[Refactor] Consolidate sequence normalization and enc-dec parsing (#33928)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/engine/protocol.py

@@ -16,6 +16,7 @@ from vllm.outputs import PoolingRequestOutput, RequestOutput
 from vllm.plugins.io_processors import IOProcessor
 from vllm.pooling_params import PoolingParams
 from vllm.renderers import BaseRenderer
+from vllm.renderers.inputs import DictPrompt, TokPrompt
 from vllm.sampling_params import SamplingParams
 from vllm.tasks import SupportedTask
 from vllm.v1.engine import EngineCoreRequest
@@ -53,7 +54,11 @@ class EngineClient(ABC):
     @abstractmethod
     def generate(
         self,
-        prompt: EngineCoreRequest | PromptType | AsyncGenerator[StreamingInput, None],
+        prompt: EngineCoreRequest
+        | PromptType
+        | DictPrompt
+        | TokPrompt
+        | AsyncGenerator[StreamingInput, None],
         sampling_params: SamplingParams,
         request_id: str,
         *,
@@ -70,7 +75,7 @@ class EngineClient(ABC):
     @abstractmethod
     def encode(
         self,
-        prompt: PromptType,
+        prompt: PromptType | DictPrompt | TokPrompt,
         pooling_params: PoolingParams,
         request_id: str,
         lora_request: LoRARequest | None = None,