[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/engine/async_llm.py

@@ -7,7 +7,6 @@ import time
 import warnings
 from collections.abc import AsyncGenerator, Iterable, Mapping
 from copy import copy
-from dataclasses import dataclass
 from typing import Any
 
 import torch
@@ -19,6 +18,7 @@ from vllm.engine.arg_utils import AsyncEngineArgs
 from vllm.engine.protocol import EngineClient
 from vllm.entrypoints.utils import _validate_truncation_size
 from vllm.inputs import PromptType
+from vllm.inputs.data import StreamingInput
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
 from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
@@ -53,18 +53,6 @@ from vllm.v1.metrics.stats import IterationStats
 logger = init_logger(__name__)
 
 
-@dataclass
-class StreamingInput:
-    """Input data for a streaming generation request.
-
-    This is used with generate() to support multi-turn streaming sessions
-    where inputs are provided via an async generator.
-    """
-
-    prompt: PromptType
-    sampling_params: SamplingParams | None = None
-
-
 class InputStreamError(Exception):
     """Wrapper for errors from the input stream generator.