更新 README.md

- CONNECT 目标解析支持 IPv6 字面量：剥掉方括号再拆 host:port，[::1]:443 不再解析成 "[::1]"（直连/白名单/内网绕过判定失效的根因）
- 更新包校验加固：release 附件按 RRPP_Lite_<tag>.exe 名字精确匹配（不再固定取第一个）；下载后与 API 记录的 size 比对；发布流程附带 .sha256 边车时校验 SHA-256
- 上游代理认证注入按完整头名精确匹配：Proxy-Authorization 不再误判为已有 Authorization；注入位置紧跟请求行保证上游读到我们的认证
- 远程同步后节点按 settings 记录的域名重匹配：服务端调整节点顺序时不再静默连到别的节点；启动校验等待订阅同步完成（最多 15 秒）后再执行
- DNS 分流判定缓存加 300 秒 TTL：域名换 IP 后不再长期沿用旧的直连/隧道判断（_is_private_ip/_is_china_ip/_is_whitelisted 换 _ttl_cache，保留 cache_clear 兼容）
- is_proxy_on 状态判定同步收紧为本程序 PAC 地址（其他软件的 PAC 不再让状态灯误亮）

.gitignore

@@ -0,0 +1,2 @@
+__pycache__/
+*.pyc

README.md

@@ -3,9 +3,13 @@
 > common 版本现在有许多未修复的 bug (如部分 UI 的显示问题) 、令人难以忍受的代码逻辑 (如 PAC 规则直接写在代码里)和安全性问题 (认证凭据的暴露与证书忽略)，而 Lite 版本修复了这些问题，并会收到持续的更新 
 - 本程序完全没有 Linux 与 CLI 功能兼容，默认构建为带有图形界面的 Windows .exe
 
-
 ## 如何区分 common 版本与 Lite 版本
 - Lite 版本的版本号前四位固定为 9999，后续为从月份开始的构建时间戳，而 common 版本的版本号为包含年份的完整的构建时间戳
 
 ## 服务器侧配置手册 (部分内容过时)
-https://rrrosmontis.icu/articles/index.html?id=00005
+- https://rrrosmontis.icu/articles/index.html?id=00005
+- 更推荐参考本仓库 server_config 文件夹中的 md 文件手册
+
+## 若要直接使用仓库客户端对应的服务
+请前往 https://sso.rrrosmontis.icu 注册 Rosmanka ID，然后与管理员商定使用细则
+royirin@rrrosmontis.icu

lite/proxy_lite.py

@@ -102,6 +102,7 @@ def _load_nodes():
 
 _SELECTED_NODE = 0  # 当前选中节点下标（默认第一条）
 _SELECTED_RESTORED = False  # 启动时是否已按设置恢复过上次选中的节点
+_NODES_SYNC_DONE = threading.Event()  # 订阅同步线程完成信号（启动校验等它）
 
 def _current_node():
     """当前使用的节点：默认取列表第一个；无节点时返回 None。"""
@@ -146,12 +147,173 @@ def _save_current_node(domain):
     data["current_node"] = str(domain or "")
     _settings_save(data)
 
+# ── IPv6-only 节点可达性（客户端标注用）─────────────────────
+# IPv6-only 节点的域名只有 AAAA 记录（无 A 记录），本机没有 IPv6 出口时连不上。
+# 判定三级：ipv6_only 字段 → 服务端 remark 里的 "(IPv6)" 标记 → 域名解析兜底。
+# 服务端 subscription?method=rrpp_client 只吐旧扁平格式，ipv6_only 字段被丢掉，
+# 只留 remark 里的 "(IPv6)"（见后端 countries_to_rrpp_client），故第 2 级是现网主路径。
+
+_IPV6_OK = {}                        # domain -> 本机能否连上该节点的 IPv6 地址
+_IPV6_ONLY = {}                      # domain -> 该域名是否只有 IPv6 地址（解析兜底）
+_IPV6_PROBE_TIMEOUT = 4              # 单次 TCP 探测超时（秒）
+
+def _local_has_ipv6():
+    """本机是否有全局 IPv6 地址（排除回环/链路本地/唯一本地地址）。
+    没有则不可能连上 IPv6-only 节点，直接短路，省去探测。"""
+    try:
+        import socket
+        import ipaddress
+        for info in socket.getaddrinfo(socket.gethostname(), None, socket.AF_INET6):
+            addr = info[4][0].split("%")[0]  # 去掉 zone id（fe80::1%eth0）
+            try:
+                ip = ipaddress.ip_address(addr)
+            except ValueError:
+                continue
+            if ip.is_loopback or ip.is_link_local or ip.is_private:
+                continue
+            return True
+    except Exception:
+        return False
+    return False
+
+def _domain_ipv6_only(domain: str) -> bool:
+    """域名是否「只有 IPv6 地址」：解析结果里有 IPv6 且没有 IPv4 才算。
+    解析失败按 False（当普通节点处理，不误伤）。结果缓存。"""
+    if domain in _IPV6_ONLY:
+        return _IPV6_ONLY[domain]
+    import ipaddress
+    only = False
+    try:
+        addrs = _resolve_host_ips(domain)
+        if addrs:
+            has6 = any(ip.version == 6 for ip in addrs)
+            has4 = any(ip.version == 4 for ip in addrs)
+            only = has6 and not has4
+    except Exception:
+        only = False
+    _IPV6_ONLY[domain] = only
+    if only:
+        log(f"节点 {domain} 解析结果只有 IPv6 地址，按 IPv6 专用节点处理")
+    return only
+
+def _node_ipv6_only(node) -> bool:
+    """节点是否 IPv6 专用。
+
+    三级判定：
+      1. ipv6_only 字段（服务端新格式；rrpp_client 输出里没有）
+      2. remark 里的 "(IPv6)" 标记（rrpp_client 现网唯一的权威线索）
+      3. 域名解析兜底：解析结果只有 IPv6 地址、没有 IPv4
+    """
+    if not isinstance(node, dict):
+        return False
+    if node.get("ipv6_only"):
+        return True
+    if "ipv6" in str(node.get("remark") or "").lower():
+        return True
+    return _domain_ipv6_only(str(node.get("domain") or "").strip())
+
+def _node_http_port(node) -> int:
+    """节点的 HTTP 代理端口（不依赖当前选中节点，供可达性探测用）。"""
+    m = (node.get("methods") or {}).get("http") or {}
+    for v in (m.get("port"), node.get("port")):
+        if v not in (None, ""):
+            try:
+                return int(v)
+            except (TypeError, ValueError):
+                pass
+    return 443
+
+def _node_reachable(node):
+    """节点在本机是否可达：普通节点恒为 True；IPv6 专用节点需真的能建立 IPv6 连接
+    （按该节点自己的 HTTP 端口探测）。结果按域名缓存（探测一次）。"""
+    if not isinstance(node, dict):
+        return True
+    domain = str(node.get("domain") or "").strip()
+    if not _node_ipv6_only(node):
+        return True
+    if not domain:
+        return False
+    if domain in _IPV6_OK:
+        return _IPV6_OK[domain]
+    port = _node_http_port(node)
+    ok = False
+    addrs = sorted((ip for ip in _resolve_host_ips(domain) if ip.version == 6), key=str)
+    if addrs and _local_has_ipv6():
+        import socket
+        for ip in addrs:
+            try:
+                s = socket.socket(socket.AF_INET6, socket.SOCK_STREAM)
+                s.settimeout(_IPV6_PROBE_TIMEOUT)
+                s.connect((str(ip), port))
+                s.close()
+                ok = True
+                break
+            except Exception:
+                try:
+                    s.close()
+                except Exception:
+                    pass
+    _IPV6_OK[domain] = ok
+    return ok
+
 def _reload_nodes():
     """强制重新读取 nodes.json（菜单打开时调用，外部改动即时生效）。"""
     global _NODES
     _NODES = None
+    _IPV6_OK.clear()    # 节点列表变了，IPv6 可达性/专用判定缓存作废
+    _IPV6_ONLY.clear()
     return _load_nodes()
 
+def _first_reachable_index(nodes):
+    """第一个本机可达的节点下标；都不可达返回 0。供选中不可达节点时回退。"""
+    for i, n in enumerate(nodes):
+        if _node_reachable(n):
+            return i
+    return 0
+
+def _rematch_saved_domain():
+    """按 settings 的 current_node（域名）在最新节点列表里重匹配下标。
+
+    远程同步后服务端可能调整节点顺序：只按下标会静默连到别的节点。
+    返回 True 表示发生了切换。"""
+    global _SELECTED_NODE
+    nodes = _load_nodes()
+    if not nodes:
+        return False
+    saved = str(_settings_load().get("current_node") or "")
+    cur = nodes[_SELECTED_NODE] if 0 <= _SELECTED_NODE < len(nodes) else None
+    if saved and (not cur or str(cur.get("domain") or "") != saved):
+        match = next((i for i, n in enumerate(nodes)
+                      if str(n.get("domain") or "") == saved), None)
+        if match is not None:
+            _SELECTED_NODE = match
+            log(f"节点列表已更新，继续使用上次选择的节点 {saved}")
+            return True
+    return False
+
+def _verify_selected_node():
+    """启动校验，返回 True 表示发生了切换（调用方据此刷新标题栏）。
+
+    两步：① 按设置记录的域名在最新节点列表里重匹配（服务端调过顺序时
+    下标会连到别的节点）；② 选中节点在本机不可达（典型：上次选了 IPv6
+    专用节点，这次本机没有 IPv6 出口）时自动回退到第一个可达节点。"""
+    global _SELECTED_NODE
+    nodes = _load_nodes()
+    switched = _rematch_saved_domain()
+    if not nodes or not (0 <= _SELECTED_NODE < len(nodes)):
+        return switched
+    cur = nodes[_SELECTED_NODE]
+    if _node_reachable(cur):
+        return switched
+    idx = _first_reachable_index(nodes)
+    old = str(cur.get("domain") or "")
+    new = str(nodes[idx].get("domain") or "")
+    _SELECTED_NODE = idx
+    _save_current_node(new)
+    if new and new != old:
+        log_red(f"节点 {old} 在本机不可用，已自动切换到 {new}")
+    return True
+
 def _node_val(key, default=""):
     node = _current_node()
     if not node:
@@ -162,18 +324,39 @@ def _node_val(key, default=""):
 def _endpoint_domain():
     return _node_val("domain", "???")
 
+def _http_method():
+    """节点 HTTP 代理方法参数（新版 nodes.json 的 methods.http）。
+
+    返回 (node, get)：get(key, default) 优先取 methods.http 下的键，
+    缺失时回退节点顶层——兼容旧扁平结构（port/username/password 直接挂在节点上）。
+    """
+    node = _current_node() or {}
+    m = (node.get("methods") or {}).get("http") or {}
+
+    def get(key, default=""):
+        v = m.get(key)
+        if v in (None, ""):
+            v = node.get(key)
+        return default if v in (None, "") else v
+
+    return node, get
+
 def _endpoint_port():
-    return int(_node_val("port", 443))
+    """HTTP 代理端口：methods.http.port（新格式）或节点顶层 port（旧格式）。"""
+    _, get = _http_method()
+    try:
+        return int(get("port", 443))
+    except (TypeError, ValueError):
+        return 443
 
 def _endpoint_country():
     return _node_val("country", "")
 
 def _proxy_auth_basic():
-    """与 nginx 正向代理使用同一套 Basic 认证凭证（取自 nodes.json）。"""
-    node = _current_node()
-    if not node:
-        return ""
-    return base64.b64encode(f"{node.get('username', '')}:{node.get('password', '')}".encode()).decode()
+    """正向代理 Basic 认证凭证：methods.http.{username,password}（新格式）
+    或节点顶层同名字段（旧格式）；用于转发器向上游代理注入 Authorization。"""
+    _, get = _http_method()
+    return base64.b64encode(f"{get('username')}:{get('password')}".encode()).decode()
 
 _tls_forwarder = None
 _pac_server = None
@@ -319,8 +502,76 @@ def _notify_proxy_change():
         pass
 
 
+def _reg_query_value(name):
+    """查询 HKCU Internet Settings 下指定值，返回 (类型, 数据)；值不存在返回 None。"""
+    r = _run(["reg", "query", PROXY_REG, "/v", name], warn=False)
+    if r.returncode != 0:
+        return None
+    for line in (r.stdout or "").splitlines():
+        if name in line:
+            parts = line.strip().split(None, 2)  # 形如 "ProxyServer    REG_SZ    127.0.0.1:7897"
+            if len(parts) >= 3:
+                return parts[1], parts[2]
+    return None
+
+def _is_our_pac_url(value: str) -> bool:
+    """value 是否指向本程序的本地 PAC 服务（http://127.0.0.1:<端口>/proxy.pac）。"""
+    import re
+    return bool(re.fullmatch(r"http://127\.0\.0\.1:\d+/proxy\.pac", value.strip()))
+
+_PROXY_BACKUP_KEY = "prev_proxy_state"   # settings.json 里的系统代理原值备份
+_proxy_prev = None                        # 运行期备份：None = 尚未备份
+
+def _backup_proxy_state():
+    """覆盖系统代理前把原值备份下来（内存 + settings.json，防崩溃丢失）。
+
+    只在尚未备份时执行：本次运行里第一次开启代理时的注册表状态才是用户
+    原有的（可能是其他代理软件的值），之后反复开/关不应覆盖首次备份。"""
+    global _proxy_prev
+    if _proxy_prev is not None:
+        return
+    state = {}
+    for name in ("ProxyEnable", "ProxyServer", "ProxyOverride", "AutoConfigURL"):
+        v = _reg_query_value(name)
+        state[name] = list(v) if v else None
+    _proxy_prev = state
+    try:
+        _settings_save({**_settings_load(), _PROXY_BACKUP_KEY: state})
+    except Exception:
+        pass  # 持久化失败只影响崩溃后的恢复，不影响本次运行
+
+def _restore_proxy_state():
+    """恢复开启代理前的系统代理原值；没有备份时退化为清空（历史行为）。"""
+    global _proxy_prev
+    state = _proxy_prev
+    if state is None:
+        state = _settings_load().get(_PROXY_BACKUP_KEY)
+    if isinstance(state, dict):
+        for name in ("ProxyEnable", "ProxyServer", "ProxyOverride", "AutoConfigURL"):
+            v = state.get(name)
+            if v:
+                _run(["reg", "add", PROXY_REG, "/v", name, "/t", v[0], "/d", v[1], "/f"], warn=False)
+            else:
+                _run(["reg", "delete", PROXY_REG, "/v", name, "/f"], warn=False)
+        log("已恢复开启代理前的系统代理设置")
+    else:
+        # 无备份（旧版本升级后首次运行）：清空，与历史行为一致
+        for name in ("ProxyServer", "ProxyOverride", "AutoConfigURL"):
+            _run(["reg", "delete", PROXY_REG, "/v", name, "/f"], warn=False)
+        _run(["reg", "add", PROXY_REG, "/v", "ProxyEnable", "/t", "REG_DWORD", "/d", "0", "/f"])
+    _proxy_prev = None
+    # 备份已消费：清除持久化副本，避免下次启动误恢复
+    try:
+        data = _settings_load()
+        if isinstance(data, dict) and _PROXY_BACKUP_KEY in data:
+            del data[_PROXY_BACKUP_KEY]
+            _settings_save(data)
+    except Exception:
+        pass
+
 def proxy_on(pac_url=None, bypass_lan=False):
     """设置 Windows 系统代理。pac_url 不为空时启用 PAC 模式，否则使用固定代理。"""
+    _backup_proxy_state()  # 覆盖前保存用户原有代理设置，关闭时恢复而不是清空
     if pac_url:
         # PAC 模式：仅设置 AutoConfigURL，ProxyEnable 置 0，
         # 避免程序读到"固定代理已启用但服务器为空"的混乱状态
@@ -341,11 +592,7 @@ def proxy_on(pac_url=None, bypass_lan=False):
 def proxy_off():
     """关闭系统代理并停止 TLS 转发器 / PAC 本地服务。"""
     log("正在关闭系统代理...")
-    _run(["reg", "add", PROXY_REG, "/v", "ProxyEnable", "/t", "REG_DWORD", "/d", "0", "/f"])
-    # 清除 ProxyServer / ProxyOverride / AutoConfigURL，确保两种模式都清理干净
-    _run(["reg", "delete", PROXY_REG, "/v", "ProxyServer", "/f"], warn=False)
-    _run(["reg", "delete", PROXY_REG, "/v", "ProxyOverride", "/f"], warn=False)
-    _run(["reg", "delete", PROXY_REG, "/v", "AutoConfigURL", "/f"], warn=False)
+    _restore_proxy_state()  # 恢复用户原有代理设置，而不是简单清空
     log("系统代理已关闭")
     _notify_proxy_change()
     _tls_stop()
@@ -468,17 +715,28 @@ def _inject_proxy_auth(data: bytes) -> bytes:
 
     nginx 的 auth_basic 校验 Authorization 而非 Proxy-Authorization，
     因此转发器向上游发送 Basic 认证时使用 Authorization 头。
+    只按完整头名精确匹配（行首到冒号）：Proxy-Authorization 等含
+    "authorization" 字样的行不算已有认证，不能阻止注入；目标站自己的
+    Authorization（明文 HTTP 请求）才跳过。注入位置紧跟请求行，
+    保证上游（取首个 Authorization）读到的是我们的认证。
     """
     if not _starts_with_method(data):
         return data
     end = data.find(b"\r\n\r\n")
     if end == -1:
         return data
-    headers_lower = data[:end].lower()
-    if b"authorization:" in headers_lower:
-        return data
+    head = data[:end]
+    first_end = head.find(b"\r\n")
+    request_line = head if first_end == -1 else head[:first_end]
+    headers = b"" if first_end == -1 else head[first_end + 2:]
+    for line in headers.split(b"\r\n"):
+        if b":" not in line:
+            continue
+        name = line[:line.index(b":")].strip().lower()
+        if name == b"authorization":
+            return data  # 已有目标站自己的 Authorization，不再注入
     auth_line = f"Authorization: Basic {_proxy_auth_basic()}\r\n".encode()
-    return data[:end + 2] + auth_line + data[end + 2:]
+    return request_line + b"\r\n" + auth_line + headers + data[end:]
 
 def _make_request_injector():
     """为明文 HTTP 代理连接生成逐条注入认证的流式处理器。
@@ -501,7 +759,43 @@ def _make_request_injector():
         return b""
     return feed
 
-@functools.lru_cache(maxsize=256)
+def _ttl_cache(seconds: int, maxsize: int = 16384):
+    """带过期时间的缓存装饰器（替代无 TTL 的 lru_cache）。
+
+    域名分流判定结果长期缓存会让域名换 IP 后继续沿用旧的直连/隧道判断；
+    到期自动失效重查。保留 cache_clear() 兼容（测试脚本在用）。"""
+    def deco(func):
+        store = {}
+        lock = threading.Lock()
+
+        def wrapper(*args, **kwargs):
+            key = (args, tuple(sorted(kwargs.items())))
+            now = time.time()
+            with lock:
+                hit = store.get(key)
+            if hit and now - hit[0] < seconds:
+                return hit[1]
+            val = func(*args, **kwargs)
+            with lock:
+                store[key] = (now, val)
+                if len(store) > maxsize:
+                    cutoff = now - seconds
+                    for k, (t, _) in list(store.items()):
+                        if t < cutoff:
+                            del store[k]
+                    if len(store) > maxsize:
+                        store.clear()  # 极端情况下全清：最坏只是多几次解析
+            return val
+
+        def cache_clear():
+            with lock:
+                store.clear()
+
+        wrapper.cache_clear = cache_clear
+        return wrapper
+    return deco
+
+@_ttl_cache(300, maxsize=2048)
 def _is_private_ip(host: str) -> bool:
     """判断目标是否命中常用内网/回环 IPv4；域名会解析后判断。"""
     try:
@@ -869,31 +1163,34 @@ def _auth_load():
 
 def _nodes_sync_remote():
     """启动时用已保存凭据拉取最新节点配置（内容比对，变化才覆盖）。"""
-    cred = _auth_load()
-    if not cred:
-        return  # 从未登录过，等登录框处理
-    url = str(_config_val("nodes_url", "") or "").strip()
-    if not url:
-        return
     try:
-        import json
-        data = _fetch_nodes_auth(url, cred["username"], cred["password"])
-        if not (isinstance(data, list) and data
-                and all(isinstance(n, dict) for n in data)
-                and str(data[0].get("domain") or "").strip()):
-            raise ValueError("返回的节点数据无效")
-        text = json.dumps(data, ensure_ascii=False, indent=2) + "\n"
-        local = _USER_DIR / "nodes.json"
-        if local.is_file() and _nodes_decrypt(local.read_bytes()) == text:
-            log("节点配置已是最新")
+        cred = _auth_load()
+        if not cred:
+            return  # 从未登录过，等登录框处理
+        url = str(_config_val("nodes_url", "") or "").strip()
+        if not url:
             return
-        _USER_DIR.mkdir(parents=True, exist_ok=True)
-        local.write_bytes(_nodes_encrypt(text))
-        global _NODES
-        _NODES = None  # 使后续读取（节点菜单/开启代理）拿到新配置
-        log(f"节点配置已更新: {len(data)} 个节点 (nodes.json)")
-    except Exception as e:
-        log(f"节点配置在线更新失败，使用本地配置: {e}")
+        try:
+            import json
+            data = _fetch_nodes_auth(url, cred["username"], cred["password"])
+            if not (isinstance(data, list) and data
+                    and all(isinstance(n, dict) for n in data)
+                    and str(data[0].get("domain") or "").strip()):
+                raise ValueError("返回的节点数据无效")
+            text = json.dumps(data, ensure_ascii=False, indent=2) + "\n"
+            local = _USER_DIR / "nodes.json"
+            if local.is_file() and _nodes_decrypt(local.read_bytes()) == text:
+                log("节点配置已是最新")
+                return
+            _USER_DIR.mkdir(parents=True, exist_ok=True)
+            local.write_bytes(_nodes_encrypt(text))
+            global _NODES
+            _NODES = None  # 使后续读取（节点菜单/开启代理）拿到新配置
+            log(f"节点配置已更新: {len(data)} 个节点 (nodes.json)")
+        except Exception as e:
+            log(f"节点配置在线更新失败，使用本地配置: {e}")
+    finally:
+        _NODES_SYNC_DONE.set()  # 无论成败都要放行启动校验
 
 class _AuthError(Exception):
     """订阅接口返回的错误：code 为 HTTP 状态码，str 为服务端返回的中文提示。"""
@@ -1048,7 +1345,7 @@ def _is_private_int(n: int) -> bool:
             (n & 0xFF000000) == 0x7F000000 or
             (n & 0xFFFF0000) == 0xA9FE0000)
 
-@functools.lru_cache(maxsize=16384)
+@_ttl_cache(300, maxsize=16384)
 def _is_china_ip(host: str):
     """判断目标是否命中中国 IP 段；域名会先解析。
 
@@ -1129,7 +1426,7 @@ def _whitelist_entry_match(host: str, entry: str) -> bool:
     # 域名条目：精确匹配或子域后缀匹配（与 pac_domains 同一规则）
     return h == entry or h.endswith("." + entry)
 
-@functools.lru_cache(maxsize=8192)
+@_ttl_cache(300, maxsize=8192)
 def _is_whitelisted(host: str, version: int) -> bool:
     """host 是否命中已启用的白名单规则（按版本键控缓存，规则变更即失效）。"""
     entries = [r["value"] for r in _whitelist_rules() if r.get("enabled")]
@@ -1154,11 +1451,24 @@ def _decide_direct(host: str, bypass_lan: bool, smart: bool) -> bool:
     return False
 
 def _parse_connect_authority(first_line: str):
-    """从 CONNECT 首行提取 (host, port)，失败返回 None。"""
+    """从 CONNECT 首行提取 (host, port)，失败返回 None。
+
+    IPv6 字面量带方括号（CONNECT [::1]:443）：剥掉括号再按最后一个冒号拆分，
+    否则会把 "[::1]" 当成主机名，直连/白名单判定全部失效。"""
     parts = first_line.split(" ")
     if len(parts) < 2:
         return None
     authority = parts[1]
+    if authority.startswith("[") and "]" in authority:
+        # 括号必须紧跟在主机名部分之后（[::1]:443）；"[::1" 无右括号等畸形输入
+        # 会因 "]:" 检查失败而走下面的普通拆分，最终被 port_s.isdigit() 拦下
+        try:
+            host = authority[1:authority.index("]")]
+            tail = authority[authority.index("]") + 1:]
+            if tail.startswith(":") and tail[1:].isdigit():
+                return host, int(tail[1:])
+        except ValueError:
+            pass
     if ":" not in authority:
         return None
     host, _, port_s = authority.rpartition(":")
@@ -1341,27 +1651,28 @@ def _release_forwarder(started_here):
 
 def is_proxy_on():
     """系统代理是否生效：固定代理已开启，或 AutoConfigURL 指向本地 PAC 服务。"""
-    r = _run(["reg", "query", PROXY_REG, "/v", "ProxyEnable"], warn=False)
-    if "0x1" in (r.stdout or ""):
+    v = _reg_query_value("ProxyEnable")
+    if v and "0x1" in (v[1] or ""):
         return True
-    r = _run(["reg", "query", PROXY_REG, "/v", "AutoConfigURL"], warn=False)
-    return "127.0.0.1" in (r.stdout or "")
+    v = _reg_query_value("AutoConfigURL")
+    return bool(v) and _is_our_pac_url(v[1] or "")
 
 def _cleanup_stale_proxy():
-    """清理上次异常退出残留的系统代理设置。
+    """清理上次异常退出残留的系统代理设置，并恢复用户原有设置。
 
     TLS 转发器 / PAC 服务只活在进程内，程序启动时若注册表仍指向
     本地转发端口或本地 PAC 地址，说明是残留——不清理会导致全网不通。
+    只认本程序的精确地址：其他软件（如 Clash）的 PAC 也在 127.0.0.1 上，
+    不能因为地址里带 127.0.0.1 就误删。
     """
     try:
-        if not is_proxy_on():
+        server = _reg_query_value("ProxyServer")
+        pac = _reg_query_value("AutoConfigURL")
+        ours = (bool(server) and (server[1] or "").strip() == f"{PROXY_HOST}:{PROXY_PORT}") or \
+               (bool(pac) and _is_our_pac_url(pac[1] or ""))
+        if not ours:
             return
-        r = _run(["reg", "query", PROXY_REG, "/v", "ProxyServer"], warn=False)
-        r2 = _run(["reg", "query", PROXY_REG, "/v", "AutoConfigURL"], warn=False)
-        stale = (f"{PROXY_HOST}:{PROXY_PORT}" in (r.stdout or "")
-                 or "127.0.0.1" in (r2.stdout or ""))
-        if stale:
-            proxy_off()
+        proxy_off()  # 内部从备份恢复用户原有设置（无备份则清空）
     except Exception:
         pass
 
