

  • warmaster@lemmy.world
    1·
    2 days ago

    How easy is it to install the software itself and AI models?

    Do you plan to integrate with any of the following?:

    • Home Assistant
    • Active Pieces / n8n / Automatisch
    • Nocodb / Baserow
  • bonn2@lemm.ee
    1·
    2 days ago

    This reminds me a lot of the Alpaca UI. I’ll have to give this one a go!