min_interruption_words

in LiveKit Agents v1.x controls how many words must be spoken (with speech-to-text enabled) for the system to treat user speech as an intentional interruption of the agent. • Configuration: Set

min_interruption_words

when creating an `AgentSession`—for example:`from livekit.agents import AgentSession session = AgentSession( min_interruption_words=3, # Requires at least 3 words to interrupt # ...other params )` • How It Works: ◦ The system first checks if speech duration exceeds

min_interruption_duration

(default 0.5s). ◦ If STT is enabled and

min_interruption_words

is set (>0), it checks the current transcript’s word count using

split_words(text, split_character=True)

. ◦ If the transcript has fewer words than this value, interruption won’t trigger. • Notes: ◦ This parameter only has an effect if STT is enabled. ◦ The default is 0 (any speech can interrupt if duration threshold is met). ◦ Setting a higher value can reduce accidental interruptions by brief utterances, but may introduce latency awaiting transcription. This allows you to tune interruption sensitivity when running agents in version 1.x. Sources: livekit-agents/livekit/agents/voice/agent.py