@@ -1998,6 +2309,7 @@ class MainWindow(QWidget):
     _update_download_done = pyqtSignal(bool, str)
 
     _status_ready = pyqtSignal(object)
+    _node_switched = pyqtSignal()   # 启动校验时代替用户切换了节点（刷新标题栏）
 
     def __init__(self):
         super().__init__()
@@ -2011,6 +2323,7 @@ class MainWindow(QWidget):
         self._update_done.connect(self._on_update_done)
         self._update_download_done.connect(self._on_update_download_done)
         self._status_ready.connect(self._apply_status)
+        self._node_switched.connect(self._on_node_switched)
 
         _user_init_local()  # 首启初始化用户数据目录（节点模板），须在 init_ui 读节点之前
 
@@ -2028,9 +2341,28 @@ class MainWindow(QWidget):
         threading.Thread(target=_user_init_remote, daemon=True).start()
         # 后台用已保存凭据同步最新节点配置（服务端改动下次启动生效）
         threading.Thread(target=_nodes_sync_remote, daemon=True).start()
+        # 后台校验上次选中的节点在本机是否还可用（IPv6 专用节点本机无 IPv6 时回退）
+        threading.Thread(target=self._verify_node_thread, daemon=True).start()
         # 节点未配置时延迟弹出登录框（等窗口显示后）
         QTimer.singleShot(600, self._maybe_show_login)
 
+    def _verify_node_thread(self):
+        """等节点配置在线同步完成后，校验当前选中节点是否可用，必要时自动回退。
+
+        探测要解析 DNS 并尝试 TCP 连接，放在后台线程，避免卡住启动。
+        订阅同步可能较慢：等同步完成（最多 15 秒），同步后才校验，
+        否则服务端换过节点顺序时校验的是旧列表。"""
+        _NODES_SYNC_DONE.wait(15)
+        try:
+            if _verify_selected_node():
+                self._node_switched.emit()
+        except Exception as e:
+            log(f"节点可用性校验失败: {e}")
+
+    def _on_node_switched(self):
+        """节点被自动切换后刷新标题栏（在 UI 线程执行）。"""
+        self.subtitle.setText(f"{_endpoint_domain()} ({_endpoint_country()})")
+
     def init_ui(self):
         """初始化主界面布局与样式。"""
         self.setWindowTitle("R.R.P.P.")
@@ -2268,6 +2600,9 @@ class MainWindow(QWidget):
         if _endpoint_domain() == "???":
             # 未配置节点时禁止开启：否则系统代理会指向连不通任何目标的转发器，导致断网
             raise RuntimeError("未配置节点：请先通过 nodes_url 在线更新或 nodes_tool 导入节点配置")
+        if not _node_reachable(_current_node()):
+            # 当前选中节点是 IPv6 专用而本机没有 IPv6 出口：转发的流量全部连不上
+            raise RuntimeError("当前节点仅提供 IPv6 地址，本机没有 IPv6 出口，请在节点菜单切换其他节点")
         _traffic_reset()
         log("TLS 直连模式")
         ep = _endpoint_domain()
@@ -2365,6 +2700,11 @@ class MainWindow(QWidget):
                                     "nodes.json 中没有可用节点。\n"
                                     "可在右下角「设置」中退出登录后重新登录获取节点配置。")
             return
+        menu = self._build_node_menu(nodes)
+        menu.exec_(self.subtitle.mapToGlobal(self.subtitle.rect().bottomLeft()))
+
+    def _build_node_menu(self, nodes):
+        """构建节点切换菜单：IPv6 专用节点加标注，本机无 IPv6 出口时置灰不可选。"""
         menu = QMenu(self)
         menu.setStyleSheet(f"""
             QMenu {{
@@ -2391,11 +2731,24 @@ class MainWindow(QWidget):
             text = f"{domain} ({country})"
             if remark and remark != domain:
                 text += f"　{remark}"
+            ipv6_only = _node_ipv6_only(n)
+            reachable = _node_reachable(n)
+            if ipv6_only:
+                # 不可达时的文案由 tooltip 改为直接写在菜单项上，用户一眼可见
+                text += "　[IPv6]" if reachable else "　[不可用：本机无 IPv6]"
             if i == cur:
                 text = "✓ " + text
-            menu.addAction(text).triggered.connect(
-                lambda _=False, idx=i: self._on_node_selected(idx))
-        menu.exec_(self.subtitle.mapToGlobal(self.subtitle.rect().bottomLeft()))
+            act = menu.addAction(text)
+            if not reachable:
+                # 本机无 IPv6 出口：置灰不可选，避免选中后连不上
+                act.setEnabled(False)
+                act.setToolTip("不可用：本机无 IPv6")
+            else:
+                if ipv6_only:
+                    act.setToolTip("该节点仅提供 IPv6 地址（本机 IPv6 可用）")
+                act.triggered.connect(
+                    lambda _=False, idx=i: self._on_node_selected(idx))
+        return menu
 
     def _on_node_selected(self, idx):
         """切换节点：更新标题。"""
@@ -2494,6 +2847,8 @@ class MainWindow(QWidget):
         # 先初始化结果，任何路径（含跳过）完成后 _on_update_done 都能安全读取
         self._latest_version = ""
         self._download_url = ""
+        self._asset_size = 0       # release API 记录的附件大小，下载后比对
+        self._checksum_url = ""    # SHA-256 边车附件链接（无则跳过校验）
         try:
             if _skip_update_check() and not manual:
                 return  # 设置跳过（仅自动检查）：不检查也不打日志
@@ -2520,11 +2875,23 @@ class MainWindow(QWidget):
                 req = urllib.request.Request(url)
                 data = opener.open(req, timeout=20).read().decode("utf-8-sig")
             info = json.loads(data)
-            # 基于 Gitea 仓库 releases API：版本取 tag_name，下载取 exe 附件链接
+            # 基于 Gitea 仓库 releases API：版本取 tag_name，下载取 exe 附件链接。
+            # 附件可能有多个（发布流程附带 .sha256 边车），按名字精确匹配而不是
+            # 固定取第一个，避免附件顺序变化时下载到错文件。
             self._latest_version = str(info.get("tag_name") or "")
             assets = info.get("assets") or []
-            self._download_url = (assets[0].get("browser_download_url") or ""
-                                  if assets else "")
+            wanted = f"RRPP_Lite_{self._latest_version}.exe" if self._latest_version else ""
+            exe_asset = next((a for a in assets if a.get("name") == wanted), None)
+            if exe_asset is None:
+                # 老 release 命名不齐时回退到任意 exe 附件
+                exe_asset = next((a for a in assets
+                                  if str(a.get("name") or "").lower().endswith(".exe")), None)
+            if exe_asset is None:
+                raise RuntimeError("release 中没有可用的 EXE 附件")
+            self._download_url = exe_asset.get("browser_download_url") or ""
+            self._asset_size = int(exe_asset.get("size") or 0)
+            sha_asset = next((a for a in assets if a.get("name") == wanted + ".sha256"), None)
+            self._checksum_url = (sha_asset.get("browser_download_url") or "") if sha_asset else ""
             log(f"当前版本: {VERSION}，最新版本: {self._latest_version or '未知'}")
         except Exception as e:
             log(f"更新检查失败: {e}")
@@ -2547,9 +2914,15 @@ class MainWindow(QWidget):
                 threading.Thread(target=self._download_update, daemon=True).start()
 
     def _download_update(self):
-        """后台下载新版本 EXE 到 EXE 所在目录。"""
+        """后台下载新版本 EXE 到 EXE 所在目录。
+
+        落盘前校验：① 大小与 release API 记录一致（传输完整性）；
+        ② 发布流程附带 .sha256 边车时校验 SHA-256（防链路被攻破/文件被替换）。
+        边车缺失（老 release）只做大小校验。
+        """
         try:
-            import urllib.request, urllib.parse, base64
+            import hashlib
+            import urllib.request, urllib.parse
             url = self._download_url
             if not url:
                 self._update_download_done.emit(False, "下载链接为空")
@@ -2561,29 +2934,40 @@ class MainWindow(QWidget):
             log(f"正在下载新版本: {url}")
             log(f"保存到: {save_path}")
 
-            # 同更新检查：按「使用代理更新」设置选择隧道或直连
-            if _update_via_proxy():
-                started_here = _ensure_forwarder()
-                try:
+            def fetch(target):
+                # 同更新检查：按「使用代理更新」设置选择隧道或直连
+                if _update_via_proxy():
                     opener = urllib.request.build_opener(urllib.request.ProxyHandler(
                         {"http": f"{PROXY_HOST}:{PROXY_PORT}",
                          "https": f"{PROXY_HOST}:{PROXY_PORT}"}))
-                    req = urllib.request.Request(url)
+                else:
+                    opener = _pac_no_proxy_opener()
+                with opener.open(urllib.request.Request(target), timeout=120) as resp:
+                    return resp.read()
 
-                    with opener.open(req, timeout=120) as resp:
-                        data = resp.read()
-                finally:
-                    _release_forwarder(started_here)
-            else:
-                opener = _pac_no_proxy_opener()
-                req = urllib.request.Request(url)
-
-                with opener.open(req, timeout=120) as resp:
-                    data = resp.read()
+            started_here = _ensure_forwarder() if _update_via_proxy() else False
+            try:
+                data = fetch(url)
+                checksum_text = ""
+                if self._checksum_url:
+                    checksum_text = fetch(self._checksum_url).decode("ascii", "replace")
+            finally:
+                _release_forwarder(started_here)
 
             if len(data) < 1024:
                 self._update_download_done.emit(False, "下载文件过小，可能不是有效的 EXE")
                 return
+            if self._asset_size and len(data) != self._asset_size:
+                self._update_download_done.emit(
+                    False, f"下载不完整：实际 {len(data)} 字节，应为 {self._asset_size} 字节")
+                return
+            if checksum_text:
+                want = checksum_text.strip().split()[0].lower()
+                got = hashlib.sha256(data).hexdigest()
+                if len(want) != 64 or want != got:
+                    self._update_download_done.emit(False, "更新包校验失败（SHA-256 不匹配），已放弃")
+                    return
+                log("更新包 SHA-256 校验通过")
 
             save_path.write_bytes(data)
             log(f"新版本已下载: {save_path}")

