Audio referenced by content digest, the same digest form as ImageContentBlock.
audio entered SurfaceKind with native signed audio input: the bytes are a
user attachment recorded through the session spine, never model output. No
adapter emits this block; the assembler and recorder refuse it the way they
refuse image, and the shape exists so the closed-union proof below stays
total instead of being relaxed.
Audio referenced by content digest, the same digest form as ImageContentBlock.
audioenteredSurfaceKindwith native signed audio input: the bytes are a user attachment recorded through the session spine, never model output. No adapter emits this block; the assembler and recorder refuse it the way they refuseimage, and the shape exists so the closed-union proof below stays total instead of being relaxed.