MCERQUA/OpenVoiceUI — reverse-engineered prompt

Build me a local voice AI assistant I can run on my own machine with Docker. I want to open it in the browser on localhost:5001, talk to it naturally, and have it actually create things on screen in a live canvas, like dashboards, simple apps, pages, and tools while I watch.

It should feel hands free, with speech input, spoken replies, and a visual face or avatar. I also want an admin area where I can switch profiles, choose different AI and voice providers, manage plugins, browse workspace files, and check system status. Please include memory across sessions, the ability to switch projects by voice, and support for parallel helper agents if that is already part of the project. Music generation and a built in player would be great too if the repo supports it.

Please wire up the setup so it is easy to configure API keys and start everything cleanly. If anything is unclear, look up the current docs online and make it work end to end.