chore: import upstream snapshot with attribution

python/ray/llm/tests/serve/benchmark/test_smoke.py

@@ -0,0 +1,60 @@
+"""Tests for benchmark smoke mode."""
+
+import sys
+import types
+
+import aiohttp
+import pytest
+
+from ray.llm._internal.serve.benchmark.http_client import send_chat_completion
+from ray.llm._internal.serve.benchmark.models import TurnResult
+from ray.llm._internal.serve.benchmark.runners import run_smoke
+
+
+@pytest.mark.asyncio
+async def test_send_chat_completion(mock_server: str) -> None:
+    """send_chat_completion should parse SSE stream and return correct TurnResult."""
+    messages = [{"role": "user", "content": "hello"}]
+    async with aiohttp.ClientSession() as session:
+        result = await send_chat_completion(
+            session=session,
+            base_url=mock_server,
+            model="test-model",
+            messages=messages,
+            first_chunk_threshold=2,
+        )
+
+    assert isinstance(result, TurnResult)
+    assert result.generated_text == "this is a test"
+    assert result.input_tokens == 10
+    assert result.output_tokens == 4
+    assert result.ttft_ms > 0
+    assert result.e2e_latency_ms > 0
+    assert result.fc_ms > 0
+    assert isinstance(result.itl_ms, float)
+
+
+def test_run_smoke(mock_server: str) -> None:
+    """run_smoke should return 0 and print JSON metrics."""
+    args = types.SimpleNamespace(
+        base_url=mock_server,
+        model="test-model",
+        first_chunk_threshold=2,
+    )
+    exit_code = run_smoke(args)
+    assert exit_code == 0
+
+
+def test_run_smoke_connection_error(free_port: int) -> None:
+    """run_smoke should return 1 when the server is unreachable."""
+    args = types.SimpleNamespace(
+        base_url=f"http://127.0.0.1:{free_port}",
+        model="test-model",
+        first_chunk_threshold=16,
+    )
+    exit_code = run_smoke(args)
+    assert exit_code == 1
+
+
+if __name__ == "__main__":
+    sys.exit(pytest.main(["-v", __file__]))