A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 05:00 UTC
[Misc] support collect_env for endpoint /server_info (#33246)
Signed-off-by: yang.xiao <yang.xiao@daocloud.io>
Xiao Yang committed
Feb 1, 2026 at 01:42 UTC
2238a12c13b9a1e27f132473a0fff1b4b233b1dc
1 file changed
+9
vllm/entrypoints/serve/instrumentator/server_info.py
+9
@@ -2,6 +2,8 @@
2
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3
4
5
+import asyncio
6
+import functools
7
from typing import Annotated, Literal
8
9
import pydantic
@@ -9,6 +11,7 @@ from fastapi import APIRouter, FastAPI, Query, Request
11
from fastapi.responses import JSONResponse
12
13
import vllm.envs as envs
14
+from vllm.collect_env import get_env_info
15
from vllm.config import VllmConfig
16
from vllm.logger import init_logger
17
@@ -32,6 +35,11 @@ def _get_vllm_env_vars():
35
return vllm_envs
36
37
38
+@functools.lru_cache(maxsize=1)
39
+def _get_system_env_info_cached():
40
+ return get_env_info()._asdict()
41
+
42
+
43
@router.get("/server_info")
44
async def show_server_info(
45
raw_request: Request,
@@ -46,6 +54,7 @@ async def show_server_info(
54
),
55
# fallback=str is needed to handle e.g. torch.dtype
56
"vllm_env": _get_vllm_env_vars(),
57
+ "system_env": await asyncio.to_thread(_get_system_env_info_cached),
58
}
59
return JSONResponse(content=server_info)
60