diff --git a/vllm_omni/entrypoints/cli/__init__.py b/vllm_omni/entrypoints/cli/__init__.py
new file mode 100644
index 000000000..2df4defa7
--- /dev/null
+++ b/vllm_omni/entrypoints/cli/__init__.py
@@ -0,0 +1,5 @@
+"""CLI helpers for vLLM-omni entrypoints."""
+
+from .serve import OmniServeCommand
+
+__all__ = ["OmniServeCommand"]
diff --git a/vllm_omni/entrypoints/cli/main.py b/vllm_omni/entrypoints/cli/main.py
index bb2447f3c..830e181a9 100644
--- a/vllm_omni/entrypoints/cli/main.py
+++ b/vllm_omni/entrypoints/cli/main.py
@@ -5,7 +5,7 @@ CLI entry point for vLLM-omni that intercepts vLLM commands.
 import sys
 import argparse
 from typing import List, Optional
-from vllm_omni.entrypoints.omni import OmniServeCommand
+from vllm_omni.entrypoints.cli.serve import OmniServeCommand
 
 
 def main():
diff --git a/vllm_omni/entrypoints/omni.py b/vllm_omni/entrypoints/cli/serve.py
similarity index 99%
rename from vllm_omni/entrypoints/omni.py
rename to vllm_omni/entrypoints/cli/serve.py
index 55f2b9db0..1fdc0f79a 100644
--- a/vllm_omni/entrypoints/omni.py
+++ b/vllm_omni/entrypoints/cli/serve.py
@@ -6,7 +6,7 @@ import argparse
 import asyncio
 from typing import List, Optional
 
-from .omni_llm import AsyncOmniLLM
+from ..omni_llm import AsyncOmniLLM
 from vllm_omni.config import (
     DiTConfig,
     create_ar_stage_config,