[Fix] Reuse resolved pipeline configuration for orchestration defaults

Signed-off-by: Sy03 <1370724210@qq.com>

tests/engine/test_single_stage_mode.py

@@ -16,7 +16,7 @@ from pytest_mock import MockerFixture
 from vllm.v1.engine.utils import EngineZmqAddresses
 
 from vllm_omni.config.config_factory import StageConfigFactory
-from vllm_omni.config.stage_config import DuplexSessionRuntimeConfig
+from vllm_omni.config.stage_config import DuplexSessionRuntimeConfig, PipelineConfig, StagePipelineConfig
 from vllm_omni.engine.async_omni_engine import AsyncOmniEngine
 from vllm_omni.engine.stage_engine_core_client import StageEngineCoreClientBase
 from vllm_omni.engine.stage_engine_startup import (
@@ -378,6 +378,7 @@ class TestSingleStageModeDetection:
         stage_cfgs: list[Any] | None = None,
         resolved_config_path: str = "/fake/path",
         patch_deploy_config: bool = True,
+        resolved_pipeline: PipelineConfig | None = None,
         **kwargs: Any,
     ) -> AsyncOmniEngine:
         mock_stage_configs = stage_cfgs or [_make_stage_cfg(0)]
@@ -392,10 +393,19 @@ class TestSingleStageModeDetection:
                 "vllm_omni.engine.async_omni_engine.load_deploy_config",
                 return_value=SimpleNamespace(duplex_session=DuplexSessionRuntimeConfig()),
             )
+
+        def resolve_stage_configs(engine, *args, **kwargs):
+            if resolved_pipeline is not None:
+                engine._config_resolution = SimpleNamespace(
+                    pipeline_config=resolved_pipeline, config_path=resolved_config_path
+                )
+            return resolved_config_path, mock_stage_configs
+
         mocker.patch.object(
             AsyncOmniEngine,
             "_resolve_stage_configs",
-            return_value=(resolved_config_path, mock_stage_configs),
+            autospec=True,
+            side_effect=resolve_stage_configs,
         )
         mocker.patch.object(AsyncOmniEngine, "_bootstrap_orchestrator")
         mock_thread_cls = mocker.patch("threading.Thread")
@@ -420,6 +430,17 @@ class TestSingleStageModeDetection:
         )
         assert engine.single_stage_mode is True
 
+    @pytest.mark.parametrize("model_type, expected", [("qwen3_tts", True), ("qwen3_omni_moe", False)])
+    def test_event_driven_default_uses_resolved_pipeline(self, mocker, model_type, expected):
+        pipeline = PipelineConfig(
+            model_type=model_type,
+            stages=(StagePipelineConfig(stage_id=0, model_stage="a", final_output=True),),
+        )
+        engine = self._make_engine_no_thread(mocker, resolved_pipeline=pipeline)
+
+        StageConfigFactory.get_pipeline_config.assert_not_called()
+        assert engine._event_driven_orch_default is expected
+
     def test_stage_id_kwarg_promotes_to_single_stage_mode(self, mocker: MockerFixture):
         engine = self._make_engine_no_thread(
             mocker,

vllm_omni/engine/async_omni_engine.py

@@ -215,32 +215,6 @@ class AsyncOmniEngine:
         # retaining the richer resolver result for pipeline-wide settings.
         # Overrides of that private seam fall back to the factory below.
         deploy_config_path = kwargs.get("deploy_config")
-        # ``trust_remote_code`` is tri-state (bool | None): ``None`` means "not
-        # specified" so stage-config resolution can defer to the deploy yaml's
-        # per-stage value (see ``with_trust_remote_code_override``). The
-        # restriction path below loads the top-level HF config via vLLM's
-        # ``get_config``, which needs a real bool, so collapse ``None`` to the
-        # default ``False`` here (#5495).
-        pipeline_config = StageConfigFactory.get_pipeline_config(
-            model=model,
-            trust_remote_code=bool(trust_remote_code),
-            deploy_config_path=deploy_config_path,
-        )
-        self.endpoint_restrictions = pipeline_config.endpoint_restrictions if pipeline_config is not None else ()
-        self._duplex_runtime_extension_path = (
-            pipeline_config.duplex_runtime_extension if pipeline_config is not None else None
-        )
-        self.duplex_serving_adapter_path = (
-            pipeline_config.duplex_serving_adapter if pipeline_config is not None else None
-        )
-        self._duplex_control_enabled = bool(pipeline_config and pipeline_config.duplex_control_enabled)
-        self._event_driven_orch_default = _event_driven_orch_default_for_pipeline(
-            pipeline_config.model_type if pipeline_config is not None else None
-        )
-        self.duplex_session_config = DuplexSessionRuntimeConfig()
-        if deploy_config_path is not None:
-            self.duplex_session_config = load_deploy_config(deploy_config_path).duplex_session
-
         # Tri-state: None means "not specified" — the deploy yaml's per-stage
         # trust_remote_code stays in effect. An explicit True/False here is a
         # global override (precedence: caller > deploy yaml > default False);
@@ -1043,6 +1017,9 @@ class AsyncOmniEngine:
             pipeline_config.duplex_serving_adapter if pipeline_config is not None else None
         )
         self._duplex_control_enabled = bool(pipeline_config and pipeline_config.duplex_control_enabled)
+        self._event_driven_orch_default = _event_driven_orch_default_for_pipeline(
+            pipeline_config.model_type if pipeline_config is not None else None
+        )
         self.duplex_session_config = DuplexSessionRuntimeConfig()
         if config_path is not None:
             self.duplex_session_config = load_deploy_config(config_path).duplex_session