@langwatch/scenario
    Preparing search index...

    Class AdapterCapabilities

    Declaration of what a voice adapter can and cannot do.

    readonly everywhere — the Python @dataclass(frozen=True) analogue in TS. Concrete adapters declare this once as a class-level constant.

    Index

    Constructors

    Properties

    dtmf: boolean

    True if the adapter can transmit DTMF tones (telephony).

    inputFormats: readonly string[]

    Wire formats the adapter accepts from the SDK for outgoing user audio (e.g. ["pcm16/24000", "mulaw/8000"]).

    interruption: boolean

    True if the adapter can send a first-class interrupt signal to the agent under test (e.g. Twilio clear, OpenAI Realtime response.cancel). When True, scenario.interrupt() uses the signal path; when False, it falls back to timing-based barge-in (audio sent over the wire while the agent is speaking, which the SUT detects via VAD).

    nativeVad: boolean

    True if the adapter itself provides voice-activity-detection events (user_start_speaking / user_stop_speaking). When False, the SDK falls back to webrtcvad on the incoming audio stream.

    outputFormats: readonly string[]

    Wire formats the adapter emits for incoming agent audio. The SDK converts these to internal PCM16/24000 mono.

    streamingTranscripts: boolean

    True if the adapter emits incremental transcript updates as the agent speaks. Required for interrupt(afterWords: N).