sonicfieldlabs/oram — reverse-engineered prompt

Build me a minimal speech operated audio looper called ORAM. I want it to run mainly in the terminal, let me record and loop sounds from my microphone, switch between four layers, overdub, mute, solo, clear, save, export a mix, and control it with simple spoken commands or keyboard input.

It should feel like an experimental instrument, not a full music app and not a chatbot. Spoken commands should get turned into safe audio actions like reverse a layer, add texture, generate a bed, or make a listening report. Please include a mock audio mode so it works even without audio hardware.

I also want each performance saved as a session folder with audio, stems, metadata, a command log, and a listening report. Add an optional small browser dashboard for local control, plus a local daemon that a desktop app could talk to. Keep it private by default, with no cloud services required, but allow optional provider keys for generation later. Look up current docs online if needed.