@langwatch/scenario
    Preparing search index...

    Type Alias InterruptionStrategy

    InterruptionStrategy: "contextual" | "random_phrase"

    Interruption configuration for proceed({ interruptions }).

    Source §4.4 L478-492. Two strategies:

    • "contextual": an LLM generates a short interruption phrase from the running conversation context.
    • "random_phrase": draw from a canned phrase list.

    Python parity: python/scenario/voice/interruption.py.

    The proposal does not supply the contextual LLM prompt — implementer-level decision. We expose a short system prompt focused on realistic user interjections so downstream callers can reuse it as-is or override it.