[Bugfix] Force continuous usage stats when CLI override is enabled (#37923)

Signed-off-by: Your Name <you@example.com>
Co-authored-by: Your Name <you@example.com>
Co-authored-by: OpenCode <noreply@openai.com>

tests/entrypoints/test_utils.py

@@ -3,7 +3,12 @@
 
 import pytest
 
-from vllm.entrypoints.utils import get_max_tokens, sanitize_message
+from vllm.entrypoints.openai.engine.protocol import StreamOptions
+from vllm.entrypoints.utils import (
+    get_max_tokens,
+    sanitize_message,
+    should_include_usage,
+)
 
 
 def test_sanitize_message():
@@ -13,6 +18,25 @@ def test_sanitize_message():
     )
 
 
+@pytest.mark.parametrize(
+    ("stream_options", "expected"),
+    [
+        (None, (True, True)),
+        (StreamOptions(include_usage=False), (True, True)),
+        (
+            StreamOptions(include_usage=False, continuous_usage_stats=False),
+            (True, True),
+        ),
+        (
+            StreamOptions(include_usage=True, continuous_usage_stats=False),
+            (True, True),
+        ),
+    ],
+)
+def test_should_include_usage_force_enables_continuous_usage(stream_options, expected):
+    assert should_include_usage(stream_options, True) == expected
+
+
 class TestGetMaxTokens:
     """Tests for get_max_tokens() to ensure generation_config's max_tokens
     acts as a default when from model author, and as a ceiling when