VectorlyApp/open-telephony-stack ? reverse-engineered prompt

Build me a production ready phone system for a voice AI app that works like a Twilio alternative, but uses Asterisk and AWS Chime instead. I want it to handle inbound and outbound calls, connect a real phone number, and stream the call audio into my own AI voice server over WebSocket.

The app should include a FastAPI shim that talks to Asterisk, manages each call, and bridges RTP audio to a Twilio style media stream format so I can plug in my own voice AI later. Also include an example voice agent server that can connect to OpenAI Realtime API, handle transcripts, interruptions, and let the AI end a call when needed.

Make it runnable locally with Docker or a simple setup flow, with clear health checks and sensible defaults. If you need to look up current docs online for Asterisk or AWS Chime, do that too.