lite/test/check_node_menu.py

@@ -0,0 +1,104 @@
+# -*- coding: utf-8 -*-
+"""离屏验证节点菜单：IPv6-only 节点的标注与置灰（真实 PyQt5，不出窗口）。
+
+需要 Windows venv 的真实 PyQt5：
+  ..\\.venv\\Scripts\\python.exe test\\check_node_menu.py
+读取 test/nodes.json 作为节点数据，只构建菜单不进事件循环。
+"""
+import os, sys
+os.environ["QT_QPA_PLATFORM"] = "offscreen"
+from pathlib import Path
+
+LITE = Path(__file__).resolve().parent.parent
+# 脚本所在目录（test/）会被解释器自动放到 sys.path 首位，那里有 PyQt5 桩，
+# 会顶掉真实 PyQt5 —— 必须移除，改用本程序自己的 venv
+sys.path = [str(LITE)] + [p for p in sys.path if Path(p).resolve() != Path(__file__).resolve().parent]
+
+sys.stdout.reconfigure(encoding="utf-8")
+
+import proxy_lite as pp
+from PyQt5.QtWidgets import QApplication
+
+# 构造窗口时不写用户数据目录（避免污染），构造后再指向真实 nodes.json
+pp._user_init_local = lambda: None
+
+app = QApplication(sys.argv)
+w = pp.MainWindow()
+w.timer.stop()                      # 停掉状态轮询，纯静态检查
+
+pp._USER_DIR = LITE / "test"        # 真实 nodes.json 所在目录
+nodes = pp._reload_nodes()
+print(f"本机有全局 IPv6: {pp._local_has_ipv6()}")
+print(f"节点数: {len(nodes)}")
+
+menu = w._build_node_menu(nodes)
+print("\n菜单项（enabled / 文案 / tooltip）:")
+for act in menu.actions():
+    print(f"  {str(act.isEnabled()):5} | {act.text()}")
+    if act.toolTip():
+        print(f"        └ {act.toolTip()}")
+
+# 点击被禁用的项不应切换节点
+before = pp._selected_index()
+for act in menu.actions():
+    if not act.isEnabled():
+        act.trigger()
+print(f"\n点击禁用项后节点下标: {before} -> {pp._selected_index()} (应不变)")
+
+# ── 上次选中的节点已不可用：启动时自动回退到第一个可达节点 ──────────
+print("\n启动校验（_verify_selected_node）:")
+bad = [i for i, n in enumerate(nodes) if not pp._node_reachable(n)]
+if not bad:
+    print("  本机所有节点均可达，无法验证回退（本机有 IPv6？）")
+else:
+    import json
+    _bak = (pp._USER_DIR / "settings.json").read_bytes() if (pp._USER_DIR / "settings.json").is_file() else None
+    bad_idx = bad[0]
+    pp._SELECTED_NODE = bad_idx           # 模拟上次选了不可用节点
+    switched = pp._verify_selected_node()
+    expect = pp._first_reachable_index(nodes)
+    ok = switched and pp._selected_index() == expect
+    print(f"  上次选中 {nodes[bad_idx].get('domain')}(下标 {bad_idx}) 不可用 -> "
+          f"回退到下标 {pp._selected_index()} ({nodes[pp._selected_index()].get('domain')}) "
+          f"{'PASS' if ok else 'FAIL'}")
+    saved = json.loads((pp._USER_DIR / "settings.json").read_text("utf-8")).get("current_node") if (pp._USER_DIR / "settings.json").is_file() else ""
+    ok2 = saved == str(nodes[expect].get("domain") or "")
+    print(f"  回退结果已写入设置 current_node={saved!r} {'PASS' if ok2 else 'FAIL'}")
+    # 再校验一次：已经是可达节点，不应再切换
+    ok3 = pp._verify_selected_node() is False
+    print(f"  再次校验不重复切换 {'PASS' if ok3 else 'FAIL'}")
+    if _bak is None:
+        (pp._USER_DIR / "settings.json").unlink(missing_ok=True)
+    else:
+        (pp._USER_DIR / "settings.json").write_bytes(_bak)
+
+# ── 服务端调整节点顺序：按域名重匹配而不是死守下标 ──────────────────
+print("\n服务端调序（_rematch_saved_domain）:")
+if len(nodes) < 2:
+    print("  节点不足 2 个，跳过")
+else:
+    import json
+    dom_old = nodes[0].get("domain")
+    ordered = [nodes[1], nodes[0]] + nodes[2:]     # 模拟服务端交换了前两个节点
+    pp._NODES = ordered
+    # 设置里记录的是原下标 0 的域名，重匹配后下标应跟随域名（新列表里是 1）
+    pp._settings_save({**pp._settings_load(), "current_node": dom_old})
+    switched = pp._rematch_saved_domain()
+    ok = switched and ordered[pp._selected_index()].get("domain") == dom_old
+    print(f"  current_node={dom_old} 调序后跟随域名 -> 下标 {pp._selected_index()} "
+          f"{'PASS' if ok else 'FAIL'}")
+    # 已是正确域名时再跑一次：不应重复切换
+    ok2 = pp._rematch_saved_domain() is False
+    print(f"  再次重匹配不重复切换 {'PASS' if ok2 else 'FAIL'}")
+    # 域名已不在列表（节点被移除）：不切换，保持现状
+    pp._settings_save({**pp._settings_load(), "current_node": "gone.example.com"})
+    ok3 = pp._rematch_saved_domain() is False
+    print(f"  节点被移除时不动 {'PASS' if ok3 else 'FAIL'}")
+    pp._NODES = None   # 还原缓存，后续用例重读文件
+    # 还原 settings.json，避免测试污染
+    if _bak is None:
+        (pp._USER_DIR / "settings.json").unlink(missing_ok=True)
+    else:
+        (pp._USER_DIR / "settings.json").write_bytes(_bak)
+
+print("OK")

