[Disclaimer: I am the author] I completely agree—I wouldn’t install it either without reviewing the code first! Thank you for being the first person to provide feedback on my tool; I really appreciate it.
Here’s how it currently works:
Press the Option key twice.
A floating window appears.
Select the action you want to perform.
Press Enter—the window closes, and the text is replaced.
The floating windows function processes all callable objects defined in the extensions package as text-handling functions.
This repository provides the backbone structure that makes it all possible. Users only need to add their custom logic as a Python callable in the format (input_text: str, *kwargs) -> str anywhere within the extensions package!
For example, implementing an OpenAI call is as simple as defining a function like this:
The function should accept a string as input and return a string as output, containing the proper implementation (e.g., HTTP calls, prompts, etc.). It's straightforward and intentionally left minimal.
Everyone is welcome to contribute, add their requirements, and customize the code to fit their specific needs with Orange Intelligence.
Ok so the answer is that these integrations are not implemented, even if the configuration has a field for an OpenAI API key (obviously just as an example).
Thank you for clarifying. Good luck with this project! I'll keep an eye on it and see how it evolves.
They aren't Open Source, so I would not expect them to be supported. The point of this project appears to be to keep full control of your intelligence pipeline.
Neat product. AI provided by the OS always creeps me out.
No offense to oranges (they're indeed one of my favorite fruits), however I can't help but feel that the name constitutes some kind of oxymoron. I know oranges are technically inanimate objects, yet I'm still at a loss to explain such a feeling. Oh well.
I have and actually find it very impressive. Herein lies the rub, however:
Non-targetability. An attacker should not be able to attempt to compromise personal data that belongs to specific, targeted Private Cloud Compute users without attempting a broad compromise of the entire PCC system.
The thing is, there's long been precedent for broad compromise:
Further consider that a likely part of any broad compromise strategy is utilizing assets who may work at Apple. There's precedent for that one, too—at least in terms of unspecified US tech companies.
Maybe because the phrase "comparing apples to oranges" is generally a defensive stance of the person making the argument? i.e. you can't imply that! It's like comparing apples to oranges! If Orange Computer, Inc created the Macintosh (okay, created the Cutie) in the 80s would you have the same reaction to "Apple Intelligence"?
Careful before installing OS-wide application with a single git commit, from someone with no public GitHub activity.
Speaking of the tool, would be great to be able to trigger the action immediately by pressing a keyboard shortcut.
[Disclaimer: I am the author] I completely agree—I wouldn’t install it either without reviewing the code first! Thank you for being the first person to provide feedback on my tool; I really appreciate it.
Here’s how it currently works:
Which part of this flow would you change?press a third time to perform a default action or replay last action
Poking around in the code, I don't see how this can be used with OpenAI or other SaaS LLMs...
Is that part still WIP?
Hi, thank you for checking out the project!
The floating windows function processes all callable objects defined in the extensions package as text-handling functions.
This repository provides the backbone structure that makes it all possible. Users only need to add their custom logic as a Python callable in the format (input_text: str, *kwargs) -> str anywhere within the extensions package!
For example, implementing an OpenAI call is as simple as defining a function like this:
https://github.com/sharingan-no-kakashi/orange-intelligence/...
The function should accept a string as input and return a string as output, containing the proper implementation (e.g., HTTP calls, prompts, etc.). It's straightforward and intentionally left minimal.
Everyone is welcome to contribute, add their requirements, and customize the code to fit their specific needs with Orange Intelligence.
Ok so the answer is that these integrations are not implemented, even if the configuration has a field for an OpenAI API key (obviously just as an example).
Thank you for clarifying. Good luck with this project! I'll keep an eye on it and see how it evolves.
Thank you!
I've just introduced a super simple integration that uses OpenAI to generate jokes based on the <selected text> topic!
Of course, it can also be used to create more meaningful content.
I’d love to hear your suggestions or feedback on how to make this even more helpful!
They aren't Open Source, so I would not expect them to be supported. The point of this project appears to be to keep full control of your intelligence pipeline.
I mean they claim to be in the write-up on this article.
I don't really see anything that substantiates your statements here, but maybe I missed something.
How does this capture all of the keyboard events?
Using pynput (https://pypi.org/project/pynput/), it listens for the option key being pressed twice, which triggers the opening of a floating window.
This serves as a signal that the user wants to process the highlighted text, prompting the floating window to open.
Once the floating window appears, the already-typed text in the focused application is captured and processed.
Neat product. AI provided by the OS always creeps me out.
No offense to oranges (they're indeed one of my favorite fruits), however I can't help but feel that the name constitutes some kind of oxymoron. I know oranges are technically inanimate objects, yet I'm still at a loss to explain such a feeling. Oh well.
It's worth reading about the expense and effort Apple goes to to ensure your privacy, whether you're creeped out or not:
https://security.apple.com/blog/private-cloud-compute/
I have and actually find it very impressive. Herein lies the rub, however:
Non-targetability. An attacker should not be able to attempt to compromise personal data that belongs to specific, targeted Private Cloud Compute users without attempting a broad compromise of the entire PCC system.
The thing is, there's long been precedent for broad compromise:
https://news.ycombinator.com/item?id=6641378
Further consider that a likely part of any broad compromise strategy is utilizing assets who may work at Apple. There's precedent for that one, too—at least in terms of unspecified US tech companies.
Such a strange pair of articles:
https://www.bloomberg.com/news/features/2018-10-04/the-big-h...
https://www.bloomberg.com/features/2021-supermicro/
Maybe because the phrase "comparing apples to oranges" is generally a defensive stance of the person making the argument? i.e. you can't imply that! It's like comparing apples to oranges! If Orange Computer, Inc created the Macintosh (okay, created the Cutie) in the 80s would you have the same reaction to "Apple Intelligence"?
Potentially. I never thought of it like that; it's certainly a clever name.