Skip to content

vllm.entrypoints.serve.instrumentator.server_info

PydanticVllmConfig module-attribute 

PydanticVllmConfig = TypeAdapter(VllmConfig)

logger module-attribute 

logger = init_logger(__name__)

router module-attribute 

router = APIRouter()

attach_router 

attach_router(app: FastAPI)
Source code in vllm/entrypoints/serve/instrumentator/server_info.py 
37
38
39
40
def attach_router(app: FastAPI):
    if not envs.VLLM_SERVER_DEV_MODE:
        return
    app.include_router(router)

show_server_info async 

show_server_info(
    raw_request: Request,
    config_format: Annotated[
        Literal["text", "json"], Query()
    ] = "text",
)
Source code in vllm/entrypoints/serve/instrumentator/server_info.py 
22
23
24
25
26
27
28
29
30
31
32
33
34
@router.get("/server_info")
async def show_server_info(
    raw_request: Request,
    config_format: Annotated[Literal["text", "json"], Query()] = "text",
):
    vllm_config: VllmConfig = raw_request.app.state.vllm_config
    server_info = {
        "vllm_config": str(vllm_config)
        if config_format == "text"
        else PydanticVllmConfig.dump_python(vllm_config, mode="json", fallback=str)
        # fallback=str is needed to handle e.g. torch.dtype
    }
    return JSONResponse(content=server_info)