IRISX-AI/IRIS-Mini — reverse-engineered prompt
Reverse engineered prompt
Build me a polished local first voice assistant called IRIS Mini that works both as a sleek web app and as a terminal assistant. I want it to use Google Gemini Live for fast, natural voice conversations, with a first run setup that asks for my Gemini API key and lets me choose between two voice personalities, Lyra or Puck. It should use the microphone, speak back, keep simple local memory, and handle useful Windows actions like opening, closing, and managing apps. It should also be able to search the web for current information when I ask.
Please make the terminal version feel clean and premium, and make the web version feel modern with an animated visualizer and smooth UI. I want a simple local development flow, plus a Docker option for headless use, with the understanding that voice and Windows integration may be limited there. Keep the whole thing practical and easy to run. If you need to, look up current docs online and wire everything together so it works end to end.
Want more depth? Deep Reverse