lite/test/test_pac_routing.py

@@ -0,0 +1,389 @@
+# -*- coding: utf-8 -*-
+"""PAC（智能分流）模式路由决策测试。
+
+核心问题：PAC 模式下，目标域名解析为国外 IP 时，是否强制走隧道代理？
+
+判定链（proxy_lite._decide_direct）：
+  0. 代理白名单命中 -> 强制直连（最优先，PAC/固定模式均生效）
+  1. pac_domains.json 域名表命中 -> 强制隧道（优先于 IP 判定）
+  2. 内网 IP 且勾选 bypass_lan -> 直连
+  3. 智能分流：中国 IP -> 直连；其余（国外 IP）-> 走隧道
+
+直接 import proxy_lite 调用真实决策函数，数据用真实的 pac_ip_cn.json /
+pac_domains.json；PyQt5 用桩替代（决策逻辑不依赖 Qt）。
+"""
+
+import sys
+from pathlib import Path
+
+TEST_DIR = Path(__file__).resolve().parent
+LITE_DIR = TEST_DIR.parent
+
+sys.path.insert(0, str(TEST_DIR))  # 假 PyQt5 桩
+sys.path.insert(0, str(LITE_DIR))
+
+import json
+import socket
+import proxy_lite as pp
+from proxy_lite import _decide_direct, _is_china_ip, _is_force_proxy, _is_private_ip, _is_whitelisted
+
+# PAC 数据已不再随项目保留：从 CDN 拉取到临时用户目录（真实数据端到端）
+import tempfile
+pp._USER_DIR = Path(tempfile.mkdtemp()) / "RRPP"
+pp._user_init_remote(base_url="https://cdn.rrrosmontis.icu/RRPP/PAC/")
+
+pp._load_cn_ranges()
+pp._load_force_proxy_suffixes()
+if not pp._CN_RANGES:
+    print("警告: 未能获取 pac_ip_cn.json（CDN 不可达？），中国 IP 判定将失效")
+print(f"中国IP段: {len(pp._CN_RANGES or [])} 段, 强制域名表: {len(pp._FORCE_PROXY_SUFFIXES or [])} 条\n")
+
+
+def resolved_ips(host):
+    try:
+        return sorted({a[4][0] for a in socket.getaddrinfo(host, 443, socket.AF_INET)})
+    except Exception:
+        return []
+
+
+def mark(ip):
+    if _is_private_ip(ip):
+        return "内网"
+    if _is_china_ip(ip):
+        return "中国IP"
+    return "国外IP"
+
+
+PATH = {True: "直连", False: "隧道"}
+MODES = [
+    ("PAC+bypass_lan", dict(bypass_lan=True, smart=True)),   # 默认勾选状态
+    ("PAC无bypass_lan", dict(bypass_lan=False, smart=True)),
+    ("固定代理模式",    dict(bypass_lan=False, smart=False)),  # 全部走隧道
+]
+
+# (名称, host, 期望 PAC+bypass_lan 决策: True=直连 / False=隧道)
+cases = [
+    ("境外域名 google.com",       "google.com",          False),
+    ("境外域名 github.com",       "github.com",          False),
+    ("境外域名 youtube.com",      "youtube.com",         False),
+    ("域名表强制 claude.ai",      "claude.ai",           False),
+    ("域名表强制 *.google 后缀",  "antigravity.google",  False),
+    ("国内域名 www.baidu.com",    "www.baidu.com",       True),
+    ("国内域名 www.qq.com",       "www.qq.com",          True),
+    ("内网 localhost",            "localhost",           True),
+    ("内网 IP 192.168.1.1",       "192.168.1.1",         True),
+    ("境外 IP 8.8.8.8",           "8.8.8.8",             False),
+    ("国内 IP 114.114.114.114",   "114.114.114.114",     True),
+]
+
+failed = 0
+for name, host, expect in cases:
+    ips = resolved_ips(host)
+    force = _is_force_proxy(host)
+    marks = f"[{','.join(mark(ip) for ip in ips)}]" if ips else "[解析失败]"
+    row = f"{name:<22} {host:<18} {str(ips) if ips else '无IP':<45} {marks}"
+    results = []
+    for mode_name, kw in MODES:
+        got = _decide_direct(host, **kw)
+        results.append(f"{mode_name}={PATH[got]}")
+    pac = _decide_direct(host, **MODES[0][1])
+    ok = pac == expect
+    failed += 0 if ok else 1
+    print(f"{row}\n    {'域名表命中' if force else '域名表未命中':<10} | {'  '.join(results)} | 期望 {PATH[expect]} -> {'PASS' if ok else 'FAIL'}")
+
+print(f"\n{'='*70}")
+print(f"结论: {'全部 PASS' if not failed else f'{failed} 个 FAIL'}")
+print("=" * 70)
+print("PAC 模式(智能分流)下: 域名解析结果全部为国外 IP -> 走隧道代理 ✓"
+      if not failed else "存在 FAIL，请检查上方明细")
+
+# ── 附加测试：不在域名表的被墙境外站 ─────────────────────
+# 这类站点不能靠域名表兜底，必须仅凭「解析结果为国外 IP」强制走隧道，
+# 否则会尝试直连 -> 被墙。
+print("\n" + "=" * 70)
+print("附加: 不在 pac_domains.json 的被墙境外站（只能靠 IP 判定走隧道）")
+print("=" * 70)
+
+extra_candidates = [
+    "dailymotion.com", "vimeo.com", "notion.so", "linear.app",
+    "quora.com", "feedly.com", "dropbox.com", "tumblr.com",
+    "flickr.com", "pastebin.com", "fandom.com", "archive.org",
+    "godaddy.com", "imgur.com", "kickstarter.com", "shopify.com",
+]
+
+extra_failed = 0
+skipped = 0
+for host in extra_candidates:
+    if _is_force_proxy(host):
+        print(f"{host:<20} 已在域名表中，跳过")
+        skipped += 1
+        continue
+    ips = resolved_ips(host)
+    if not ips:
+        print(f"{host:<20} 解析失败，跳过")
+        skipped += 1
+        continue
+    foreign = all(not _is_private_ip(ip) and not _is_china_ip(ip) for ip in ips)
+    if not foreign:
+        print(f"{host:<20} 解析到非国外 IP {ips}，跳过（不是有效的境外用例）")
+        skipped += 1
+        continue
+    got = _decide_direct(host, bypass_lan=True, smart=True)  # PAC 模式
+    if got is not False:
+        # 决策函数内部会重新解析域名，CDN 轮询 DNS 可能与其上方展示的解析结果
+        # 不一致：清缓存重试一次，仍不一致则按 DNS 不稳定跳过（非代码问题）
+        _is_china_ip.cache_clear()
+        _is_private_ip.cache_clear()
+        ips = resolved_ips(host)
+        got = _decide_direct(host, bypass_lan=True, smart=True)
+        if got is not False:
+            print(f"{host:<20} 解析/判定随 DNS 轮询不一致 {ips}，跳过（非代码问题）")
+            skipped += 1
+            continue
+        print(f"{host:<20} 重试后走隧道（DNS 瞬时差异） PASS")
+        continue
+    extra_failed += 0
+    print(f"{host:<20} {str(ips):<50} [{'，'.join(mark(ip) for ip in ips)}] -> "
+          f"{PATH[got]} (期望隧道) PASS")
+
+print(f"\n附加测试: {len(extra_candidates)} 个候选, 有效用例 {len(extra_candidates) - skipped} 个, "
+      f"{'全部 PASS' if not extra_failed else f'{extra_failed} 个 FAIL'}")
+if extra_failed:
+    failed += extra_failed
+print("=" * 70)
+
+# ── 代理白名单测试（用户可配置：命中即直连，最优先，含 IPv4/IPv6/CIDR） ─────
+print("\n" + "=" * 70)
+print("代理白名单: 命中规则强制直连（优先于域名表/固定代理，支持 IPv4/IPv6/CIDR）")
+print("=" * 70)
+
+wl_cases = [
+    # (规则, 测试host, 期望命中)
+    ("example.com",         "example.com",          True),
+    ("example.com",         "www.example.com",      True),
+    ("example.com",         "sub.www.example.com",  True),
+    ("example.com",         "example.org",          False),
+    ("claude.ai",           "claude.ai",            True),   # 域名表强制隧道，白名单优先
+    ("localhost",           "localhost",            True),
+    ("8.8.8.8",             "8.8.8.8",              True),   # IPv4 字面量
+    ("8.8.8.8",             "8.8.4.4",              False),
+    ("10.0.0.0/8",          "10.1.2.3",             True),   # IPv4 CIDR
+    ("192.0.2.0/24",        "192.0.2.55",           True),   # IPv4 CIDR
+    ("2001:db8::/32",       "2001:db8::1",          True),   # IPv6 网段
+    ("fe80::1",             "fe80::1",              True),   # IPv6 单地址
+]
+
+# 注入白名单规则（绕过 settings.json 直接改内存缓存；版本 +1 使匹配缓存失效），
+# 另加一条"禁用"的规则验证不参与匹配
+pp._WHITELIST = [{"value": "203.0.113.0/24", "enabled": False}] + \
+                [{"value": r, "enabled": True} for r, _, _ in wl_cases]
+wl_ver = pp._WHITELIST_VERSION + 1
+pp._WHITELIST_VERSION = wl_ver
+
+wl_failed = 0
+for rule, host, expect in wl_cases:
+    got = _is_whitelisted(host, wl_ver)
+    ok = got == expect
+    wl_failed += 0 if ok else 1
+    print(f"规则 {rule:<16} host {host:<22} 命中={got} 期望={expect} -> {'PASS' if ok else 'FAIL'}")
+got = _is_whitelisted("203.0.113.9", wl_ver)  # 禁用的规则不应命中
+print(f"禁用规则 203.0.113.0/24 host 203.0.113.9 -> 命中={got} 期望=False -> {'PASS' if not got else 'FAIL'}")
+wl_failed += 0 if not got else 1
+
+# 白名单在三种模式下都强制直连（固定代理模式原本全部走隧道）
+print("\n白名单在三种模式下的决策（命中白名单应为直连）:")
+for mode_name, kw in MODES:
+    got = _decide_direct("www.example.com", **kw)
+    ok = got is True
+    wl_failed += 0 if ok else 1
+    print(f"  {mode_name:<16} www.example.com -> {PATH[got]} (期望 直连) {'PASS' if ok else 'FAIL'}")
+# 白名单优先于域名表：claude.ai 在 pac_domains.json 中强制隧道，白名单命中应直连
+got = _decide_direct("claude.ai", bypass_lan=True, smart=True)
+ok = got is True
+wl_failed += 0 if ok else 1
+print(f"  {'域名表冲突':<14} claude.ai -> {PATH[got]} (期望 直连，白名单优先) {'PASS' if ok else 'FAIL'}")
+# 未命中白名单的境外域名仍走隧道（白名单不影响其他路由）
+got = _decide_direct("google.com", bypass_lan=True, smart=True)
+ok = got is False
+wl_failed += 0 if ok else 1
+print(f"  {'未命中白名单':<14} google.com -> {PATH[got]} (期望 隧道) {'PASS' if ok else 'FAIL'}")
+
+# 恢复空白名单，避免影响后续用例
+pp._WHITELIST = None
+pp._WHITELIST_VERSION += 1
+if wl_failed:
+    failed += wl_failed
+print("=" * 70)
+
+# ── IPv6-only 节点标注 ─────────────────────────────────────
+# IPv6 专用节点只提供 AAAA 记录，本机没有 IPv6 出口时连不上：
+# 客户端据此在节点菜单标注并置灰。判定三级：ipv6_only 字段 →
+# remark 里的 "(IPv6)"（服务端 rrpp_client 现网唯一的线索）→ 域名解析兜底。
+print("\n" + "=" * 70)
+print("IPv6 专用节点：识别（三级判定）与可达性")
+print("=" * 70)
+
+ipv6_failed = 0
+has_v6 = pp._local_has_ipv6()
+print(f"本机是否有全局 IPv6 地址: {has_v6}")
+
+# 用真实 nodes.json（test 目录）；没有则跳过
+_nodes_file = TEST_DIR / "nodes.json"
+real_nodes = json.loads(_nodes_file.read_text("utf-8")) if _nodes_file.is_file() else []
+if real_nodes:
+    pp._IPV6_OK.clear()   # 清缓存，做真实探测
+    for n in real_nodes:
+        c = n.get("country") or "?"
+        io = pp._node_ipv6_only(n)
+        reach = pp._node_reachable(n)
+        mark = "[IPv6]" if reach else "[不可用：本机无 IPv6]"
+        print(f"  [{c}] {n.get('domain'):<24} ipv6_only={str(io):<6} "
+              f"可达={str(reach):<6} 菜单文案: {n.get('domain')} ({c})"
+              + (f"　{mark}" if io else ""))
+        # ipv6_only 节点在本机无 IPv6 时必须判定为不可达，否则必须可达
+        if io:
+            expect = has_v6 or reach  # 本机有 IPv6 -> 应可达；无 IPv6 -> 不应可达
+            ok = reach == expect
+            ipv6_failed += 0 if ok else 1
+            print(f"        IPv6 专用节点可达性: {reach} (期望 {expect}) "
+                  f"{'PASS' if ok else 'FAIL'}")
+        else:
+            ok = reach is True
+            ipv6_failed += 0 if ok else 1
+            print(f"        普通节点恒可达: {reach} {'PASS' if ok else 'FAIL'}")
+    # 缓存生效：再次判定不发新探测（结果一致）
+    cached = list(pp._IPV6_OK.items())
+    if cached:
+        print(f"  可达性缓存: {cached}")
+
+# 三级判定：字段 → remark 里的 "(IPv6)" 标记 → 域名解析兜底。
+# 服务端 subscription?method=rrpp_client 只吐旧扁平格式，ipv6_only 字段被丢掉，
+# 只在 remark 里留 "(IPv6)"（后端 countries_to_rrpp_client），故第 2 级是现网主路径。
+synthetic = [
+    ({"domain": "example.com"}, False, "无标记（example.com 有 A 记录）"),
+    ({"domain": "example.com", "ipv6_only": False}, False, "ipv6_only=false"),
+    ({"domain": "example.com", "ipv6_only": True}, True, "ipv6_only=true"),
+    ({"domain": "example.com", "remark": "GlobalVM 350G (IPv6)"}, True, "remark 含 (IPv6)"),
+    ({"domain": "example.com", "remark": "Azure 15G"}, False, "remark 无 IPv6 标记"),
+    ({"domain": "example.com", "remark": "IPv6 专用机"}, True, "remark 含 IPv6（无括号）"),
+    ({"domain": "example.com", "remark": "some_ipv6_thing"}, True, "remark 小写 ipv6"),
+]
+for node, expect, desc in synthetic:
+    got = pp._node_ipv6_only(node)
+    ok = got == expect
+    ipv6_failed += 0 if ok else 1
+    print(f"  {desc:<24} _node_ipv6_only={got} (期望 {expect}) {'PASS' if ok else 'FAIL'}")
+pp._IPV6_OK.clear()   # 合成用例可能污染缓存
+
+if ipv6_failed:
+    failed += ipv6_failed
+print("=" * 70)
+
+# ── CONNECT authority 解析：IPv6 字面量带方括号 ──────────────────
+print("\n" + "=" * 70)
+print("CONNECT authority 解析（IPv6 字面量剥方括号）")
+print("=" * 70)
+
+conn_cases = [
+    ("CONNECT [::1]:443 HTTP/1.1",           ("::1", 443)),
+    ("CONNECT example.com:443 HTTP/1.1",     ("example.com", 443)),
+    ("CONNECT iam.aeolianchi.me:443 HTTP/1.1", ("iam.aeolianchi.me", 443)),
+    ("CONNECT [::1] HTTP/1.1",               None),   # 无端口
+    ("CONNECT example.com HTTP/1.1",         None),   # 无端口
+    ("CONNECT [fe80::1%25eth0]:8443 HTTP/1.1", ("fe80::1%25eth0", 8443)),
+    ("CONNECT 127.0.0.1:1080 HTTP/1.1",      ("127.0.0.1", 1080)),
+]
+conn_failed = 0
+for line, expect in conn_cases:
+    got = pp._parse_connect_authority(line)
+    ok = got == expect
+    conn_failed += 0 if ok else 1
+    print(f"  {line:<45} -> {got} (期望 {expect}) {'PASS' if ok else 'FAIL'}")
+if conn_failed:
+    failed += conn_failed
+print("=" * 70)
+
+# ── 上游代理认证注入：按完整头名精确匹配 ────────────────────────
+print("\n" + "=" * 70)
+print("认证注入：仅精确的 authorization 头名阻止注入")
+print("=" * 70)
+
+def _auth_lines(data: bytes):
+    head = data.split(b"\r\n\r\n", 1)[0]
+    return [l for l in head.split(b"\r\n")[1:] if l.lower().startswith(b"authorization:")]
+
+auth_cases = [
+    # (描述, 请求, 期望)：数字=注入后 authorization 行数；"unchanged"=原样跳过
+    ("无任何认证头",      b"CONNECT iam.aeolianchi.me:443 HTTP/1.1\r\nHost: iam.aeolianchi.me\r\n\r\n", 1),
+    ("Proxy-Authorization", b"CONNECT iam.aeolianchi.me:443 HTTP/1.1\r\nProxy-Authorization: Basic xyz\r\n\r\n", 1),
+    ("已有 Authorization", b"GET http://example.com/ HTTP/1.1\r\nAuthorization: Bearer tok\r\n\r\n", "unchanged"),
+    ("小写 authorization", b"GET http://example.com/ HTTP/1.1\r\nauthorization: basic abc\r\n\r\n", "unchanged"),
+    ("X-Authorization 前缀", b"CONNECT iam.aeolianchi.me:443 HTTP/1.1\r\nX-Authorization-Go: 1\r\n\r\n", 1),
+    ("无头空块",          b"GET http://example.com/ HTTP/1.1\r\n\r\n", 1),
+]
+auth_failed = 0
+for desc, req, expect in auth_cases:
+    out = pp._inject_proxy_auth(req)
+    if expect == "unchanged":
+        ok = out == req
+        print(f"  {desc:<22} 原样跳过 {'PASS' if ok else 'FAIL'}")
+    else:
+        lines = _auth_lines(out)
+        ok = len(lines) == expect and out.split(b"\r\n", 2)[1] == lines[0]
+        # 注入位置必须紧跟请求行（上游取首个 Authorization 时读到我们的认证）
+        print(f"  {desc:<22} authorization 行数={len(lines)} (期望 {expect}) "
+              f"{'PASS' if ok else 'FAIL'}")
+    auth_failed += 0 if ok else 1
+if auth_failed:
+    failed += auth_failed
+print("=" * 70)
+
+# ── 本程序 PAC 地址判定（启动清理只认自己的残留） ────────────────
+print("\n" + "=" * 70)
+print("PAC 地址归属判定（_is_our_pac_url）")
+print("=" * 70)
+
+pac_cases = [
+    ("http://127.0.0.1:39453/proxy.pac", True),     # 本程序 PAC 服务
+    ("http://127.0.0.1:7897/pac?auth=xyz", False),  # 其他代理软件（如 Clash）
+    ("http://127.0.0.1:1080/other.pac", False),     # 路径不符
+    ("http://192.168.1.1:8080/proxy.pac", False),   # 非本机
+    ("", False),
+]
+pac_failed = 0
+for url, expect in pac_cases:
+    got = pp._is_our_pac_url(url)
+    ok = got == expect
+    pac_failed += 0 if ok else 1
+    print(f"  {url:<42} 命中={got} (期望 {expect}) {'PASS' if ok else 'FAIL'}")
+if pac_failed:
+    failed += pac_failed
+print("=" * 70)
+
+# ── DNS 分流 TTL 缓存 ────────────────────────────────────────
+print("\n" + "=" * 70)
+print("TTL 缓存：cache_clear 兼容 + 缓存命中")
+print("=" * 70)
+
+ttl_failed = 0
+assert hasattr(pp._is_china_ip, "cache_clear"), "cache_clear 缺失"
+assert hasattr(pp._is_private_ip, "cache_clear"), "cache_clear 缺失"
+assert hasattr(pp._is_whitelisted, "cache_clear"), "cache_clear 缺失"
+pp._is_china_ip.cache_clear()
+pp._is_private_ip.cache_clear()
+pp._is_whitelisted.cache_clear()
+a1, a2 = pp._is_private_ip("10.1.2.3"), pp._is_private_ip("10.1.2.3")
+ok = a1 is True and a2 is True
+ttl_failed += 0 if ok else 1
+print(f"  内网 IP 判定缓存命中: {a1}/{a2} {'PASS' if ok else 'FAIL'}")
+b1, b2 = pp._is_china_ip("114.114.114.114"), pp._is_china_ip("114.114.114.114")
+ok = b1 is True and b2 is True
+ttl_failed += 0 if ok else 1
+print(f"  中国 IP 判定缓存命中: {b1}/{b2} {'PASS' if ok else 'FAIL'}")
+pp._is_china_ip.cache_clear()
+pp._is_private_ip.cache_clear()
+print(f"  cache_clear 后重查: {pp._is_china_ip('114.114.114.114')} "
+      f"{'PASS' if pp._is_china_ip('114.114.114.114') is True else 'FAIL'}")
+if ttl_failed:
+    failed += ttl_failed
+print("=" * 70)

