From ff9563eb17c307980f08560423632fee9ab65704 Mon Sep 17 00:00:00 2001
From: Stephen Chin <1290231+steveonjava@users.noreply.github.com>
Date: Thu, 13 Aug 2026 03:39:54 -0700
Subject: [PATCH] fix(responses): sync logging stream state

Providers such as ChatGPT can force a Responses API request to stream after the client requested a non-streaming chat completion. Record the resolved wire behavior on the shared logging object so chunk logging does not latch non-stream dedup state before the assembled response is logged.

Tests: uv run pytest tests/test_litellm/llms/custom_httpx/test_llm_http_handler.py tests/test_litellm/completion_extras/litellm_responses_transformation/test_completion_extras_litellm_responses_transformation_handler.py tests/test_litellm/llms/chatgpt/responses/test_chatgpt_responses_transformation.py -q
---
 litellm/llms/custom_httpx/llm_http_handler.py |  4 +++
 .../custom_httpx/test_llm_http_handler.py     | 28 ++++++++++++++++++-
 2 files changed, 31 insertions(+), 1 deletion(-)

diff --git a/litellm/llms/custom_httpx/llm_http_handler.py b/litellm/llms/custom_httpx/llm_http_handler.py
index 9eccfe12e71..56ed57daede 100644
--- a/litellm/llms/custom_httpx/llm_http_handler.py
+++ b/litellm/llms/custom_httpx/llm_http_handler.py
@@ -2423,6 +2423,8 @@ class BaseLLMHTTPHandler:
         if extra_body:
             data.update(extra_body)
         stream = bool(stream or data.get("stream"))
+        logging_obj.stream = stream
+        logging_obj.model_call_details.update(stream=stream)
 
         # Preserve the OpenAI-style request context (not sent to the provider) for streaming
         # hooks/metadata; the streaming iterator now consumes this to run deployment hooks
@@ -2611,6 +2613,8 @@ class BaseLLMHTTPHandler:
         if extra_body:
             data.update(extra_body)
         stream = bool(stream or data.get("stream"))
+        logging_obj.stream = stream
+        logging_obj.model_call_details.update(stream=stream)
 
         # Preserve the OpenAI-style request context (not sent to the provider) for streaming
         # hooks/metadata; the streaming iterator now consumes this to run deployment hooks
diff --git a/tests/unit/llms/custom_httpx/test_llm_http_handler.py b/tests/unit/llms/custom_httpx/test_llm_http_handler.py
index 399e4dbf206..4b2f054cf19 100644
--- a/tests/unit/llms/custom_httpx/test_llm_http_handler.py
+++ b/tests/unit/llms/custom_httpx/test_llm_http_handler.py
@@ -197,6 +197,10 @@ def test_prepare_fake_stream_request():
 
 
 def test_response_api_handler_streams_when_provider_transform_adds_stream():
+    from datetime import datetime
+
+    from litellm.litellm_core_utils.litellm_logging import Logging
+
     handler = BaseLLMHTTPHandler()
     config = Mock()
     config.validate_environment.return_value = {}
@@ -214,7 +218,22 @@ def test_response_api_handler_streams_when_provider_transform_adds_stream():
             request=httpx.Request("POST", "https://chatgpt.example.com/responses"),
         )
     )
-    logging_obj = Mock()
+    logging_obj = Logging(
+        model="gpt-5.3-codex",
+        messages=[{"role": "user", "content": "hi"}],
+        stream=False,
+        call_type="completion",
+        start_time=datetime.now(),
+        litellm_call_id="test-call",
+        function_id="test-function",
+    )
+    logging_obj.update_environment_variables(
+        model="gpt-5.3-codex",
+        user=None,
+        optional_params={"stream": False},
+        litellm_params={},
+        custom_llm_provider="chatgpt",
+    )
 
     handler.response_api_handler(
         model="gpt-5.3-codex",
@@ -229,6 +248,10 @@ def test_response_api_handler_streams_when_provider_transform_adds_stream():
 
     assert client.post.call_args.kwargs["stream"] is True
     assert client.post.call_args.kwargs["json"]["stream"] is True
+    assert logging_obj.stream is True
+    assert logging_obj.model_call_details["stream"] is True
+    logging_obj.has_run_logging(event_type="async_success")
+    assert logging_obj.should_run_logging(event_type="async_success") is True
 
 
 def test_response_api_handler_runs_agentic_hooks_in_sync_path(monkeypatch):
@@ -356,6 +379,7 @@ async def test_async_response_api_handler_streams_when_provider_transform_adds_s
         )
     )
     logging_obj = Mock()
+    logging_obj.model_call_details = {}
 
     await handler.async_response_api_handler(
         model="gpt-5.3-codex",
@@ -370,6 +394,8 @@ async def test_async_response_api_handler_streams_when_provider_transform_adds_s
 
     assert client.post.call_args.kwargs["stream"] is True
     assert client.post.call_args.kwargs["json"]["stream"] is True
+    assert logging_obj.stream is True
+    assert logging_obj.model_call_details["stream"] is True
 
 
 @pytest.mark.asyncio