You're asking how to make an agent speak after the user finishes speaking, but only if a certain amount of time has passed since the session started. Below is a concise pattern you can use and adapt. Key idea: • Record session start time in

on_enter()

. • Detect end of user speech with

on_user_turn_completed()

. • In that hook, check elapsed time and trigger the agent reply (once or per your rules) with

session.generate_reply()

or

session.say()

. Example (Python — minimal):

from livekit.agents import Agent
from livekit.agents.voice import AgentSession
import asyncio
from datetime import datetime

class TimedPostUserAgent(Agent):
    def __init__(self):
        super().__init__(instructions="You are a helpful assistant.")
        self.session_start = None
        self.triggered = False  # ensure one-time behavior

    async def on_enter(self):
        self.session_start = datetime.utcnow()

    async def on_user_turn_completed(self, turn_ctx, new_message):
        if self.triggered:
            return

        delay_seconds = 10  # time since session start required
        elapsed = (datetime.utcnow() - self.session_start).total_seconds()
        if elapsed >= delay_seconds:
            self.triggered = True
            await asyncio.sleep(0.5)  # optional slight pause
            await self.session.generate_reply(instructions="Say a brief greeting and ask how to help.")

If you want variations I can help with: • Triggering repeatedly (every N minutes) instead of once • Cancelling the behavior if user is active • Using

session.say()

directly for fixed TTS text Which variation do you want to implement? Sources: Pipeline nodes and hooks | LiveKit docs | Agent speech and audio | LiveKit docs