[Entrypoint] Add realtime OpenPI robot serving API (#3673)

Signed-off-by: Yangshen Deng <yangshen.d@outlook.com>
Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
Co-authored-by: Meng <meng_chen99@163.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

vllm_omni/entrypoints/openai/api_server.py

@@ -133,6 +133,7 @@ from vllm_omni.entrypoints.openai.storage import STORAGE_MANAGER
 from vllm_omni.entrypoints.openai.stores import VIDEO_STORE, VIDEO_TASKS
 from vllm_omni.entrypoints.openai.utils import get_stage_type, parse_lora_request
 from vllm_omni.entrypoints.openai.video_api_utils import decode_input_reference
+from vllm_omni.entrypoints.openpi.serving import ServingRealtimeRobotOpenPI
 from vllm_omni.inputs.data import OmniDiffusionSamplingParams, OmniTextPrompt
 
 logger = init_logger(__name__)
@@ -633,6 +634,10 @@ async def omni_init_app_state(
         )
         state.openai_streaming_speech = None
         state.openai_streaming_video = None
+        state.openai_serving_realtime_robot = ServingRealtimeRobotOpenPI.create_policy_server(
+            engine_client=engine_client,
+            model_name=model_name,
+        )
 
         state.enable_server_load_tracking = getattr(args, "enable_server_load_tracking", False)
         state.server_load_metrics = 0
@@ -963,6 +968,7 @@ async def omni_init_app_state(
         model_name=served_model_names[0] if served_model_names else None,
         stage_configs=state.stage_configs,
     )
+    state.openai_serving_realtime_robot = None
 
     state.enable_server_load_tracking = args.enable_server_load_tracking
     state.server_load_metrics = 0
@@ -1406,6 +1412,23 @@ async def realtime_websocket(websocket: WebSocket):
     await connection.handle_connection()
 
 
+@router.websocket("/v1/realtime/robot/openpi")
+async def realtime_robot_openpi(websocket: WebSocket):
+    """WebSocket endpoint for robot policy inference via OpenPI messages."""
+    from vllm_omni.entrypoints.openpi.connection import (
+        RobotRealtimeConnection,
+    )
+
+    serving = getattr(websocket.app.state, "openai_serving_realtime_robot", None)
+    if serving is None:
+        await websocket.accept()
+        await websocket.send_json({"type": "error", "error": "Robot policy not available", "code": "unsupported"})
+        await websocket.close()
+        return
+    connection = RobotRealtimeConnection(websocket, serving)
+    await connection.handle_connection()
+
+
 # Health and Model endpoints for diffusion mode