addyosmani/jarvis ? reverse-engineered prompt

Build me a web app like Jarvis, a real time AI assistant that can talk with me using voice, see my camera, and answer back naturally while the conversation is happening.

I want it to feel polished and simple to use, with a clear main screen for chatting, starting and stopping live audio and video, and showing the assistant’s responses as they come in. It should support asking normal questions, searching the web for current info, and generating or reimagining images when I ask for that. Please make it work with Gemini Live API style streaming so the replies feel instant and conversational.

Use the existing design as inspiration, keep the interface modern, and include the setup needed for API key based local use. If anything needs current docs or best practices, look them up online first.