fix(responses): enforce spec object on completion bridge (#26327)

Ensure Chat Completions -> Responses bridge always emits object="response" so non-native providers return the same top-level schema as native OpenAI Responses.

Made-with: Cursor

litellm/responses/litellm_completion_transformation/transformation.py

@@ -1652,7 +1652,7 @@ class LiteLLMCompletionResponsesConfig:
             id=chat_completion_response.id,
             created_at=chat_completion_response.created,
             model=chat_completion_response.model,
-            object=chat_completion_response.object,
+            object="response",
             error=getattr(chat_completion_response, "error", None),
             incomplete_details=getattr(
                 chat_completion_response, "incomplete_details", None

tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py

@@ -369,6 +369,7 @@ class TestLiteLLMCompletionResponsesConfig:
         ]
         assert len(message_items) == 1, "Should have exactly one message item"
         assert message_items[0].content[0].text == "Just a regular answer."
+        assert responses_api_response.object == "response"
 
     def test_transform_chat_completion_response_multiple_choices_with_reasoning(self):
         """Test that only reasoning from first choice is included when multiple choices exist"""