feat: add benchmark serving (#657)

python/sglang/srt/openai_protocol.py

@@ -7,6 +7,23 @@ from pydantic import BaseModel, Field
 from typing_extensions import Literal
 
 
+class ModelCard(BaseModel):
+    """Model cards."""
+
+    id: str
+    object: str = "model"
+    created: int = Field(default_factory=lambda: int(time.time()))
+    owned_by: str = "sglang"
+    root: Optional[str] = None
+
+
+class ModelList(BaseModel):
+    """Model list consists of model cards."""
+
+    object: str = "list"
+    data: List[ModelCard] = []
+
+
 class ErrorResponse(BaseModel):
     object: str = "error"
     message: str

python/sglang/srt/server.py

@@ -44,6 +44,7 @@ from sglang.srt.openai_api_adapter import (
     v1_chat_completions,
     v1_completions,
 )
+from sglang.srt.openai_protocol import ModelCard, ModelList
 from sglang.srt.server_args import PortArgs, ServerArgs
 from sglang.srt.utils import (
     API_KEY_HEADER_NAME,
@@ -73,6 +74,21 @@ async def health() -> Response:
     return Response(status_code=200)
 
 
+def get_model_list():
+    """Available models."""
+    model_names = [tokenizer_manager.model_path]
+    return model_names
+
+
+@app.get("/v1/models")
+def available_models():
+    """Show available models."""
+    model_cards = []
+    for model_name in get_model_list():
+        model_cards.append(ModelCard(id=model_name, root=model_name))
+    return ModelList(data=model_cards)
+
+
 @app.get("/get_model_info")
 async def get_model_info():
     result = {