[Refactor][OutputProcessor 2/8]: OmniRequestOutput should inherit RequestOutput and no nested wrap-up (#5146)

Signed-off-by: Boao Shi <aoibosh@connect.hku.hk>
Signed-off-by: bowieshi <aoibosh@connect.hku.hk>
Signed-off-by: princepride <wangzhipeng628@gmail.com>
Signed-off-by: R2-Y <ruiruyang2@gmail.com>
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
Signed-off-by: Shi Boao Bowie <aoibosh@connect.hku.hk>
Co-authored-by: 汪志鹏 <wangzhipeng628@gmail.com>
Co-authored-by: ruirui(rein) yang <73573651+R2-Y@users.noreply.github.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>
Co-authored-by: Gao Han <hgaoaf@connect.ust.hk>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>

tests/entrypoints/openai_api/test_duplex_handler.py

@@ -2156,12 +2156,12 @@ def test_direct_listen_decision_survives_inner_completion_metadata():
     )
     assert decision is not None
     output = attach_duplex_output_decision(
-        OmniRequestOutput(
+        OmniRequestOutput.from_stage_output(
+            inner_output,
             request_id="duplex-direct-listen",
             finished=True,
             stage_id=0,
             final_output_type=decision.final_output_type,
-            request_output=inner_output,
             metrics={
                 "stage_metrics": {
                     "0": {
@@ -3049,12 +3049,12 @@ async def test_minicpmo_auto_response_listen_without_response_does_not_defer_com
     )
     assert decision is not None
     listen_output = attach_duplex_output_decision(
-        OmniRequestOutput(
+        OmniRequestOutput.from_stage_output(
+            inner_output,
             request_id=request_id,
             finished=True,
             stage_id=0,
             final_output_type=decision.final_output_type,
-            request_output=inner_output,
         ),
         decision,
     )

tests/entrypoints/openai_api/test_image_server.py

@@ -119,7 +119,6 @@ class MockGenerationResult:
 
     def __init__(self, images):
         self.images = images
-        self.request_output = SimpleNamespace(images=images)
         self.stage_durations = {}
         self.peak_memory_mb = 0.0
 
@@ -137,10 +136,7 @@ class MockStageResult:
             outputs = [SimpleNamespace(text=text, index=0)]
         else:
             outputs = []
-        self.request_output = SimpleNamespace(
-            outputs=outputs,
-            images=self.images,
-        )
+        self.outputs = outputs
         self.stage_durations = {}
         self.peak_memory_mb = 0.0
 
@@ -2080,22 +2076,20 @@ def test_extract_images_from_result():
     assert all(isinstance(img, Image.Image) for img in images)
     assert all(img.size == (64, 64) for img in images)
 
-    # Test dict path: result.request_output["images"]
+    # Test result with "images" attribute set in __init__
     class DictRequestOutput:
         def __init__(self):
-            self.request_output = {"images": [np.random.randint(0, 255, (64, 64, 3), dtype=np.uint8)]}
+            self.images = [np.random.randint(0, 255, (64, 64, 3), dtype=np.uint8)]
 
     result = DictRequestOutput()
     images = _extract_images_from_result(result)
     assert len(images) == 1
     assert isinstance(images[0], Image.Image)
 
-    # Test attribute path: result.request_output.images
+    # Test result with "images" attribute from an inner object
     class AttrRequestOutput:
         def __init__(self):
-            self.request_output = type(
-                "obj", (), {"images": [np.random.randint(0, 255, (32, 32, 3), dtype=np.uint8)]}
-            )()
+            self.images = [np.random.randint(0, 255, (32, 32, 3), dtype=np.uint8)]
 
     result = AttrRequestOutput()
     images = _extract_images_from_result(result)

tests/entrypoints/openai_api/test_serving_speech.py

@@ -180,10 +180,10 @@ def create_mock_audio_output_for_test(
     audio_tensor = torch.sin(torch.linspace(0, 440 * 2 * torch.pi, num_samples))
     mock_request_output = MockRequestOutput(request_id=request_id, audio_tensor=audio_tensor)
 
-    return OmniRequestOutput(
+    return OmniRequestOutput.from_stage_output(
+        mock_request_output,
         stage_id=0,
         final_output_type="audio",
-        request_output=mock_request_output,
     )
 
 
@@ -2292,10 +2292,10 @@ class TestStreamingResponse:
                     self.prompt_logprobs = None
                     self.kv_transfer_params = None
 
-            return OmniRequestOutput(
+            return OmniRequestOutput.from_stage_output(
+                MockRequestOutput(audio_tensor=chunk),
                 stage_id=0,
                 final_output_type="audio",
-                request_output=MockRequestOutput(audio_tensor=chunk),
                 finished=finished,
             )
 
@@ -3441,10 +3441,10 @@ class TestWAVStreaming:
                     self.prompt_logprobs = None
                     self.kv_transfer_params = None
 
-            return OmniRequestOutput(
+            return OmniRequestOutput.from_stage_output(
+                MockRequestOutput(audio_tensor=chunk),
                 stage_id=0,
                 final_output_type="audio",
-                request_output=MockRequestOutput(audio_tensor=chunk),
                 finished=finished,
             )
 

tests/entrypoints/openai_api/test_serving_video_stream.py

@@ -47,7 +47,7 @@ def _text_result(text: str) -> OmniRequestOutput:
     output.text = text
     request_output = RequestOutput()
     request_output.outputs = [output]
-    return OmniRequestOutput(final_output_type="text", request_output=request_output)
+    return OmniRequestOutput.from_stage_output(request_output, final_output_type="text")
 
 
 def _audio_result(audio_data: Any) -> OmniRequestOutput:
@@ -61,7 +61,7 @@ def _audio_result(audio_data: Any) -> OmniRequestOutput:
     output.multimodal_output = {"audio": audio_data}
     request_output = RequestOutput()
     request_output.outputs = [output]
-    return OmniRequestOutput(final_output_type="audio", request_output=request_output)
+    return OmniRequestOutput.from_stage_output(request_output, final_output_type="audio")
 
 
 class MockWebSocket: