Majiks-Studio/majiks-music-studio — reverse-engineered prompt

Build me an open source desktop app for local AI music creation, basically like an AI producer I can chat with. I want to type in a vibe, genre, mood, BPM, key, instruments, and even lyrics, then have it generate a full song with vocals plus stems in a few minutes. It should feel like a real studio, so include a chat based creation flow, a more detailed session view for tweaking parameters and making variants, and a library where I can browse, tag, search, and organize tracks with album art and metadata.

Please make it work with the ACE Step 1.5 music model, support export to WAV, FLAC, MP3, and M4A, and include stem separation for vocals, drums, bass, and other parts. Keep everything local after the model download, no cloud account required. If possible, match the repo setup with a native macOS app and a cross platform desktop version for Linux and future Windows support, sharing the same backend. Look up the current docs online if you need to.