Pull the first audio chunk out of a ModelMessage-shaped object.
Recognizes the canonical AI-SDK file part (mediaType: "audio/*"), and
— for robustness at the transition / adapter edge — the legacy OpenAI
input_audio and alternate audio conventions. Picks up a transcript
from a sibling text part. Returns null if no audio part is present.
Pull the first audio chunk out of a ModelMessage-shaped object.
Recognizes the canonical AI-SDK
filepart (mediaType: "audio/*"), and — for robustness at the transition / adapter edge — the legacy OpenAIinput_audioand alternateaudioconventions. Picks up a transcript from a siblingtextpart. Returnsnullif no audio part is present.