优化网络与系统检测为缓存优先，避免页面长时间停留在检测中。

将网络/IP 路由拆分并加入去重缓存，系统检测改为线程执行并支持强制刷新，前端改为先展示缓存快照后按需手动重测，降低阻塞与重复探测开销。

Co-authored-by: Cursor <cursoragent@cursor.com>

web2_host_caps.py

@@ -4,6 +4,8 @@ from __future__ import annotations
 import platform
 import socket
 import subprocess
+import threading
+import time
 from typing import Any
 
 
@@ -20,8 +22,10 @@ def detect_nvidia_gpu() -> bool:
         return False
 
 
-def detect_nvenc_available() -> bool:
-    if not detect_nvidia_gpu():
+def detect_nvenc_available(nvidia_present: bool | None = None) -> bool:
+    if nvidia_present is None:
+        nvidia_present = detect_nvidia_gpu()
+    if not nvidia_present:
         return False
     try:
         _kw = {"capture_output": True, "text": True, "timeout": 8}
@@ -33,13 +37,27 @@ def detect_nvenc_available() -> bool:
         return False
 
 
-def gpu_status_payload() -> dict[str, Any]:
+_GPU_STATUS_CACHE_TTL_SECONDS = 60.0
+_gpu_status_lock = threading.Lock()
+_gpu_status_cache: tuple[float, dict[str, Any]] | None = None
+
+
+def gpu_status_payload(force_refresh: bool = False) -> dict[str, Any]:
+    global _gpu_status_cache
+    now = time.time()
+    with _gpu_status_lock:
+        if not force_refresh and _gpu_status_cache and (now - _gpu_status_cache[0]) < _GPU_STATUS_CACHE_TTL_SECONDS:
+            return dict(_gpu_status_cache[1])
+
     nv = detect_nvidia_gpu()
-    enc = detect_nvenc_available()
-    return {
+    enc = detect_nvenc_available(nv)
+    payload = {
         "nvidia_detected": nv,
         "nvenc_available": enc,
     }
+    with _gpu_status_lock:
+        _gpu_status_cache = (time.time(), payload)
+    return dict(payload)
 
 
 def machine_summary_payload() -> dict[str, Any]: