[OAI] patch origin request_id logic (#7508)

python/sglang/srt/entrypoints/openai/protocol.py

@@ -196,6 +196,9 @@ class CompletionRequest(BaseModel):
     bootstrap_port: Optional[int] = None
     bootstrap_room: Optional[int] = None
 
+    # For request id
+    rid: Optional[Union[List[str], str]] = None
+
     @field_validator("max_tokens")
     @classmethod
     def validate_max_tokens_positive(cls, v):
@@ -430,8 +433,8 @@ class ChatCompletionRequest(BaseModel):
     stream_reasoning: bool = True
     chat_template_kwargs: Optional[Dict] = None
 
-    # The request id.
-    rid: Optional[str] = None
+    # For request id
+    rid: Optional[Union[List[str], str]] = None
 
     # For PD disaggregation
     bootstrap_host: Optional[str] = None
@@ -529,7 +532,7 @@ class EmbeddingRequest(BaseModel):
     user: Optional[str] = None
 
     # The request id.
-    rid: Optional[str] = None
+    rid: Optional[Union[List[str], str]] = None
 
 
 class EmbeddingObject(BaseModel):

python/sglang/srt/entrypoints/openai/serving_chat.py

@@ -95,6 +95,7 @@ class OpenAIServingChat(OpenAIServingBase):
             bootstrap_port=request.bootstrap_port,
             bootstrap_room=request.bootstrap_room,
             return_hidden_states=request.return_hidden_states,
+            rid=request.rid,
         )
 
         return adapted_request, request

python/sglang/srt/entrypoints/openai/serving_completions.py

@@ -87,6 +87,7 @@ class OpenAIServingCompletion(OpenAIServingBase):
             bootstrap_port=request.bootstrap_port,
             bootstrap_room=request.bootstrap_room,
             return_hidden_states=request.return_hidden_states,
+            rid=request.rid,
         )
 
         return adapted_request, request

python/sglang/srt/entrypoints/openai/serving_embedding.py

@@ -119,6 +119,7 @@ class OpenAIServingEmbedding(OpenAIServingBase):
 
         adapted_request = EmbeddingReqInput(
             **prompt_kwargs,
+            rid=request.rid,
         )
 
         return adapted_request, request