From 276b5695681ac4f545362df9cf4127b2e1f92788 Mon Sep 17 00:00:00 2001
From: Tom Ron <rontom@gmail.com>
Date: Fri, 2 Oct 2026 11:18:57 +0300
Subject: [PATCH] fix(responses): emit response.incomplete on replayed streams
 and default temperature to 0

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
---
 .../transformation.py                         |  4 ++--
 litellm/responses/streaming_iterator.py       |  8 +++++++
 .../test_incomplete_status_transformation.py  | 21 +++++++++++++++++++
 3 files changed, 31 insertions(+), 2 deletions(-)

diff --git a/litellm/responses/litellm_completion_transformation/transformation.py b/litellm/responses/litellm_completion_transformation/transformation.py
index a922e95912d..2b050c56098 100644
--- a/litellm/responses/litellm_completion_transformation/transformation.py
+++ b/litellm/responses/litellm_completion_transformation/transformation.py
@@ -243,7 +243,7 @@ class LiteLLMCompletionResponsesConfig:
             "tools": tools,
             "top_p": responses_api_request.get("top_p"),
             "user": responses_api_request.get("user"),
-            "temperature": responses_api_request.get("temperature"),
+            "temperature": responses_api_request.get("temperature", 0),
             "parallel_tool_calls": responses_api_request.get("parallel_tool_calls"),
             "max_tokens": responses_api_request.get("max_output_tokens"),
             "stream": stream,
@@ -1625,7 +1625,7 @@ class LiteLLMCompletionResponsesConfig:
                 responses_api_request=responses_api_request,
             ),
             parallel_tool_calls=getattr(chat_completion_response, "parallel_tool_calls", False),
-            temperature=responses_api_request.get("temperature"),
+            temperature=responses_api_request.get("temperature", 0),
             tool_choice=getattr(chat_completion_response, "tool_choice", "auto"),
             tools=getattr(chat_completion_response, "tools", []),
             top_p=responses_api_request.get("top_p"),
diff --git a/litellm/responses/streaming_iterator.py b/litellm/responses/streaming_iterator.py
index dab666ff0d9..5844c34eaa2 100644
--- a/litellm/responses/streaming_iterator.py
+++ b/litellm/responses/streaming_iterator.py
@@ -1253,6 +1253,14 @@ def _build_synthetic_response_events(
             )
         )
 
+    if transformed.status == "incomplete":
+        events.append(
+            openai_types.ResponseIncompleteEvent(
+                type=openai_types.ResponsesAPIStreamEvents.RESPONSE_INCOMPLETE,
+                response=transformed,
+            )
+        )
+        return events
     events.append(
         openai_types.ResponseCompletedEvent(
             type=openai_types.ResponsesAPIStreamEvents.RESPONSE_COMPLETED,
diff --git a/tests/test_litellm/responses/litellm_completion_transformation/test_incomplete_status_transformation.py b/tests/test_litellm/responses/litellm_completion_transformation/test_incomplete_status_transformation.py
index 92785b52849..4ae7b8c75a5 100644
--- a/tests/test_litellm/responses/litellm_completion_transformation/test_incomplete_status_transformation.py
+++ b/tests/test_litellm/responses/litellm_completion_transformation/test_incomplete_status_transformation.py
@@ -8,6 +8,7 @@ from litellm.responses.litellm_completion_transformation.streaming_iterator impo
 from litellm.responses.litellm_completion_transformation.transformation import (
     LiteLLMCompletionResponsesConfig,
 )
+from litellm.responses.streaming_iterator import _build_synthetic_response_events
 from litellm.types.llms.openai import ResponsesAPIStreamEvents
 from litellm.types.utils import Choices, Message, ModelResponse, Usage
 
@@ -70,3 +71,23 @@ def test_stream_terminal_event_follows_status(finish_reason, event_type):
     event = iterator._emit_response_completed_event(_chat_response(finish_reason))
     assert event is not None
     assert event.type == event_type
+
+
+@pytest.mark.parametrize(
+    "finish_reason, event_type",
+    [
+        ("length", ResponsesAPIStreamEvents.RESPONSE_INCOMPLETE),
+        ("stop", ResponsesAPIStreamEvents.RESPONSE_COMPLETED),
+    ],
+)
+def test_replayed_stream_terminal_event_follows_status(finish_reason, event_type):
+    events = _build_synthetic_response_events(
+        transformed=_transform(finish_reason, {}),
+        logging_obj=None,
+        chunk_size=10,
+    )
+    assert events[-1].type == event_type
+
+
+def test_omitted_temperature_defaults_to_zero():
+    assert _transform("stop", {}).temperature == 0