server_config/README.md

@@ -1,26 +1,457 @@
-# server_config
+# RRPP 服务端部署规范
+
+> 一套统一的跨境代理节点部署流程，适用于 Debian/Ubuntu 系 Linux。
+> 每台节点运行相同的服务栈：**BBR 拥塞控制 + nginx CONNECT 隧道 + sing-box(VLESS+REALITY / Shadowsocks) + mtg(MTProto)**。
+>
+> 本文档为**通用规范**，所有 IP / 域名 / 端口 / 密钥均以**占位符**表示，部署时替换为实际值。
+> 凡标注 `<...>` 的均为占位符，不要原样照抄。
+
+---
+
+## 目录
+
+- [0. 架构总览](#0-架构总览)
+- [1. 端口与服务规划](#1-端口与服务规划)
+- [2. 系统基础优化（BBR 跨境丢包优化）](#2-系统基础优化bbr-跨境丢包优化)
+- [3. 伪装站点页面](#3-伪装站点页面)
+- [4. HTTPS 证书（acme.sh DNS-01）](#4-https-证书acmesh-dns-01)
+- [5. nginx 源码编译 + CONNECT 隧道](#5-nginx-源码编译--connect-隧道)
+- [6. sing-box（VLESS+REALITY / Shadowsocks）](#6-sing-boxvlessreality--shadowsocks)
+- [7. mtg（MTProto 代理）](#7-mtgmtproto-代理)
+- [8. 本机公网 IP 自动识别（防自代理回环）](#8-本机公网-ip-自动识别防自代理回环)
+- [9. systemd 服务文件](#9-systemd-服务文件)
+- [10. 启动顺序与验证清单](#10-启动顺序与验证清单)
+- [11. 目录结构与文件索引](#11-目录结构与文件索引)
+
+---
+
+## 0. 架构总览
+
+```
+客户端 ──► nginx(443/<nginx端口>) ──┬─► CONNECT 隧道(任意端口, 全端口放行)
+        │                          └─► 伪装页面(直接访问/IP访问)
+        │
+        ├─► sing-box(<vless端口>) ──► VLESS+REALITY(vision)
+        ├─► sing-box(<ss端口>)    ──► Shadowsocks(2022-blake3-aes-256-gcm)
+        └─► mtg(<mtproto端口>)    ──► MTProto(Fake TLS, Telegram 内置代理)
+```
+
+**四层代理各司其职：**
+
+| 层 | 组件 | 协议 | 适用场景 |
+|---|---|---|---|
+| HTTP 隧道 | nginx | CONNECT + Basic Auth | 浏览器/curl/系统代理，全端口 WebSocket 透传 |
+| 通用代理 | sing-box | VLESS+REALITY | Shadowrocket/Clash/v2rayN 客户端 |
+| 兼容旧客户端 | sing-box | Shadowsocks | Shadowrocket 2.x 等不支持 REALITY 的客户端 || Telegram 专用 | mtg | MTProto Fake TLS | Telegram 内置代理，抗封锁 |
+
+**核心密钥体系（2026-09 起全部每节点独立）：**
+- VLESS+REALITY：每节点**独立** UUID、REALITY 密钥对、short_id；SNI 仍全节点统一（用 `<伪装SNI域名>`，选一个高流量、TLS 指纹稳定的知名站点域名）
+- Shadowsocks：每节点**独立** PSK，方法 `2022-blake3-aes-256-gcm`（SIP022，抗重放探测）
+- nginx CONNECT：每节点**独立** Basic Auth 密码（客户端按节点携带凭据，无兼容性问题）
+- MTProto：每台**独立 secret**（Fake TLS 类型，fronting 到 `<伪装SNI域名>`）
+
+> 独立化的目的：单节点密钥泄漏/被污染时不波及其余节点，且换钥粒度更小。
+> 代价是订阅/nodes.json 必须逐节点维护，改任一项后必须做端到端验证（见 6.3、第 10 节）。
+
+---
+
+## 1. 端口与服务规划
+
+端口全部落在**同一区段**便于防火墙管理（示例用 1200–1220，可按需调整）：
+
+| 端口 | 服务 | 说明 |
+|---|---|---|
+| `<nginx端口>` | nginx | HTTPS 站点 + CONNECT 隧道（示例 `1205`） |
+| `<vless端口>` | sing-box | VLESS+REALITY（示例 `1206`） |
+| `<ss端口>` | sing-box | Shadowsocks（示例 `1207`） |
+| `<mtproto端口>` | mtg | MTProto（示例 `1210`） |
+
+> 节点 A 用 `80xx` 区段、节点 B 用 `120xx` 区段皆可，**只要同节点内不冲突、且运营商/防火墙放行**。
+> 防火墙放行：`iptables -A INPUT -p tcp -m multiport --dports <nginx端口>,<vless端口>,<ss端口>,<mtproto端口> -j ACCEPT`
+
+**域名规划：** 每个节点一个独立域名 `<节点域名>`（如 `node1.example.com`），A 记录指向节点公网 IP。acme.sh 用 DNS-01 签证书，无需 80 端口可达。
+
+---
+
+## 2. 系统基础优化（BBR 跨境丢包优化）
+
+BBR（Bottleneck Bandwidth and RTT）+ fq 队列，显著改善跨境 TCP 丢包/高延迟下的吞吐。
+
+### 2.1 启用 BBR
+
+写入 `/etc/sysctl.d/99-bbr.conf`：
+
+```conf
+net.core.default_qdisc=fq
+net.ipv4.tcp_congestion_control=bbr
+```
+
+应用并验证：
+
+```bash
+modprobe tcp_bbr 2>/dev/null || true   # 部分内核需手动加载模块
+sysctl --system
+sysctl net.ipv4.tcp_congestion_control net.core.default_qdisc
+# 期望输出: bbr / fq
+```
+
+> **内核要求**：BBR 需 Linux 4.9+。`default_qdisc=fq` 在部分新内核（6.12+ cloud 内核）可能无此参数，仅保留 `tcp_congestion_control=bbr` 即可，不影响 BBR 生效。
+
+### 2.2 文件描述符
+
+nginx worker 高并发需要提高 limit，在 `/etc/security/limits.conf` 追加：
+
+```
+* soft nofile 65535
+* hard nofile 65535
+```
+
+### 2.3 时区与 NTP
+
+代理证书校验、日志排障都依赖准确时间：
+
+```bash
+timedatectl set-timezone Asia/Shanghai
+apt install -y chrony && systemctl enable --now chrony
+```
+
+### 2.4 PMTUD 黑洞缓解（tcp_mtu_probing）
+
+**症状**：部分客户端网络上出现"TCP 能连上、但 TLS 握手稳定超时"
+（如 `_ssl.c:1011 handshake operation timed out`），而服务器从其他网络探测一切正常。
+原因是客户端↔服务器路径上某跳 MTU 变小且 ICMP Fragmentation Needed 被过滤
+（PMTUD 黑洞），服务器发出的满尺寸大包永远到不了客户端。
+
+开启 RFC 4821 PLPMTUD，让服务器在探测失败后自动降低 MSS 恢复通信。
+写入 `/etc/sysctl.d/99-mtu-probing.conf`：
+
+```conf
+net.ipv4.tcp_mtu_probing = 1
+```
+
+```bash
+sysctl --system
+sysctl net.ipv4.tcp_mtu_probing   # 期望输出: 1
+```
+
+> 只影响命中黑洞的连接，正常连接无感知，可安全全节点开启。
+
+### 2.5 fail2ban（sshd 爆破防护）
+
+节点 SSH 暴露在公网，爆破尝试是常态（实测单节点累计 3 万+ 次）。装 fail2ban 护 sshd：
+
+```bash
+apt install -y fail2ban
+cat > /etc/fail2ban/jail.d/sshd.local << 'EOF'
+[sshd]
+enabled = true
+backend = systemd
+EOF
+systemctl enable --now fail2ban
+fail2ban-client status sshd   # 确认 jail 在跑
+```
+
+> - `backend = systemd`：无 `/var/log/auth.log` 的最小化系统（无 rsyslog）必须加这行，
+>   否则 fail2ban 启动报 "Have not found any log file for sshd jail"。
+> - LXC 容器也能正常封禁（实测 ban/unban 链路可用）。
+> - 只加 fail2ban、**不动 sshd 认证方式**（root 登录等现状保留）。
+> - 如需同时护 nginx CONNECT 的 401 爆破，可再加 nginx jail（凭据够强时可不做）。
+
+---
+
+## 3. 伪装站点页面
+
+直接访问节点/IP 时返回一个普通网页（而非代理错误），降低被探测的特征。
+
+放置一个静态 `index.html` 到 `/var/www/html/index.html`：
+- 任意抓取的合法 HTML 页面（博客/门户首页快照均可）
+- **所有节点用同一份**（md5 一致），避免逐节点特征差异
+
+```bash
+mkdir -p /var/www/html
+# index.html 为本仓库 fake-site/index.html 的拷贝
+cp fake-site/index.html /var/www/html/index.html
+```
+
+---
+
+## 4. HTTPS 证书（acme.sh DNS-01）
+
+nginx 站点与 CONNECT 隧道共用一张证书。用 [acme.sh](https://github.com/acmesh-official/acme.sh) + DNS-01（以阿里云 DNS 为例，其他 DNS 商同理）：
+
+```bash
+curl https://get.acme.sh | sh -s email=<你的邮箱>
+export Ali_Key="<阿里云AccessKey>"
+export Ali_Secret="<阿里云Secret>"
+
+~/.acme.sh/acme.sh --issue --dns dns_ali -d <节点域名> --keylength ec-256
+
+# 安装到 nginx 读取路径
+mkdir -p /etc/nginx/ssl/<节点域名>
+~/.acme.sh/acme.sh --install-cert -d <节点域名> --ecc \
+  --fullchain-file /etc/nginx/ssl/<节点域名>/fullchain.pem \
+  --key-file       /etc/nginx/ssl/<节点域名>/privkey.pem
+```
+
+acme.sh 自带 cron 自动续期。证书路径固定为 `/etc/nginx/ssl/<节点域名>/{fullchain,privkey}.pem`。
+
+---
+
+## 5. nginx 源码编译 + CONNECT 隧道
+
+标准 nginx 不自带 CONNECT 方法支持，需打 [ngx_http_proxy_connect_module](https://github.com/chobits/ngx_http_proxy_connect_module) 补丁后**源码编译**。
+
+### 5.1 编译
+
+```bash
+apt install -y build-essential libssl-dev libpcre3-dev zlib1g-dev curl wget
+
+NGINX_VER=1.26.3
+cd /usr/local/src
+curl -LO http://nginx.org/download/nginx-$NGINX_VER.tar.gz
+tar xzf nginx-$NGINX_VER.tar.gz
+
+git clone https://github.com/chobits/ngx_http_proxy_connect_module.git
+cd nginx-$NGINX_VER
+patch -p1 < ../ngx_http_proxy_connect_module/patch/proxy_connect_rewrite_102101.patch
+
+./configure \
+  --prefix=/usr/local/nginx \
+  --conf-path=/etc/nginx/nginx.conf \
+  --pid-path=/var/run/nginx.pid \
+  --error-log-path=/var/log/nginx/error.log \
+  --http-log-path=/var/log/nginx/access.log \
+  --with-http_ssl_module \
+  --with-http_v2_module \
+  --with-http_realip_module \
+  --with-http_stub_status_module \
+  --with-http_gzip_static_module \
+  --add-module=../ngx_http_proxy_connect_module
+make -j$(nproc) && make install
+```
+
+> 若发行版 nginx 已自带 proxy_connect 补丁（极少见），可用 `apt install nginx` 跳过编译，但源码编译是通用做法。
+
+### 5.2 Basic Auth 凭据
+
+CONNECT 代理需要 Basic Auth（用户名/密码），**每节点独立密码**（客户端配置按节点携带凭据，无兼容性问题）。生成 htpasswd：
+
+```bash
+apt install -y apache2-utils
+htpasswd -bc /etc/nginx/.htpasswd <代理用户名> '<本节点独立密码>'
+chmod 644 /etc/nginx/.htpasswd   # worker 以 nobody 运行，必须可读
+```
+
+> 无 apache2-utils 的机器可用 openssl 生成（sha512-crypt，nginx 同样支持）：
+> `echo "<用户名>:$(openssl passwd -6 '<密码>')" > /etc/nginx/.htpasswd`
+> htpasswd 是**每次请求实时读取**的，改密码无需 reload nginx。
+
+### 5.3 nginx.conf
+
+完整配置见 [nginx/nginx.conf](nginx/nginx.conf)。**关键设计**：
+
+- **两个 server 块共用 `<nginx端口>`**：
+  - 主块 `server_name <节点域名>`：证书站点 + `location /` 返回伪装页；非本域名一律 rewrite 到 `/__proxy` 转发
+  - fallback 块 `default_server`：其他 SNI 也接受 CONNECT 代理；本机/私网/公网 IP 直连返回伪装页，其余走 `/__proxy`
+- `proxy_connect_allow all;`：**全端口** CONNECT 放行（codex/WebSocket 走任意端口均通）。若需限制，改回 `proxy_connect_allow 443 563;`
+- `auth_basic`：仅 CONNECT/正向代理需要认证，`location /` 伪装页 `auth_basic off`
+- **自代理回环保护**：`/__proxy` 转发时带 `X-Proxy-Loop-Guard` 头，二次进入返回 508 防死循环
+- `/__local`：本机 IP 直连时剥前缀返回伪装页
+- `map $host $is_local`：私网/回环段判定；公网 IP 由 cron 脚本动态写入 `local_ip_rules.conf`
+
+### 5.4 自代理回环保护说明
+
+正向代理转发目标若指向**本机自己**（如域名公网 IP 变更、或代理目标就是本节点），会无限循环。两层防护：
+1. `map` 判定私网/回环段 + cron 维护的本机公网 IP 规则 → 直连伪装页
+2. `/__proxy` 的 Loop-Guard 头兜底 → 二次进入返回 508
+
+---
+
+## 6. sing-box（VLESS+REALITY / Shadowsocks）
+
+### 6.1 安装
+
+```bash
+SINGBOX_VER=1.13.19
+curl -LO https://github.com/SagerNet/sing-box/releases/download/v$SINGBOX_VER/sing-box-$SINGBOX_VER-linux-amd64.tar.gz
+tar xzf sing-box-*.tar.gz
+install -m 755 sing-box-$SINGBOX_VER-linux-amd64/sing-box /usr/local/bin/sing-box
+```
+
+### 6.2 配置
+
+完整配置见 [sing-box/config.json](sing-box/config.json)。要点：
+- **VLESS+REALITY**：每节点独立 UUID/密钥对/short_id，SNI 全节点统一，flow `xtls-rprx-vision`；REALITY handshake server 用 `<伪装SNI域名>:443`
+- **Shadowsocks**：`2022-blake3-aes-256-gcm`（SIP022：会话子密钥 + 抗重放探测，显著优于老 AEAD `aes-256-gcm`）+ 每节点独立 PSK，独立端口（兼容不支持 REALITY 的老客户端）
+  - **PSK 格式要求**：密码必须是 base64 编码的 32 字节密钥（`openssl rand -base64 32`），不是任意字符串；改错格式 sing-box 启动即报错
+  - sing-box / Shadowrocket / mihomo 均支持 2022-blake3 系列
+- 无 SOCKS5 inbound（已按需移除；如需可在 inbounds 追加 `type: socks`）
+
+### 6.3 REALITY 密钥生成
+
+**每台节点独立生成一套**（互不共用）：
+
+```bash
+# 生成私钥/公钥（sing-box 1.13+ 内置 reality 密钥生成）
+sing-box generate reality-keypair
+# 输出 private_key / public_key，记录备用
+# short_id 用 8 字节随机 hex：
+openssl rand -hex 8
+# UUID:
+sing-box generate uuid
+```
+
+> 私钥只放服务端配置；公钥给客户端订阅。UUID/密钥对/short_id 每节点独立。
+
+**公钥必须推导验证，禁止手抄。** 订阅/节点配置里的 `public_key` 与服务端
+`private_key` 必须严格配对（X25519 推导），曾发生手抄错一个字符导致全节点
+REALITY 握手失败（客户端报 `reality verification failed`）的事故。改配置后验证：
+
+```bash
+python3 - << 'EOF'
+import base64
+from cryptography.hazmat.primitives.asymmetric.x25519 import X25519PrivateKey
+from cryptography.hazmat.primitives import serialization
+priv = base64.urlsafe_b64decode('<REALITY_PRIVATE_KEY>' + '==')
+pub = X25519PrivateKey.from_private_bytes(priv).public_key().public_bytes(
+    serialization.Encoding.Raw, serialization.PublicFormat.Raw)
+print(base64.urlsafe_b64encode(pub).decode().rstrip('='))  # 应与下发的 public_key 逐字符一致
+EOF
+```
+
+---
+
+## 7. mtg（MTProto 代理）
+
+Telegram 内置代理，Fake TLS 模式抗封锁。用 [mtg](https://github.com/9seconds/mtg)（Go 单二进制）：
+
+### 7.1 安装
+
+```bash
+MTG_VER=2.2.8
+curl -LO https://github.com/9seconds/mtg/releases/download/v$MTG_VER/mtg-$MTG_VER-linux-amd64.tar.gz
+tar xzf mtg-*.tar.gz
+install -m 755 mtg-$MTG_VER-linux-amd64/mtg /usr/local/bin/mtg
+```
+
+### 7.2 生成 secret（每台独立）
+
+```bash
+mtg generate-secret --hex <伪装SNI域名>
+# 输出 ee<32hex><域名hex>，以 ee 开头 = Fake TLS 类型
+```
+
+### 7.3 运行
+
+```bash
+mtg simple-run --prefer-ip=prefer-ipv4 0.0.0.0:<mtproto端口> <上面生成的secret>
+```
+
+> `--prefer-ip=prefer-ipv4`：强制 IPv4 出站，避免容器/双栈环境下 IPv6 不可达导致回退失败。
+
+### 7.4 Telegram 客户端参数
+
+- 服务器：`<节点域名>`
+- 端口：`<mtproto端口>`
+- Secret：上面生成的 hex（Telegram 自动识别 Fake TLS）
+
+---
+
+## 8. 本机公网 IP 自动识别（防自代理回环）
+
+正向代理转发时，若目标 Host 是本机公网 IP，会被识别为"本机直连"返回伪装页（而非经代理转发本机，造成回环）。公网 IP 可能变动，用 cron 脚本每 5 分钟自动维护。
+
+脚本见 [scripts/sync-nginx-local-ips.sh](scripts/sync-nginx-local-ips.sh)，cron 见 [cron/sync-nginx-local-ips](cron/sync-nginx-local-ips)：
+
+```cron
+*/5 * * * * root /usr/local/sbin/sync-nginx-local-ips.sh >/dev/null 2>&1
+```
+
+脚本逻辑：取 `api.ipify.org` 出站 IP + 节点域名 A 记录，比对重写 `/etc/nginx/local_ip_rules.conf`，**仅变更时** `nginx -t && systemctl reload nginx`。
+
+---
+
+## 9. systemd 服务文件
+
+| 服务 | 文件 |
+|---|---|
+| nginx | [systemd/nginx.service](systemd/nginx.service) |
+| sing-box | [systemd/sing-box.service](systemd/sing-box.service) |
+| mtg | [systemd/mtg.service](systemd/mtg.service) |
+
+```bash
+cp systemd/*.service /etc/systemd/system/
+systemctl daemon-reload
+systemctl enable --now nginx sing-box mtg
+```
+
+---
+
+## 10. 启动顺序与验证清单
+
+### 启动顺序
+
+```bash
+sysctl --system                      # 1. BBR
+systemctl start sing-box             # 2. VLESS/SS
+systemctl start mtg                  # 3. MTProto
+systemctl start nginx                # 4. nginx（最后，依赖证书/伪装页/规则文件）
+```
+
+### 验证清单
+
+```bash
+# BBR
+sysctl net.ipv4.tcp_congestion_control   # => bbr
+sysctl net.ipv4.tcp_mtu_probing          # => 1（PMTUD 黑洞缓解，见 2.4）
+
+# fail2ban
+fail2ban-client status sshd              # => jail 存在（见 2.5）
+
+# sing-box
+sing-box check -c /etc/sing-box/config.json
+systemctl is-active sing-box
+
+# mtg（Fake TLS：非法 ClientHello 应返回 TLS alert 0x15）
+timeout 8 python3 -c "
+import socket,os,struct
+s=socket.connect(('<节点域名>',<mtproto端口>)) if False else socket.create_connection(('<节点域名>',<mtproto端口>),timeout=6)
+s.sendall(b'\x16\x03\x01'+struct.pack('>H',100)+os.urandom(100))
+print('resp:',s.recv(8).hex())"   # 期望首字节 15 (TLS alert)
+
+# nginx CONNECT（带 Basic Auth，B64 = base64('<代理用户名>:<代理密码>')）
+B64=$(printf '<代理用户名>:<代理密码>' | base64 -w0)
+curl -sk --proxy-insecure -x https://<节点域名>:<nginx端口> \
+  --proxy-header "Authorization: Basic $B64" \
+  -o /dev/null -w '%{http_code}\n' http://example.com/    # => 200
+
+# 伪装页（直连 IP/域名，无认证）
+curl -sk https://<节点域名>:<nginx端口>/ -o /dev/null -w '%{http_code}\n'   # => 200
+```
+
+---
+
+## 11. 目录结构与文件索引
 
 ```
 server_config/
+├── README.md                        ← 本文档
 ├── nginx/
-│   ├── nginx.conf
-│   ├── conf.d/
-│   │   ├── api.aeolianchi.me.conf
-│   │   └── autodiscover.aeolianchi.me.conf
-│   └── local_ip_rules.conf.example
+│   ├── nginx.conf                   ← 完整 nginx 配置（占位符版）
+│   ├── local_ip_rules.conf.example  ← 本机 IP 规则文件示例（cron 自动生成，勿手改）
+│   └── conf.d/                      ← 预留：额外站点配置
 ├── sing-box/
-│   └── config.json
-├── traffic-monitor/
-│   ├── config.json
-│   ├── traffic_monitor.py
-│   └── traffic-monitor.service
+│   └── config.json                  ← sing-box 配置（VLESS+REALITY + SS，占位符版）
+├── systemd/
+│   ├── nginx.service
+│   ├── sing-box.service
+│   └── mtg.service
 ├── scripts/
-│   ├── sync-nginx-local-ips.sh
-│   └── rrpp-stats
-├── pac/
-│   ├── proxy.pac
-│   └── proxy_wg.pac
-└── cron/
-    ├── crontab.txt
-    └── sync-nginx-local-ips
+│   └── sync-nginx-local-ips.sh      ← 本机公网 IP 规则自动维护
+├── cron/
+│   └── sync-nginx-local-ips         ← cron 定义（每 5 分钟）
+└── fake-site/
+    └── index.html                   ← 伪装站点页面（所有节点共用同一份）
 ```
+
+> 旧的 `traffic-monitor/`、`pac/`、`conf.d/*.conf` 已下线或迁移，不再维护，已从本目录移除。

