Thanks — you’re asking about best practice for redirecting callers to a backup phone number when part of your pipeline is down. I’ll answer with what you can do using LiveKit’s telephony primitives and a couple quick questions so I can tailor the recommendation. Short guidance (what you can do) • If you simply need to move the caller to another number with no briefing: use a cold transfer via

TransferSIPParticipant

(SIP REFER). This is the simplest automatic redirect. • If you need context preserved and a human to screen/brief before connecting: use a warm transfer flow — create a SIP participant for the backup number (

CreateSIPParticipant

), brief them privately, then connect the caller (

MoveParticipant

or

WarmTransferTask

). • Add hold/announcement audio and pre-handoff verification so callers don’t hear dead air and the backup agent knows call context. • Use multiple SIP trunk providers (Twilio, Telnyx, Plivo, etc.) and webhook/health checks to detect provider or pipeline failures and trigger the transfer logic automatically. • For non-telephony failures (STT/TTS/LLM), implement model fallback adapters to keep the voice path working; telephony transfer is separate from model-provider fallbacks. Questions to tailor this: 1. Do you want transfers to be fully automatic (no human) or require a human to accept/brief before connection? 2. Which parts of your pipeline are at risk (SIP trunk, model provider, LiveKit service)? 3. Are you using a single SIP provider today or multiple providers/trunks? If you tell me your answers I’ll give a concise implementation pattern (API calls and health-check logic) matching your needs. Sources: The Human-in-the-Loop (HITL) Pattern for Voice Agents | LiveKit | Agents telephony integration | LiveKit Documentation | LiveKit Phone Numbers | Connect voice agents to phone lines | LiveKit | Fallback strategies | LiveKit Documentation, with interactive elements