feat: sync latest ARR implementation

tests/test_agent_service.py

@@ -80,7 +80,14 @@ class AgentServiceTests(unittest.TestCase):
         client = FakeAgentClient()
         with SQLiteSessionStore(":memory:") as store:
             service = OpenAgentService(client, store)  # type: ignore[arg-type]
-            events = list(service.stream_message("conversation-001", "你好", message_id="message-001"))
+            events = list(
+                service.stream_message(
+                    "conversation-001",
+                    "你好",
+                    message_id="message-001",
+                    include_trace=True,
+                )
+            )
             status = service.get_run("conversation-001", "run-001")
             cancelled = service.cancel_run("conversation-001", "run-001")
             deleted = service.forget_session("conversation-001")
@@ -91,6 +98,7 @@ class AgentServiceTests(unittest.TestCase):
         self.assertEqual(status["status"], "completed")
         self.assertEqual(cancelled["status"], "cancelling")
         self.assertTrue(deleted)
+        self.assertTrue(client.stream_calls[0][2]["include_trace"])
         self.assertEqual(
             client.run_calls,
             [