server_config/cron/crontab.txt

@@ -1,2 +0,0 @@
-* * * * * /usr/local/bin/rrpp-stats | sudo tee /root/backups/stats.json > /dev/null
-27 0,6,12,18 * * * "/home/RRRosmontis/.acme.sh"/acme.sh --cron --home "/home/RRRosmontis/.acme.sh" > /dev/null

server_config/nginx/conf.d/api.aeolianchi.me.conf

@@ -1,19 +0,0 @@
-server {
-    listen 443 ssl;
-    http2 on;
-    server_name api.aeolianchi.me;
-
-    ssl_certificate     /etc/nginx/ssl/api.aeolianchi.me/fullchain.pem;
-    ssl_certificate_key /etc/nginx/ssl/api.aeolianchi.me/privkey.pem;
-    ssl_protocols TLSv1.2 TLSv1.3;
-    ssl_ciphers HIGH:!aNULL:!MD5;
-    ssl_session_cache shared:SSL:10m;
-    ssl_session_timeout 10m;
-
-    location /vpn/status/traffic {
-        proxy_pass http://127.0.0.1:18080;
-        proxy_set_header Host $host;
-        proxy_set_header X-Real-IP $remote_addr;
-        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-    }
-}

server_config/nginx/conf.d/autodiscover.aeolianchi.me.conf

@@ -1,44 +0,0 @@
-# autodiscover.aeolianchi.me -> jp 反代 -> bj autodiscover.rrrosmontis.icu
-server {
-    listen 80;
-    listen [::]:80;
-    server_name autodiscover.aeolianchi.me;
-    location /.well-known/acme-challenge/ {
-        root /var/www/acme;
-    }
-
-    location / {
-        return 301 https://$host$request_uri;
-    }
-}
-
-# autodiscover.aeolianchi.me 443 -> 反代 -> bj autodiscover.rrrosmontis.icu
-server {
-    listen 443 ssl;
-    http2 on;
-    server_name autodiscover.aeolianchi.me;
-
-    ssl_certificate    /etc/nginx/ssl/aeolianchi.me/fullchain.pem;
-    ssl_certificate_key    /etc/nginx/ssl/aeolianchi.me/privkey.pem;
-    ssl_protocols TLSv1.2 TLSv1.3;
-    ssl_ciphers EECDH+CHACHA20:EECDH+AES128:RSA+AES128:EECDH+AES256:RSA+AES256:!MD5;
-    ssl_prefer_server_ciphers on;
-    ssl_session_cache shared:SSL:10m;
-    ssl_session_timeout 10m;
-    add_header Strict-Transport-Security "max-age=31536000";
-    error_page 497 https://$host$request_uri;
-
-    location / {
-        proxy_pass https://autodiscover.rrrosmontis.icu;
-        proxy_set_header Host autodiscover.rrrosmontis.icu;
-        proxy_set_header X-Real-IP $remote_addr;
-        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-        proxy_set_header X-Forwarded-Proto $scheme;
-        proxy_http_version 1.1;
-        proxy_ssl_server_name on;
-        proxy_ssl_name autodiscover.rrrosmontis.icu;
-    }
-
-    access_log /var/log/nginx/autodiscover.aeolianchi.me.log;
-    error_log /var/log/nginx/autodiscover.aeolianchi.me.error.log;
-}

server_config/nginx/local_ip_rules.conf.example

@@ -1,4 +1,5 @@
 # nginx 本机公网 IP 直连规则（自动生成，勿手改；由 sync-nginx-local-ips.sh 维护）
 # 公网 IP 变更后本文件会被重写并 reload nginx
-if ($host = "<PUBLIC_IP>") { rewrite ^(.*)$ /__local$1 last; }
-if ($host = "<PUBLIC_IP>") { rewrite ^(.*)$ /__local$1 last; }
+# 每条规则对应一个本机公网 IP（出站 IP + 节点域名 A 记录）
+if ($host = "<PUBLIC_IP_1>") { rewrite ^(.*)$ /__local$1 last; }
+if ($host = "<PUBLIC_IP_2>") { rewrite ^(.*)$ /__local$1 last; }

server_config/nginx/nginx.conf

@@ -1,4 +1,3 @@
-load_module /usr/lib/nginx/modules/ngx_http_proxy_connect_module.so;
 worker_processes auto;
 worker_rlimit_nofile 65535;
 events {
@@ -25,45 +24,15 @@ http {
     keepalive_timeout 65;
     keepalive_requests 100;
 
-    # 80 端口：iam.aeolianchi.me 强制跳转到 HTTPS（保留 ACME 验证路径）
+    # <nginx端口>：主域名 HTTPS 站点 + 认证 CONNECT 代理
     server {
-        listen 80;
-        server_name iam.aeolianchi.me;
-
-        location /.well-known/acme-challenge/ {
-            root /var/www/acme;
-        }
-
-        location / {
-            return 301 https://$host$request_uri;
-        }
-    }
-
-    # 80 端口：默认站点只做静态伪装，不提供代理、不弹认证
-    server {
-        listen 80 default_server;
-        server_name _;
-
-        root /var/www/html;
-        index index.html;
-
-        location /.well-known/acme-challenge/ {
-            root /var/www/acme;
-        }
-
-        location / {
-            try_files $uri $uri/ =404;
-        }
-    }
-
-    server {
-        listen 443 ssl;
+        listen <nginx端口> ssl;
         http2 on;
-        server_name XXX;
+        server_name <节点域名>;
         root /var/www/html;
 
