From d5d84d52c182e720ccb419b25cab18aa8da5f878 Mon Sep 17 00:00:00 2001
From: Niklas Semmler <niklas@langfuse.com>
Date: Tue, 10 Nov 2026 18:29:09 +0100
Subject: [PATCH] test(live-provider): request stream usage from OpenAI in
 streaming tests

The Langfuse server stopped inferring token usage from content
(langfuse/langfuse#19802). OpenAI streams only report usage when
stream_options.include_usage is set, so these tests got empty
usage_details and failed with KeyError: 'input'.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 tests/live_provider/test_decorators.py | 1 +
 tests/live_provider/test_openai.py     | 5 +++++
 2 files changed, 6 insertions(+)

diff --git a/tests/live_provider/test_decorators.py b/tests/live_provider/test_decorators.py
index 87f9097e7..1e732023c 100644
--- a/tests/live_provider/test_decorators.py
+++ b/tests/live_provider/test_decorators.py
@@ -140,6 +140,7 @@ async def level_2_function():
             temperature=0,
             metadata={"someKey": "someResponse"},
             stream=True,
+            stream_options={"include_usage": True},
         )
 
         with propagate_attributes(
diff --git a/tests/live_provider/test_openai.py b/tests/live_provider/test_openai.py
index 6021046ef..c11993dc6 100644
--- a/tests/live_provider/test_openai.py
+++ b/tests/live_provider/test_openai.py
@@ -104,6 +104,7 @@ def test_openai_chat_completion_stream(openai):
         temperature=0,
         metadata={"someKey": "someResponse"},
         stream=True,
+        stream_options={"include_usage": True},
     )
 
     assert iter(completion)
@@ -159,6 +160,7 @@ def test_openai_chat_completion_stream_with_next_iteration(openai):
         temperature=0,
         metadata={"someKey": "someResponse"},
         stream=True,
+        stream_options={"include_usage": True},
     )
 
     assert iter(completion)
@@ -664,6 +666,7 @@ async def test_async_chat_stream(openai):
         model="gpt-4o-mini",
         name=generation_name,
         stream=True,
+        stream_options={"include_usage": True},
     )
 
     async for c in completion:
@@ -711,6 +714,7 @@ async def test_async_chat_stream_with_anext(openai):
         model="gpt-4o-mini",
         name=generation_name,
         stream=True,
+        stream_options={"include_usage": True},
     )
 
     result = ""
@@ -1100,6 +1104,7 @@ async def test_close_async_stream(openai):
         model="gpt-4o-mini",
         name=generation_name,
         stream=True,
+        stream_options={"include_usage": True},
     )
 
     async for token in stream: