[Feature] integrate Structural Tag in xgrammar backend for function calling (#3566)

Co-authored-by: shuaills <shishuaiuoe@gmail.com>

python/sglang/srt/openai_api/adapter.py

@@ -994,10 +994,17 @@ def v1_chat_generate_request(
             "ignore_eos": request.ignore_eos,
             "skip_special_tokens": request.skip_special_tokens,
         }
+
         if request.response_format and request.response_format.type == "json_schema":
             sampling_params["json_schema"] = convert_json_schema_to_str(
                 request.response_format.json_schema.schema_
             )
+        elif (
+            request.response_format and request.response_format.type == "structural_tag"
+        ):
+            sampling_params["structural_tag"] = convert_json_schema_to_str(
+                request.response_format.model_dump(by_alias=True)
+            )
         sampling_params_list.append(sampling_params)
 
         image_data_list.append(image_data)

python/sglang/srt/openai_api/protocol.py

@@ -258,6 +258,18 @@ class ResponseFormat(BaseModel):
     json_schema: Optional[JsonSchemaResponseFormat] = None
 
 
+class StructuresResponseFormat(BaseModel):
+    begin: str
+    schema_: Optional[Dict[str, object]] = Field(alias="schema", default=None)
+    end: str
+
+
+class StructuralTagResponseFormat(BaseModel):
+    type: Literal["structural_tag"]
+    structures: List[StructuresResponseFormat]
+    triggers: List[str]
+
+
 class Function(BaseModel):
     """Function descriptions."""
 
@@ -298,7 +310,7 @@ class ChatCompletionRequest(BaseModel):
     max_tokens: Optional[int] = None
     n: int = 1
     presence_penalty: float = 0.0
-    response_format: Optional[ResponseFormat] = None
+    response_format: Union[ResponseFormat, StructuralTagResponseFormat] = None
     seed: Optional[int] = None
     stop: Optional[Union[str, List[str]]] = None
     stream: bool = False