[router][CI] Clean up imports and prints statements in sgl-router/py_test (#12024)

sgl-router/py_test/e2e_grpc/basic/test_openai_server.py

@@ -13,14 +13,11 @@ Run with:
 """
 
 import json
-
-# CHANGE: Import router launcher instead of server launcher
 import sys
 import unittest
 from pathlib import Path
 
 import openai
-import requests
 
 _TEST_DIR = Path(__file__).parent
 sys.path.insert(0, str(_TEST_DIR.parent))
@@ -225,7 +222,6 @@ class TestOpenAIServer(CustomTestCase):
         try:
             js_obj = json.loads(text)
         except (TypeError, json.decoder.JSONDecodeError):
-            print("JSONDecodeError", text)
             raise
         assert isinstance(js_obj["name"], str)
         assert isinstance(js_obj["population"], int)

sgl-router/py_test/e2e_grpc/conftest.py

@@ -7,7 +7,7 @@ This module provides shared fixtures that can be used across all gRPC router tes
 import sys
 from pathlib import Path
 
-import pytest
+import pytest  # noqa: F401
 
 # Ensure router py_src is importable
 _ROUTER_ROOT = Path(__file__).resolve().parents[2]

sgl-router/py_test/e2e_grpc/features/test_enable_thinking.py

@@ -6,17 +6,11 @@ python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinki
 python3 -m unittest openai_server.features.test_enable_thinking.TestEnableThinking.test_stream_chat_completion_without_reasoning
 """
 
-import asyncio
 import json
-import os
 import sys
-import time
 import unittest
-
-# CHANGE: Import router launcher instead of server launcher
 from pathlib import Path
 
-import openai
 import requests
 
 _TEST_DIR = Path(__file__).parent
@@ -24,10 +18,8 @@ sys.path.insert(0, str(_TEST_DIR.parent))
 from fixtures import popen_launch_workers_and_router
 from util import (
     DEFAULT_ENABLE_THINKING_MODEL_PATH,
-    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
-    get_tokenizer,
     kill_process_tree,
 )
 
@@ -131,7 +123,6 @@ class TestEnableThinking(CustomTestCase):
         has_reasoning = False
         has_content = False
 
-        print("\n=== Stream With Reasoning ===")
         for line in response.iter_lines():
             if line:
                 line = line.decode("utf-8")
@@ -176,7 +167,6 @@ class TestEnableThinking(CustomTestCase):
         has_reasoning = False
         has_content = False
 
-        print("\n=== Stream Without Reasoning ===")
         for line in response.iter_lines():
             if line:
                 line = line.decode("utf-8")

sgl-router/py_test/e2e_grpc/features/test_reasoning_content.py

@@ -9,15 +9,11 @@ python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningC
 python3 -m unittest openai_server.features.test_reasoning_content.TestReasoningContentStartup.test_streaming
 """
 
-import json
-
-# CHANGE: Import router launcher instead of server launcher
 import sys
 import unittest
 from pathlib import Path
 
 import openai
-import requests
 
 _TEST_DIR = Path(__file__).parent
 sys.path.insert(0, str(_TEST_DIR.parent))

sgl-router/py_test/e2e_grpc/fixtures.py

@@ -8,6 +8,7 @@ This module provides fixtures for launching SGLang workers and gRPC router separ
 This approach gives more control and matches production deployment patterns.
 """
 
+import logging
 import socket
 import subprocess
 import time
@@ -15,6 +16,8 @@ from typing import Optional
 
 import requests
 
+logger = logging.getLogger(__name__)
+
 
 def find_free_port() -> int:
     """Find an available port on localhost."""
@@ -56,9 +59,11 @@ def wait_for_workers_ready(
             attempt += 1
             elapsed = int(time.time() - start_time)
 
-            # Print progress every 10 seconds
+            # Log progress every 10 seconds
             if elapsed > 0 and elapsed % 10 == 0 and attempt % 10 == 0:
-                print(f"  Still waiting for workers... ({elapsed}/{timeout}s elapsed)")
+                logger.info(
+                    f"  Still waiting for workers... ({elapsed}/{timeout}s elapsed)"
+                )
 
             try:
                 response = session.get(
@@ -69,7 +74,7 @@ def wait_for_workers_ready(
                     total_workers = data.get("total", 0)
 
                     if total_workers == expected_workers:
-                        print(
+                        logger.info(
                             f"  All {expected_workers} workers connected after {elapsed}s"
                         )
                         return
@@ -161,14 +166,14 @@ def popen_launch_workers_and_router(
     else:
         router_port = find_free_port()
 
