ci: pin sglang to v0.5.10rc0 for proto compat

e2e_test/chat_completions/test_openai_server.py

@@ -11,7 +11,6 @@
 import logging
 
 import pytest
-from infra import is_sglang
 
 logger = logging.getLogger(__name__)
 
@@ -392,8 +391,6 @@ def test_chat_completion_stream(self, model, api_client, logprobs, parallel_samp
         super().test_chat_completion_stream(model, api_client, logprobs, parallel_sample_num)
 
     def test_stop_sequences_stream(self, model, api_client):
-        if is_sglang():
-            self.STOP_SEQUENCE_TRIMMED = True
         super().test_stop_sequences_stream(model, api_client)
 
     @pytest.mark.skip(reason="gpt-oss models don't support regex constraints")

grpc_servicer/pyproject.toml

@@ -31,7 +31,7 @@ classifiers = [
 
 [project.optional-dependencies]
 vllm = ["vllm>=0.17.0"]
-sglang = ["sglang>=0.5.9"]
+sglang = ["sglang>=0.5.10rc0"]
 
 [project.urls]
 Homepage = "https://github.com/lightseekorg/smg"

grpc_servicer/smg_grpc_servicer/sglang/request_manager.py

@@ -35,7 +35,8 @@
     real_time,
 )
 from sglang.srt.server_args import PortArgs, ServerArgs
-from sglang.srt.utils import get_or_create_event_loop, get_zmq_socket, kill_process_tree
+from sglang.srt.utils import get_or_create_event_loop, kill_process_tree
+from sglang.srt.utils.network import get_zmq_socket
 from sglang.utils import get_exception_traceback
 
 logger = logging.getLogger(__name__)

grpc_servicer/smg_grpc_servicer/sglang/servicer.py

@@ -11,7 +11,7 @@
 import os
 import time
 from collections.abc import AsyncIterator
-from datetime import UTC, datetime
+from datetime import datetime, timezone
 
 import grpc
 import msgspec
@@ -537,7 +537,7 @@ async def GetLoads(
         loads = [_convert_loads_to_protobuf(r) for r in results]
 
         return sglang_scheduler_pb2.GetLoadsResponse(
-            timestamp=datetime.now(UTC).isoformat(),
+            timestamp=datetime.now(timezone.utc).isoformat(),
             version=sglang.__version__,
             dp_rank_count=len(loads),
             loads=loads,

ruff.toml

@@ -11,6 +11,8 @@ max-line-length = 120
 [lint.per-file-ignores]
 "e2e_test/**" = ["F841", "E402"]
 "bindings/python/tests/**" = ["F841", "E402"]
+# grpc_servicer supports Python 3.10+ — don't upgrade to 3.11+ syntax (e.g. datetime.UTC)
+"grpc_servicer/**" = ["UP017"]
 
 [format]
 # Black-compatible defaults

scripts/ci_install_sglang.sh

@@ -40,8 +40,13 @@ else
 fi
 
 # Install SGLang with all dependencies
+# TODO(changsu): Revert to `uv pip install "sglang[all]"` once sglang releases
+# a stable version (>= 0.5.10) that includes the smg_grpc_servicer migration
+# (sgl-project/sglang#20478). The rc0 is needed because the old built-in
+# grpc_server.py references proto types (EmbedError, EmbedComplete) that were
+# removed from our proto definitions.
 echo "Installing SGLang..."
-uv pip install "sglang[all]"
+uv pip install --prerelease=allow "sglang[all]==0.5.10rc0"
 
 # Install flashinfer-jit-cache: sglang bundles flashinfer_python but only for attention ops.
 # Multi-GPU models need trtllm_comm kernels (fused allreduce + layernorm) which FlashInfer