After my agent finishes speaking its response, the session just sits in silence waiting for the user to say something. If the user does not reply, nothing else happens—there’s no follow-up prompt, timeout, or automatic session end. It essentially stays in an idle state indefinitely. What I’m Looking For I’d like guidance on how to detect this “post-speech silence” and either: • automatically send a follow-up message (e.g., “Are you still there?”), or • gracefully end the session after a configurable period of inactivity. I understand this is different from the

resume_false_interruption

feature (which handles pauses during speech). This is specifically about what happens after the agent has completely finished speaking and the user never responds.

Thanks for reaching out, Manoj. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

I’m using the LiveKit Agents Python SDK to run a voice assistant. The core flow is inside an

entrypoint(ctx: JobContext)

where I create an

AgentSession

(with VAD, STT, TTS, and OpenAI LLM) and call

await session.start(agent=Agent(instructions="..."), room=ctx.room)

The silence happens inside this AgentSession after the agent has finished speaking its reply. No special webhooks or custom event handlers are running at the moment—just the default

AgentSession

start/listen cycle.

To gently manage post-speech silence in your

AgentSession

, here are a few friendly approaches you might find helpful: • Detect User Inactivity: You can use the

user_away_timeout

parameter to automatically notice when the user becomes inactive. For example:

session = AgentSession(user_away_timeout=12.5)  # Timeout in seconds

When the user’s state changes to "away," the

user_state_changed

event will be triggered. You can listen for this event to send a gentle follow-up message or take other thoughtful actions:

@session.on("user_state_changed")
def _user_state_changed(ev: UserStateChangedEvent):
    if ev.new_state == "away":
        asyncio.create_task(user_presence_task())

• Send Follow-Up Messages: Consider creating a task that checks if the user is still present and sends polite prompts to keep the conversation going. For example:

async def user_presence_task():
    for _ in range(3):
        await session.generate_reply(instructions="Are you still there?")
        await asyncio.sleep(10)
    await session.aclose()  # Close session after three gentle checks

• Gracefully End the Session: You might want to set up a background task that monitors user interaction and ends the session kindly after a period of silence. For example:

SILENCE_THRESHOLD = 5  # seconds

async def monitor_interaction():
    while True:
        if time.time() - user_last_spoke_time > SILENCE_THRESHOLD:
            await ctx.room.disconnect()
            break
        await asyncio.sleep(1)

These friendly methods can help you handle post-speech silence smoothly—either by re-engaging your user or ending the session with care. If you’d like, I’m here to help you implement these solutions or answer any questions you have along the way! Sources: Agent Testing & Evaluation | livekit/livekit_composite | DeepWiki | livekit-agents/livekit/agents/voice/agent_activity.py | Events and error handling | LiveKit Docs | Events and error handling | LiveKit Docs | Speech & Audio Processing | livekit/livekit_composite | DeepWiki