Here’s a friendly way to add a delay-based prompt right after your user finishes speaking: 1. Detect the End of Speech: Keep an eye on the

END_OF_SPEECH

event from LiveKit’s STT component. This handy event lets you know when the user has stopped talking, based on the

min_endpointing_delay

you set up. 2. Introduce a Delay: You can tweak the delay by adjusting the

min_endpointing_delay

parameter in your agent constructor. This controls how long the system waits before deciding the user’s done speaking. 3. Use the Callback: The

on_user_turn_completed

callback is your friend here—it lets you run some code right after the user’s turn wraps up. Perfect for adding that little pause before you jump in. 4. Trigger TTS Prompt: Once the delay is over, use your TTS system to gently ask, “Are you still there?”—a nice way to keep the conversation flowing. By playing around with the delay and keeping an eye on metrics like

end_of_utterance_delay

, you’ll create a smooth, natural experience that feels really responsive to your users. Sources: AI voice agents | LiveKit Docs | Capturing metrics | LiveKit Docs | AI Agents Framework | livekit/livekit_composite | DeepWiki