[Realtime API] Adds minimal realtime API based on websockets (#33187)

Signed-off-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Nick Hill <nickhill123@gmail.com>

vllm/v1/worker/gpu_model_runner.py

@@ -68,6 +68,7 @@ from vllm.model_executor.models.interfaces import (
     supports_eagle3,
     supports_mrope,
     supports_multimodal_pruning,
+    supports_realtime,
     supports_transcription,
     supports_xdrope,
 )
@@ -2541,6 +2542,9 @@ class GPUModelRunner(
 
             supported_tasks.append("transcription")
 
+        if supports_realtime(model):
+            supported_tasks.append("realtime")
+
         return supported_tasks
 
     def get_supported_pooling_tasks(self) -> list[PoolingTask]: