Python: keep an in-progress continuation_token on the aggregated response

python/packages/core/agent_framework/_types.py

@@ -2236,7 +2236,12 @@ def _process_update(response: ChatResponse | AgentResponse, update: ChatResponse
         and update.finish_reason is not None
     ):
         response.finish_reason = update.finish_reason
-    response.continuation_token = update.continuation_token
+    if update.continuation_token is not None:
+        response.continuation_token = update.continuation_token
+    elif update.finish_reason is not None:
+        # A terminal update withdraws the token: its presence on the aggregate
+        # means the operation is still in progress.
+        response.continuation_token = None
 
 
 def _merge_function_call_content(message: Message, content: Content) -> None:

python/packages/core/tests/core/test_types.py

@@ -1889,6 +1889,40 @@ def test_agent_run_response_from_updates_uses_last_non_none_agent_id() -> None:
     assert response.agent_id == "source-agent"
 
 
+def test_chat_response_from_updates_preserves_continuation_token_while_in_progress() -> None:
+    # A token stamped on an earlier update must survive later updates that do not
+    # repeat it: presence on the aggregate means the operation is still in progress.
+    response = ChatResponse.from_updates([
+        ChatResponseUpdate(
+            contents=[Content.from_text(text="working")],
+            response_id="resp_1",
+            model="gpt-5",
+            continuation_token={"response_id": "resp_1"},
+        ),
+        ChatResponseUpdate(contents=[Content.from_text(text="...still working")]),
+    ])
+
+    assert response.response_id == "resp_1"
+    assert response.model == "gpt-5"
+    assert response.continuation_token == {"response_id": "resp_1"}
+
+
+def test_chat_response_from_updates_clears_continuation_token_on_terminal_update() -> None:
+    # The update that carries a finish_reason is terminal, so it withdraws the token
+    # from the aggregate: "None means the operation is complete".
+    response = ChatResponse.from_updates([
+        ChatResponseUpdate(
+            contents=[Content.from_text(text="working")],
+            response_id="resp_1",
+            continuation_token={"response_id": "resp_1"},
+        ),
+        ChatResponseUpdate(contents=[Content.from_text(text="done")], finish_reason="stop"),
+    ])
+
+    assert response.finish_reason == "stop"
+    assert response.continuation_token is None
+
+
 def test_agent_run_response_str_method(chat_message: Message) -> None:
     response = AgentResponse(messages=chat_message)
     assert str(response) == "Hello"