rapidaai/voice-ai — reverse-engineered prompt

I want a self hosted voice AI platform that I can run locally and use as the base for real time conversational voice agents. Please set up the whole project so it starts cleanly with Docker and the provided make commands, with a working UI, API gateway, assistant service, endpoint service, and integration service. I want to be able to plug in my own model and voice providers using API keys, create and manage agents, stream audio with low latency, and support the usual voice flow like speech to text, text to speech, voice activity detection, and phone or web based conversations.

Please make the local setup feel production ready, with health checks, solid error handling, logs, call traces, and useful observability in the UI where possible. If the optional knowledge service is already supported here, wire that in too so documents can be added later. Keep everything easy to run and easy to inspect, and use the existing docs if you need current details.