Automatically nudge users back into the conversation after a period of inactivity
A Wakeup Call automatically triggers an action when the user has been inactive for a specified period of time. Instead of leaving a silent gap when the caller goes quiet, the agent can gently check in, re-prompt, or offer help, keeping the conversation alive and maintaining engagement.
Set wake_up on your Pipeline to the number of seconds of caller silence to allow before nudging. When the caller stays quiet for that long, the runtime calls your agent’s on_wake_up method, where you decide what happens next.
pipeline = Pipeline( stt=CartesiaSTT(model="ink-2"), llm=OpenAILLM(model="gpt-5.4-nano-2026-03-17", streaming=True), tts=SarvamAITTS(streaming=True), vad=SileroVAD(), turn_detector=TurnDetector(model="echo_large"), wake_up=10, # nudge after 10s of caller silence)class PatientAgent(Agent): def __init__(self) -> None: super().__init__( name="PatientAgent", agent_id=AGENT_ID, instructions=( "You are a patient assistant. Answer questions and help the caller. If they go " "quiet, you'll gently check in on them." ), pipeline=pipeline, ) self._nudges = 0 async def on_enter(self) -> None: await self.session.say("Hi! Take your time — I'm here whenever you're ready.") async def on_exit(self) -> None: await self.session.say("Goodbye!") async def on_wake_up(self) -> None: # Called by the runtime when the caller has been silent for `wake_up` seconds. self._nudges += 1 await self.session.say("Are you still there? I'm happy to keep helping.")
Track state across nudges (like the _nudges counter above) to escalate your response — for example, offer more help on the first nudge and end the call after several unanswered check-ins.