-    print(f"\n{'='*70}")
-    print(f"Launching gRPC cluster (separate workers + router)")
-    print(f"{'='*70}")
-    print(f"  Model: {model}")
-    print(f"  Router port: {router_port}")
-    print(f"  Workers: {num_workers}")
-    print(f"  TP size: {tp_size}")
-    print(f"  Policy: {policy}")
+    logger.info(f"\n{'='*70}")
+    logger.info(f"Launching gRPC cluster (separate workers + router)")
+    logger.info(f"{'='*70}")
+    logger.info(f"  Model: {model}")
+    logger.info(f"  Router port: {router_port}")
+    logger.info(f"  Workers: {num_workers}")
+    logger.info(f"  TP size: {tp_size}")
+    logger.info(f"  Policy: {policy}")
 
     # Step 1: Launch workers with gRPC enabled
     workers = []
@@ -179,9 +184,9 @@ def popen_launch_workers_and_router(
         worker_url = f"grpc://127.0.0.1:{worker_port}"
         worker_urls.append(worker_url)
 
-        print(f"\n[Worker {i+1}/{num_workers}]")
-        print(f"  Port: {worker_port}")
-        print(f"  URL: {worker_url}")
+        logger.info(f"\n[Worker {i+1}/{num_workers}]")
+        logger.info(f"  Port: {worker_port}")
+        logger.info(f"  URL: {worker_url}")
 
         # Build worker command
         worker_cmd = [
@@ -226,17 +231,19 @@ def popen_launch_workers_and_router(
             )
 
         workers.append(worker_proc)
-        print(f"  PID: {worker_proc.pid}")
+        logger.info(f"  PID: {worker_proc.pid}")
 
     # Give workers a moment to start binding to ports
     # The router will check worker health when it starts
-    print(f"\nWaiting for {num_workers} workers to initialize (20s)...")
+    logger.info(f"\nWaiting for {num_workers} workers to initialize (20s)...")
     time.sleep(20)
 
     # Quick check: make sure worker processes are still alive
     for i, worker in enumerate(workers):
         if worker.poll() is not None:
-            print(f"  ✗ Worker {i+1} died during startup (exit code: {worker.poll()})")
+            logger.error(
+                f"  ✗ Worker {i+1} died during startup (exit code: {worker.poll()})"
+            )
             # Cleanup: kill all workers
             for w in workers:
                 try:
@@ -245,12 +252,14 @@ def popen_launch_workers_and_router(
                     pass
             raise RuntimeError(f"Worker {i+1} failed to start")
 
-    print(f"✓ All {num_workers} workers started (router will verify connectivity)")
+    logger.info(
+        f"✓ All {num_workers} workers started (router will verify connectivity)"
+    )
 
     # Step 2: Launch router pointing to workers
-    print(f"\n[Router]")
-    print(f"  Port: {router_port}")
-    print(f"  Worker URLs: {', '.join(worker_urls)}")
+    logger.info(f"\n[Router]")
+    logger.info(f"  Port: {router_port}")
+    logger.info(f"  Worker URLs: {', '.join(worker_urls)}")
 
     # Build router command
     router_cmd = [
@@ -284,7 +293,7 @@ def popen_launch_workers_and_router(
         router_cmd.extend(router_args)
 
     if show_output:
-        print(f"  Command: {' '.join(router_cmd)}")
+        logger.info(f"  Command: {' '.join(router_cmd)}")
 
     # Launch router
     if show_output:
@@ -296,19 +305,19 @@ def popen_launch_workers_and_router(
             stderr=subprocess.PIPE,
         )
 
-    print(f"  PID: {router_proc.pid}")
+    logger.info(f"  PID: {router_proc.pid}")
 
     # Wait for router to be ready
     router_url = f"http://127.0.0.1:{router_port}"
-    print(f"\nWaiting for router to start at {router_url}...")
+    logger.info(f"\nWaiting for router to start at {router_url}...")
 
