Mystery-Machine 3 days ago

I'm using Raycast AI (paid subscription) to interact with AI models through the use of hotkeys. It can even act upon selected text. I've got several hotkeys set up: - translate (selected text) to English/Spanish/Italian (3 hotkeys) - fix (selected text) English/Spanish/Italian grammar (3 hotkeys) - add (selected text) to current chat prompt where I can continue typing

There's also a built-in "Summarize webpage" AI action that I assigned a hotkey to and I can get a nice summary of the currently opened webpage without having to read the whole article.

These save me so much time and help me better my language skills. Especially when fixing grammar. When fixing grammar, it shows you the diff so you can see what you wrote incorrectly and how to write it correctly.

It also allows you to paste the text, from the app window into the last used app simply by pressing ENTER. I also use this a lot.

Just some feedback.

SimianSci 3 days ago

Id personally be interested to know how this compares to a solution such as Ollama, which I have been using to run local models.

  • talldayo 3 days ago

    Linux users have a pretty excellent native Ollama client called Alpaca, it's well worth checking out: https://flathub.org/apps/com.jeffser.Alpaca

    It's still early in development but supports multimodal and some nice looking code blocks.

rahimnathwani 3 days ago

Slightly off topic: some of the tools in HuggingChat don't work, even one of the official ones (Image Editor):

  Error calling tool Image Editor
  Parameters
  image: output_2_0.webp
  prompt: add a smiling emoji
  Error
  An error occurred while calling the tool edit_image: No API found


  Error calling tool TTS (fast)
  Parameters
  text: hello
  description: Laura's voice is monotone yet slightly fast in delivery, with a very close recording that almost has no background noise.
  Error
  An error occurred while calling the tool gen_tts_fast: Error: Error: Could not resolve app config.
kelsey98765431 3 days ago

first impression: it's slick, looks a bit rough but promising, looking forward to using this and hope it improves.

isodev 3 days ago

What does it do? Isn’t HuggingChat already available as a dedicated web app (https://huggingface.co/chat/)?

  • archiv 3 days ago

    Yes but you now have the ability to run queries with a shortcut from your Mac, as well as switch to a local model using a dedicated keyboard shortcut. See demo: https://x.com/cyrilzakka/status/1838618605648490974

    • isodev 3 days ago

      Sorry, this link is not accessible for me. It would be nice to add some more information to the repo’s readme.

      • archiv 3 days ago

        @isodev sorry about that! Just uploaded a video to the README

    • Sidnicious 3 days ago

      Would it be possible to let you run local models without an account?

geniium 3 days ago

Interesting. I am an ai enthusiast but I don’t understand exactly what your product does. Run locally ? Run remote on HF? Would love more info.

  • archiv 3 days ago

    Both! You have the option of using HF’s HuggingChat or switching to a local model using a keyboard shortcut

privacyis1mp 3 days ago

Congrats! I noticed it was initially closed-source and now it's opened, nice! Is it a project funded by HuggingFace? (Not immediately clear to me).

I've built a macOS assistant too (more advanced though), with focus on privacy and easy of use (https://getfluid.app). I'd love to open-source it, but not sure about sustainability of such business model. Right now I'm experimenting with a fully private paid Llama hosting (for GPU poors).

Good luck :)