module-attribute ¶
CHUNK_METADATA_KEYS: frozenset[str] = frozenset(
{
"audio_text_total_chars",
"duplex_epoch",
"duplex_turn_id",
"llm_output_text_utf8",
"segment_end",
"tts_is_last_chunk",
"turn_end",
}
)
drain_delta_payload(payload: MultimodalPayload) -> None
Remove client-facing delta data while retaining request-level state.
is_non_final_delta_audio_chunk(
payload: MultimodalPayload, mm_type: str | None
) -> bool
Return whether an audio delta explicitly declares more chunks.
replace_snapshot_keys(
accumulated: MultimodalPayload,
incoming: MultimodalPayload,
) -> None
Replace per-chunk metadata with the latest values.