Fix: Enable /v1/models endpoint for pure diffusion mode (#805)

Signed-off-by: majiayu000 <1835304752@qq.com>

tests/entrypoints/openai_api/test_image_server.py

@@ -168,7 +168,13 @@ def test_client(mock_async_diffusion):
     app.state.engine_client = mock_async_diffusion
     app.state.diffusion_engine = mock_async_diffusion  # Also set for health endpoint
     app.state.stage_configs = [SimpleNamespace(stage_type="diffusion")]
-    app.state.diffusion_model_name = "Qwen/Qwen-Image"  # For models endpoint
+    from vllm.entrypoints.openai.models.protocol import BaseModelPath
+
+    from vllm_omni.entrypoints.openai.api_server import _DiffusionServingModels
+
+    app.state.openai_serving_models = _DiffusionServingModels(
+        [BaseModelPath(name="Qwen/Qwen-Image", model_path="Qwen/Qwen-Image")]
+    )
     app.state.args = Namespace(
         default_sampling_params='{"0": {"num_inference_steps":4, "guidance_scale":7.5}}',
         max_generated_image_size=4096,  # 64*64

vllm_omni/entrypoints/openai/api_server.py

@@ -1207,38 +1207,16 @@ _remove_route_from_router(router, "/v1/models")
 
 @router.get("/v1/models")
 async def show_available_models(raw_request: Request) -> JSONResponse:
-    """Show available models endpoint that works for both LLM and diffusion modes.
+    """Show available models for both LLM and diffusion modes.
 
-    Returns model information in OpenAI-compatible format.
+    Delegates to state.openai_serving_models which is set to either
+    OpenAIServingModels (LLM) or _DiffusionServingModels (pure diffusion).
     """
-    # Check if we're in diffusion mode
-    diffusion_model_name = getattr(raw_request.app.state, "diffusion_model_name", None)
-    if diffusion_model_name is not None:
-        # Diffusion mode - return the loaded model
-        return JSONResponse(
-            content={
-                "object": "list",
-                "data": [
-                    {
-                        "id": diffusion_model_name,
-                        "object": "model",
-                        "created": 0,
-                        "owned_by": "vllm-omni",
-                        "permission": [],
-                    }
-                ],
-            }
-        )
-
-    # LLM mode - delegate to openai_serving_models
-    openai_serving_models = getattr(raw_request.app.state, "openai_serving_models", None)
-    if openai_serving_models is not None:
-        models = await openai_serving_models.show_available_models()
+    handler = getattr(raw_request.app.state, "openai_serving_models", None)
+    if handler is not None:
+        models = await handler.show_available_models()
         return JSONResponse(content=models.model_dump())
-
-    return JSONResponse(
-        content={"object": "list", "data": []},
-    )
+    return JSONResponse(content={"object": "list", "data": []})
 
 
 # Image generation API endpoints