@langwatch/scenario
    Preparing search index...

    Variable PCM16_SAMPLE_RATEConst

    PCM16_SAMPLE_RATE: 24000

    AudioChunk — the canonical internal audio representation.

    Per the AudioChunk normalization locked decision (Python parity: see python/scenario/voice/audio_chunk.py), every piece of audio flowing through the SDK is PCM16 @ 24kHz mono at the framework boundary. Adapters convert to/from their transport-native format at the send/recv edge.

    This keeps the combinatorial complexity of N adapters x M formats collapsed to N conversions at the adapter edge.