diff --git a/docker/Dockerfile b/docker/Dockerfile
index 69e9ba13540..779a2fa5d17 100644
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@@ -3,10 +3,10 @@
 #
 # 2. radixark/miles:dev-cu13-arm64
 #    build-arg:ENABLE_CUDA_13=1 \
-#    build-arg:SGLANG_IMAGE_TAG=v0.5.9-cu130-arm64 \
+#    build-arg:SGLANG_IMAGE_TAG=v0.5.10-cu130 \
 #    build-arg:WHEELS_TAG=cu130-aarch64 \
 
-ARG SGLANG_IMAGE_TAG=v0.5.9
+ARG SGLANG_IMAGE_TAG=v0.5.10
 FROM lmsysorg/sglang:${SGLANG_IMAGE_TAG} AS sglang
 
 # ======================================== Arguments =============================================
@@ -88,7 +88,7 @@ RUN pip install megatron-energon --no-deps
 RUN pip install multi-storage-client --no-deps
 
 COPY requirements.txt /tmp/requirements.txt
-RUN pip install -r /tmp/requirements.txt
+RUN rm -rf /usr/lib/python3/dist-packages/jwt /usr/lib/python3/dist-packages/PyJWT* && pip install -r /tmp/requirements.txt
 
 # https://github.com/pytorch/pytorch/issues/168167
 RUN if [ "${ENABLE_CUDA_13}" = "1" ]; then \
diff --git a/miles/rollout/session/session_server.py b/miles/rollout/session/session_server.py
index 0377117bdf9..bc2633350ea 100644
--- a/miles/rollout/session/session_server.py
+++ b/miles/rollout/session/session_server.py
@@ -36,7 +36,7 @@ def __init__(self, args, backend_url: str):
         )
 
         # Close the httpx connection pool when uvicorn shuts down to avoid FD leaks.
-        self.app.add_event_handler("shutdown", self.client.aclose)
+        self.app.router.on_shutdown.append(self.client.aclose)
 
         setup_session_routes(self.app, self, args)
 
diff --git a/miles/router/router.py b/miles/router/router.py
index 09be44b033f..51194be4cf1 100644
--- a/miles/router/router.py
+++ b/miles/router/router.py
@@ -36,7 +36,7 @@ def __init__(self, args, verbose=False):
         self.verbose = verbose
 
         self.app = FastAPI()
-        self.app.add_event_handler("startup", self._start_background_health_check)
+        self.app.router.on_startup.append(self._start_background_health_check)
 
         # URL -> Active Request Count (load state)
         self.worker_request_counts: dict[str, int] = {}
diff --git a/tests/fast/utils/chat_template_utils/test_template.py b/tests/fast/utils/chat_template_utils/test_template.py
index 225a9bf1783..39ac412954b 100644
--- a/tests/fast/utils/chat_template_utils/test_template.py
+++ b/tests/fast/utils/chat_template_utils/test_template.py
@@ -60,6 +60,7 @@ def _make_serving(tokenizer) -> OpenAIServingChat:
     serving.use_dpsk_v32_encoding = False
     serving.is_gpt_oss = False
     serving.tool_call_parser = None
+    serving.reasoning_parser = None
     return serving