I just watched a short video about Substage and I like the idea of a nice input box below a window (e.g. the Finder, Terminal, any window).
Actions typed in the input box would be executed upon the selection in the window. Of course there’s no need to replicate the functionality of Substage: it’s just a nice idea that I would like to explore. As I see it, the commands don’t necessarily require AI.
Would it be possible to attach an input box to windows via Keyboard Maestro and use the selection in the window above it as an object to manipulate?
It would be, in theory, feasible to position a custom HTML prompt at the bottom of a window. But that box would not (easily) move or resize when you moved that window, and arbitrarily attaching one to every single window that exists seems like it would be very very difficult. (Well beyond anything I've ever tried with custom HTML prompts.)
And then, once you have the box there, you have to make it interact with the main window it's affecting, which is another can of worms, given the difficulty of determining the selection in some apps. And even in those apps where you can do it, you have to do so by copying the selection, working with the copied data, and pasting it back. Depending on what you're trying to do, this may make it impossible to accomplish.
Finally, if you wish to speak to it in natural language, you'll either have to interface with an LLM, or write a parser for the input box to try to determine what's being requested.
In short, my advice would be to leave such solutions to full-blown apps.
Substage only works in Finder, where it is relatively easy to target the current selection -- from a skim of the website it's passing the path(s) of the selected items to a natural language-generated shell command.
If you aren't going to use AI you'll be limited to pre-defined commands -- effectively a keyboard-driven Services and Contextual Menu system, which seems to be a lot of work for very little gain.
All the screenshots show one Finder window with one Substage bar underneath that. Relatively easy to implement with a Custom HTML Prompt, which you could move between active windows and resize to match with something like:
...assuming only one Engine window is open. You could match by window name, defined in your HTML, to be more precise. But movement is a bit jumpy, and I'll bet Substage is a lot smoother!
So you could create a pale imitation of Substage using KM. But is it worth it?
