fix(openai): skip realtime truncate when no audio was played fixes : #6157

livekit-plugins/livekit-plugins-openai/livekit/plugins/openai/realtime/realtime_model.py

@@ -1615,14 +1615,25 @@ def truncate(
         audio_transcript: NotGivenOr[str] = NOT_GIVEN,
     ) -> None:
         if "audio" in modalities:
-            self.send_event(
-                ConversationItemTruncateEvent(
-                    type="conversation.item.truncate",
-                    content_index=0,
-                    item_id=message_id,
-                    audio_end_ms=audio_end_ms,
+            if audio_end_ms > 0:
+                self.send_event(
+                    ConversationItemTruncateEvent(
+                        type="conversation.item.truncate",
+                        content_index=0,
+                        item_id=message_id,
+                        audio_end_ms=audio_end_ms,
+                    )
                 )
-            )
+            else:
+                remote_ids = {item.id for item in self._remote_chat_ctx.to_chat_ctx().items}
+                if message_id in remote_ids:
+                    self.send_event(
+                        ConversationItemDeleteEvent(
+                            type="conversation.item.delete",
+                            item_id=message_id,
+                            event_id=utils.shortuuid("chat_ctx_delete_"),
+                        )
+                    )
         elif utils.is_given(audio_transcript):
             # sync the forwarded text to the remote chat ctx
             chat_ctx = self.chat_ctx.copy(

tests/test_openai_realtime_chat_ctx.py

@@ -54,6 +54,67 @@ async def test_update_chat_ctx_keeps_existing_remote_empty_messages() -> None:
     assert session._sent_events == []
 
 
+def test_truncate_deletes_item_when_no_audio_played() -> None:
+    """Interrupting before any audio frame plays (audio_end_ms == 0) must not
+    send a conversation.item.truncate (the Realtime API rejects it with
+    "Only model output audio messages can be truncated"). When the server holds
+    the generated-but-unplayed item, it is deleted instead so it is not left
+    dangling in the remote chat ctx."""
+    from openai.types.realtime import ConversationItemDeleteEvent
+
+    session = _create_session()
+    # server holds the unplayed assistant audio message
+    session._remote_chat_ctx.insert(
+        None, llm.ChatMessage(role="assistant", content=["hi"], id="item_1")
+    )
+
+    session.truncate(
+        message_id="item_1",
+        modalities=["audio", "text"],
+        audio_end_ms=0,
+    )
+
+    assert len(session._sent_events) == 1
+    event = session._sent_events[0]
+    assert isinstance(event, ConversationItemDeleteEvent)
+    assert event.item_id == "item_1"
+
+
+def test_truncate_noop_when_no_audio_and_item_not_on_server() -> None:
+    """If audio_end_ms == 0 and the item was never committed to the server, send
+    nothing: there is no audio to truncate and deleting a non-existent item would
+    itself surface an error."""
+    session = _create_session()
+
+    session.truncate(
+        message_id="missing-item",
+        modalities=["audio", "text"],
+        audio_end_ms=0,
+    )
+
+    assert session._sent_events == []
+
+
+def test_truncate_sends_event_when_audio_played() -> None:
+    """When audio has actually played (audio_end_ms > 0) the truncate event is
+    still sent for audio modalities."""
+    from openai.types.realtime import ConversationItemTruncateEvent
+
+    session = _create_session()
+
+    session.truncate(
+        message_id="item_1",
+        modalities=["audio", "text"],
+        audio_end_ms=500,
+    )
+
+    assert len(session._sent_events) == 1
+    event = session._sent_events[0]
+    assert isinstance(event, ConversationItemTruncateEvent)
+    assert event.item_id == "item_1"
+    assert event.audio_end_ms == 500
+
+
 def test_response_done_handles_string_status_details(monkeypatch) -> None:
     session = _create_session()
     session._realtime_model = types.SimpleNamespace(_provider_label="xAI")