     try:
         wait_for_workers_ready(
             router_url, expected_workers=num_workers, timeout=180, api_key=api_key
         )
-        print(f"✓ Router ready at {router_url}")
+        logger.info(f"✓ Router ready at {router_url}")
     except TimeoutError:
-        print(f"✗ Router failed to start")
+        logger.error(f"✗ Router failed to start")
         # Cleanup: kill router and all workers
         try:
             router_proc.kill()
@@ -321,11 +330,11 @@ def popen_launch_workers_and_router(
                 pass
         raise
 
-    print(f"\n{'='*70}")
-    print(f"✓ gRPC cluster ready!")
-    print(f"  Router: {router_url}")
-    print(f"  Workers: {len(workers)}")
-    print(f"{'='*70}\n")
+    logger.info(f"\n{'='*70}")
+    logger.info(f"✓ gRPC cluster ready!")
+    logger.info(f"  Router: {router_url}")
+    logger.info(f"  Workers: {len(workers)}")
+    logger.info(f"{'='*70}\n")
 
     return {
         "workers": workers,

sgl-router/py_test/e2e_grpc/function_call/test_openai_function_calling.py

@@ -13,10 +13,7 @@ Run with:
 """
 
 import json
-
-# CHANGE: Import router launcher instead of server launcher
 import sys
-import time
 import unittest
 from pathlib import Path
 

sgl-router/py_test/e2e_grpc/function_call/test_tool_choice.py

@@ -8,8 +8,6 @@ Tests: required, auto, and specific function choices in both streaming and non-s
 """
 
 import json
-
-# CHANGE: Import router launcher instead of server launcher
 import sys
 import unittest
 from pathlib import Path

sgl-router/py_test/e2e_grpc/util.py

@@ -8,6 +8,7 @@ Extracted and adapted from:
 - sglang.test.test_utils (constants and CustomTestCase)
 """
 
+import logging
 import os
 import signal
 import threading
@@ -17,6 +18,8 @@ from typing import Optional, Union
 
 import psutil
 
+logger = logging.getLogger(__name__)
+
 try:
     from transformers import (
         AutoTokenizer,
@@ -204,8 +207,8 @@ def get_tokenizer(
         raise RuntimeError(err_msg) from e
 
     if not isinstance(tokenizer, PreTrainedTokenizerFast):
-        print(
-            f"Warning: Using a slow tokenizer. This might cause a performance "
+        logger.warning(
+            f"Using a slow tokenizer. This might cause a performance "
             f"degradation. Consider using a fast tokenizer instead."
         )
 
@@ -245,14 +248,10 @@ class CustomTestCase(unittest.TestCase):
                 return super(CustomTestCase, self)._callTestMethod(method)
             except Exception as e:
                 if attempt < max_retry:
-                    print(
+                    logger.info(
                         f"Test failed on attempt {attempt + 1}/{max_retry + 1}, retrying..."
                     )
                     continue
                 else:
                     # Last attempt, re-raise the exception
                     raise
-
-    def setUp(self):
-        """Print test method name at the start of each test."""
-        print(f"[Test Method] {self._testMethodName}", flush=True)

sgl-router/py_test/e2e_grpc/validation/test_large_max_new_tokens.py

@@ -3,8 +3,6 @@ python3 -m unittest openai_server.validation.test_large_max_new_tokens.TestLarge
 """
 
 import os
-
-# CHANGE: Import router launcher instead of server launcher
 import sys
 import time
 import unittest
@@ -104,7 +102,6 @@ class TestLargeMaxNewTokens(CustomTestCase):
                 self.stderr.flush()
                 lines = open(STDERR_FILENAME).readlines()
                 for line in lines[pt:]:
-                    print(line, end="", flush=True)
                     if f"#running-req: {num_requests}" in line:
                         all_requests_running = True
                         pt = -1

sgl-router/py_test/e2e_grpc/validation/test_openai_server_ignore_eos.py

@@ -12,7 +12,6 @@ Run with:
     pytest py_test/e2e_grpc/e2e_grpc/validation/test_openai_server_ignore_eos.py -v
 """
 
-# CHANGE: Import router launcher instead of server launcher
 import sys
 from pathlib import Path