fix(streaming): keep the served service tier on the finish chunk the stream adds

litellm/litellm_core_utils/streaming_handler.py

@@ -1915,6 +1915,16 @@ class CustomStreamWrapper:
         model_response: Final = self.model_response_creator(
             hidden_params=None if self.received_model_id is None else {"received_model_id": self.received_model_id}
         )
+        served_tier: Final = next(
+            (
+                tier
+                for chunk in reversed(self.chunks)
+                if isinstance(tier := getattr(chunk, "service_tier", None), str) and tier
+            ),
+            None,
+        )
+        if served_tier is not None:
+            setattr(model_response, "service_tier", served_tier)  # noqa: B010  # pydantic extra, not a declared field
         _finish_reason: Final = self.received_finish_reason or self.intermittent_finish_reason
         if _finish_reason is not None:
             model_response.choices[0].finish_reason = _finish_reason

tests/unit/litellm_core_utils/test_streaming_handler.py

@@ -7547,6 +7547,38 @@ async def test_openai_stream_relays_the_served_service_tier_on_every_chunk_inclu
     assert relayed[-1]["usage"]["total_tokens"] == 11
 
 
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_stream_relays_the_served_service_tier_on_the_finish_chunk_it_adds(logging_obj: Logging, sync_mode: bool):
+    from litellm.utils import ModelResponseListIterator
+
+    def _chunk(content: str, finish_reason: str | None) -> ModelResponseStream:
+        return ModelResponseStream(
+            id="chatcmpl-tier",
+            created=1742056047,
+            model="gpt-4.1-mini",
+            choices=[StreamingChoices(finish_reason=finish_reason, index=0, delta=Delta(content=content))],
+            service_tier="priority",
+        )
+
+    logging_obj.update_environment_variables(
+        model="gpt-4.1-mini", optional_params={}, litellm_params={}, custom_llm_provider="openai"
+    )
+    wrapper = CustomStreamWrapper(
+        completion_stream=ModelResponseListIterator(model_responses=[_chunk("Hi", None), _chunk("!", "stop")]),
+        model="gpt-4.1-mini",
+        custom_llm_provider="openai",
+        logging_obj=logging_obj,
+    )
+
+    relayed = (
+        [chunk.model_dump() for chunk in wrapper] if sync_mode else [chunk.model_dump() async for chunk in wrapper]
+    )
+
+    assert [chunk["choices"][0]["finish_reason"] for chunk in relayed] == [None, None, "stop"], relayed
+    assert [chunk.get("service_tier") for chunk in relayed] == ["priority"] * len(relayed), relayed
+
+
 def _last_chunk_carries_finish_reason_wrapper(
     logging_obj: Logging, finish_reason: str, sync_stream: bool
 ) -> CustomStreamWrapper: