Add "/server_info" endpoint in api_server to retrieve the vllm_config. (#16572)

Signed-off-by: Xihui Cang <xihuicang@gmail.com>

vllm/engine/protocol.py

@@ -5,7 +5,7 @@ from abc import ABC, abstractmethod
 from typing import AsyncGenerator, List, Mapping, Optional
 
 from vllm.beam_search import BeamSearchSequence, create_sort_beams_key_function
-from vllm.config import DecodingConfig, ModelConfig
+from vllm.config import DecodingConfig, ModelConfig, VllmConfig
 from vllm.core.scheduler import SchedulerOutputs
 from vllm.inputs.data import PromptType, TokensPrompt
 from vllm.inputs.parse import is_explicit_encoder_decoder_prompt
@@ -220,6 +220,11 @@ class EngineClient(ABC):
         """
         ...
 
+    @abstractmethod
+    async def get_vllm_config(self) -> VllmConfig:
+        """Get the vllm configuration of the vLLM engine."""
+        ...
+
     @abstractmethod
     async def get_model_config(self) -> ModelConfig:
         """Get the model configuration of the vLLM engine."""