Hi there! When working with LiveKit 1.x, you have several ways to detect if a user has interrupted an agent. Here are some helpful methods you can use: • The

interrupted

flag in the

ChatMessage

object, which gets set whenever an interruption happens. • The

agent_speech_interrupted

event, which is triggered when the agent is interrupted while speaking. • The

agent_state_changed

event, where the state changes from "speaking" to "listening" once an interruption occurs. • The

PlaybackFinishedEvent

, where the

interrupted

field will be

True

if the playback was interrupted. On top of that, you can customize how interruptions behave by adjusting the

AgentSession

parameters to fit your needs: • allow_interruptions: Set this to

False

if you want to disable interruptions (by default, it’s

True

). • min_interruption_duration: This sets the minimum speech duration in seconds required to trigger an interruption (default is 0.5 seconds). • min_interruption_words: This defines the minimum number of transcribed words needed to consider it an interruption (default is 0). And if you want to take manual control, you can always call the

session.interrupt()

method to immediately stop any ongoing or queued speech. Sources: livekit-agents/livekit/agents/voice/agent_session.py | Turn detection and interruptions | LiveKit Docs | VoicePipelineAgent | LiveKit Docs | VoicePipelineAgent | LiveKit Docs