@langwatch/scenario
    Preparing search index...

    Interface AudioSegment

    A contiguous span of audio attributed to one speaker.

    transcriptTruncated is true when an agent segment was cut short by a user_interrupt event during the run — the audio bytes are authoritative; the transcript may reflect what the agent INTENDED to say, not what the user actually heard.

    interface AudioSegment {
        audio: Uint8Array;
        endTime: number;
        speaker: SpeakerRole;
        startTime: number;
        transcript?: string;
        transcriptTruncated?: boolean;
    }
    Index

    Properties

    audio: Uint8Array

    PCM16 little-endian bytes, mono, 24kHz.

    endTime: number
    speaker: SpeakerRole
    startTime: number
    transcript?: string
    transcriptTruncated?: boolean