Equipa

Sirena is a voice assistant for the Mac that lets you use the whole computer without touching the keyboard or mouse. You say "Sirena" and then what you want (Ex: "switch to Chrome", "go to the first tab", "click Send", "read the article", "move this window to the other screen") and she does it. You can also say "Sirena, start dictating" and she types everything you say into whatever app you're in, cleaned up, until you tell her to stop. I had a different project in mind for this hackathon, but on Tuesday (27), I got injured at the gym :( and when I got home, I couldn't get up off my bed or hold the computer in my hands comfortably enough to program or do anything else. Then, I installed some dictation and controlling tools, but they didn't do the job, and it was then that "Sirena" came to my mind. I wanted her to feel fast and safe, not like a chatbot you have to wait for. So Sirena reads the screen through the macOS Accessibility API (app names, tabs, buttons, text fields) and acts with keyboard events and AppleScript. Simple commands like switching tabs or scrolling never reach a model at all. Harder ones go to Muse Spark with a small text snapshot of the screen. Only when that isn't enough does she take a screenshot and use computer use. Anything risky, like sending a message or deleting something, gets a short question first ("Send 'see you at 5' in WhatsApp? Yes / Cancel"). She can't act until you answer. There's a lot more around that: a numbers mode that puts a badge on every button so you can say "click 7", a shortcut book for your own multi-step commands, notes and snippets, read-aloud, multi-monitor support, Portuguese commands, and a history of everything she did with timings. It's a native Swift menu bar app.