-        ssl_certificate /etc/nginx/ssl/XXX/fullchain.pem;
-        ssl_certificate_key /etc/nginx/ssl/XXX/privkey.pem;
+        ssl_certificate /etc/nginx/ssl/<节点域名>/fullchain.pem;
+        ssl_certificate_key /etc/nginx/ssl/<节点域名>/privkey.pem;
         ssl_protocols TLSv1.2 TLSv1.3;
         ssl_ciphers HIGH:!aNULL:!MD5;
         ssl_session_cache shared:SSL:10m;
@@ -74,14 +43,14 @@ http {
         auth_basic_user_file /etc/nginx/.htpasswd;
 
         proxy_connect;
-        proxy_connect_allow 443 563;
+        proxy_connect_allow all;
         proxy_connect_connect_timeout 60s;
         proxy_connect_read_timeout 120s;
         proxy_connect_send_timeout 120s;
 
         location / {
             auth_basic off;
-            if ($host != "iam.aeolianchi.me") {
+            if ($host != "<节点域名>") {
                 rewrite ^ /__proxy last;
             }
             index index.html;
@@ -104,14 +73,14 @@ http {
         }
     }
 
-    # 443 端口 fallback：其他 SNI 也可用 CONNECT 代理；普通 GET 只返回伪装页
+    # <nginx端口> fallback：其他 SNI 也可用 CONNECT 代理；普通 GET 只返回伪装页
     server {
-        listen 443 ssl default_server;
+        listen <nginx端口> ssl default_server;
         http2 on;
         server_name _;
 
-        ssl_certificate /etc/nginx/ssl/iam.aeolianchi.me/fullchain.pem;
-        ssl_certificate_key /etc/nginx/ssl/iam.aeolianchi.me/privkey.pem;
+        ssl_certificate /etc/nginx/ssl/<节点域名>/fullchain.pem;
+        ssl_certificate_key /etc/nginx/ssl/<节点域名>/privkey.pem;
         ssl_protocols TLSv1.2 TLSv1.3;
         ssl_ciphers HIGH:!aNULL:!MD5;
         ssl_session_cache shared:SSL:10m;
@@ -122,7 +91,7 @@ http {
         auth_basic_user_file /etc/nginx/.htpasswd;
 
         proxy_connect;
-        proxy_connect_allow 443 563;
+        proxy_connect_allow all;
         proxy_connect_connect_timeout 60s;
         proxy_connect_read_timeout 120s;
         proxy_connect_send_timeout 120s;

server_config/pac/proxy.pac

@@ -1,647 +0,0 @@
-function FindProxyForURL(url, host) {
-    // RRPP 规则代理列表 — 仅以下域名走加密隧道
-    var proxy = "PROXY 127.0.0.1:1080";
-    // Google 全系 + Gemini + CDN
-    if (shExpMatch(host, "google.com") ||
-        shExpMatch(host, "*.google.com") ||
-        shExpMatch(host, "google.com.hk") ||
-        shExpMatch(host, "*.google.com.hk") ||
-        shExpMatch(host, "google.co.jp") ||
-        shExpMatch(host, "*.google.co.jp") ||
-        shExpMatch(host, "googleapis.com") ||
-        shExpMatch(host, "*.googleapis.com") ||
-        shExpMatch(host, "googleusercontent.com") ||
-        shExpMatch(host, "*.googleusercontent.com") ||
-        shExpMatch(host, "googletagmanager.com") ||
-        shExpMatch(host, "*.googletagmanager.com") ||
-        shExpMatch(host, "google-analytics.com") ||
-        shExpMatch(host, "*.google-analytics.com") ||
-        shExpMatch(host, "gstatic.com") ||
-        shExpMatch(host, "*.gstatic.com") ||
-        shExpMatch(host, "ggpht.com") ||
-        shExpMatch(host, "*.ggpht.com") ||
-        shExpMatch(host, "googlehosted.com") ||
-        shExpMatch(host, "*.googlehosted.com") ||
-        shExpMatch(host, "1e100.net") ||
-        shExpMatch(host, "*.1e100.net") ||
-        shExpMatch(host, "youtube.com") ||
-        shExpMatch(host, "*.youtube.com") ||
-        shExpMatch(host, "ytimg.com") ||
-        shExpMatch(host, "*.ytimg.com") ||
-        shExpMatch(host, "googlevideo.com") ||
-        shExpMatch(host, "*.googlevideo.com") ||
-        shExpMatch(host, "gemini.google.com") ||
-        shExpMatch(host, "*.gemini.google.com") ||
-        shExpMatch(host, "generativeai.googleapis.com") ||
-        shExpMatch(host, "*.generativeai.googleapis.com") ||
-        shExpMatch(host, "aistudio.google.com") ||
-        shExpMatch(host, "*.aistudio.google.com") ||
-        shExpMatch(host, "deepmind.google.com") ||
-        shExpMatch(host, "*.deepmind.google.com") ||
-        shExpMatch(host, "deepmind.com") ||
-        shExpMatch(host, "*.deepmind.com")) {
-        return proxy;
-    }
-
-    // OpenAI / ChatGPT 全系 + CDN
-    if (shExpMatch(host, "openai.com") ||
-        shExpMatch(host, "*.openai.com") ||
-        shExpMatch(host, "chatgpt.com") ||
-        shExpMatch(host, "*.chatgpt.com") ||
-        shExpMatch(host, "oaistatic.com") ||
-        shExpMatch(host, "*.oaistatic.com") ||
-        shExpMatch(host, "oaiusercontent.com") ||
-        shExpMatch(host, "*.oaiusercontent.com") ||
-        shExpMatch(host, "openaiusercontent.com") ||
-        shExpMatch(host, "*.openaiusercontent.com")) {
-        return proxy;
-    }
-
-    // Grok / xAI 全系 + CDN
-    if (shExpMatch(host, "grok.com") ||
-        shExpMatch(host, "*.grok.com") ||
-        shExpMatch(host, "x.com") ||
-        shExpMatch(host, "*.x.com") ||
-        shExpMatch(host, "twitter.com") ||
-        shExpMatch(host, "*.twitter.com") ||
-        shExpMatch(host, "twimg.com") ||
-        shExpMatch(host, "*.twimg.com") ||
-        shExpMatch(host, "x.ai") ||
-        shExpMatch(host, "*.x.ai") ||
-        shExpMatch(host, "xai.com") ||
-        shExpMatch(host, "*.xai.com")) {
-        return proxy;
-    }
-
-    // Meta 全系 + CDN
-    if (shExpMatch(host, "meta.com") ||
-        shExpMatch(host, "*.meta.com") ||
-        shExpMatch(host, "meta.ai") ||
-        shExpMatch(host, "*.meta.ai") ||
-        shExpMatch(host, "facebook.com") ||
-        shExpMatch(host, "*.facebook.com") ||
-        shExpMatch(host, "fb.com") ||
-        shExpMatch(host, "*.fb.com") ||
-        shExpMatch(host, "fbcdn.net") ||
-        shExpMatch(host, "*.fbcdn.net") ||
-        shExpMatch(host, "fbsbx.com") ||
-        shExpMatch(host, "*.fbsbx.com") ||
-        shExpMatch(host, "instagram.com") ||
-        shExpMatch(host, "*.instagram.com") ||
-        shExpMatch(host, "cdninstagram.com") ||
-        shExpMatch(host, "*.cdninstagram.com") ||
-        shExpMatch(host, "threads.net") ||
-        shExpMatch(host, "*.threads.net") ||
-        shExpMatch(host, "whatsapp.com") ||
-        shExpMatch(host, "*.whatsapp.com") ||
-        shExpMatch(host, "whatsapp.net") ||
-        shExpMatch(host, "*.whatsapp.net") ||
-        shExpMatch(host, "messenger.com") ||
-        shExpMatch(host, "*.messenger.com")) {
-        return proxy;
-    }
-
-    // Claude / Anthropic 全系 + CDN
-    if (shExpMatch(host, "anthropic.com") ||
-        shExpMatch(host, "*.anthropic.com") ||
-        shExpMatch(host, "claude.ai") ||
-        shExpMatch(host, "*.claude.ai") ||
-        shExpMatch(host, "claudeusercontent.com") ||
-        shExpMatch(host, "*.claudeusercontent.com")) {
-        return proxy;
-    }
-
-    // Dify 全系
-    if (shExpMatch(host, "dify.ai") ||
-        shExpMatch(host, "*.dify.ai") ||
-        shExpMatch(host, "udify.app") ||
-        shExpMatch(host, "*.udify.app")) {
-        return proxy;
-    }
-
-    // Cloudflare 全系
-    if (shExpMatch(host, "cloudflare.com") ||
-        shExpMatch(host, "*.cloudflare.com") ||
-        shExpMatch(host, "cloudflare.net") ||
-        shExpMatch(host, "*.cloudflare.net") ||
-        shExpMatch(host, "cloudflareinsights.com") ||
-        shExpMatch(host, "*.cloudflareinsights.com") ||
-        shExpMatch(host, "cloudflarestatus.com") ||
-        shExpMatch(host, "*.cloudflarestatus.com") ||
-        shExpMatch(host, "cloudflarecdn.com") ||
-        shExpMatch(host, "*.cloudflarecdn.com") ||
-        shExpMatch(host, "cf-cdn.com") ||
-        shExpMatch(host, "*.cf-cdn.com") ||
-        shExpMatch(host, "workers.dev") ||
-        shExpMatch(host, "*.workers.dev") ||
-        shExpMatch(host, "pages.dev") ||
-        shExpMatch(host, "*.pages.dev")) {
-        return proxy;
-    }
-
-    // GitHub 全系
-    if (shExpMatch(host, "github.com") ||
-        shExpMatch(host, "*.github.com") ||
-        shExpMatch(host, "githubusercontent.com") ||
-        shExpMatch(host, "*.githubusercontent.com") ||
-        shExpMatch(host, "githubassets.com") ||
-        shExpMatch(host, "*.githubassets.com") ||
-        shExpMatch(host, "github.io") ||
-        shExpMatch(host, "*.github.io") ||
-        shExpMatch(host, "githubcopilot.com") ||
-        shExpMatch(host, "*.githubcopilot.com") ||
-        shExpMatch(host, "ghcr.io") ||
-        shExpMatch(host, "*.ghcr.io")) {
-        return proxy;
-    }
-
-    // Hugging Face 全系 + CDN
-    if (shExpMatch(host, "huggingface.co") ||
-        shExpMatch(host, "*.huggingface.co") ||
-        shExpMatch(host, "hf.co") ||
-        shExpMatch(host, "*.hf.co") ||
-        shExpMatch(host, "huggingfacecdn.com") ||
-        shExpMatch(host, "*.huggingfacecdn.com")) {
-        return proxy;
-    }
-
-    // Microsoft Copilot 全系 + Bing + Azure CDN
-    if (shExpMatch(host, "microsoft.com") ||
-        shExpMatch(host, "*.microsoft.com") ||
-        shExpMatch(host, "bing.com") ||
-        shExpMatch(host, "*.bing.com") ||
-        shExpMatch(host, "bingapis.com") ||
-        shExpMatch(host, "*.bingapis.com") ||
-        shExpMatch(host, "bing.net") ||
-        shExpMatch(host, "*.bing.net") ||
-        shExpMatch(host, "live.com") ||
-        shExpMatch(host, "*.live.com") ||
-        shExpMatch(host, "microsoftonline.com") ||
-        shExpMatch(host, "*.microsoftonline.com") ||
-        shExpMatch(host, "microsoft365.com") ||
-        shExpMatch(host, "*.microsoft365.com") ||
-        shExpMatch(host, "office.com") ||
-        shExpMatch(host, "*.office.com") ||
-        shExpMatch(host, "office.net") ||
-        shExpMatch(host, "*.office.net") ||
-        shExpMatch(host, "azure.com") ||
-        shExpMatch(host, "*.azure.com") ||
-        shExpMatch(host, "azureedge.net") ||
-        shExpMatch(host, "*.azureedge.net") ||
-        shExpMatch(host, "azurefd.net") ||
-        shExpMatch(host, "*.azurefd.net") ||
-        shExpMatch(host, "msn.com") ||
-        shExpMatch(host, "*.msn.com") ||
-        shExpMatch(host, "msecnd.net") ||
-        shExpMatch(host, "*.msecnd.net") ||
-        shExpMatch(host, "msftncsi.com") ||
-        shExpMatch(host, "*.msftncsi.com") ||
-        shExpMatch(host, "s-microsoft.com") ||
-        shExpMatch(host, "*.s-microsoft.com")) {
-        return proxy;
-    }
-
-    // Steam 全系 + CDN / 社区 / 下载
-    if (shExpMatch(host, "steampowered.com") ||
-        shExpMatch(host, "*.steampowered.com") ||
-        shExpMatch(host, "steamcommunity.com") ||
-        shExpMatch(host, "*.steamcommunity.com") ||
-        shExpMatch(host, "steam.com") ||
-        shExpMatch(host, "*.steam.com") ||
-        shExpMatch(host, "steamgames.com") ||
-        shExpMatch(host, "*.steamgames.com") ||
-        shExpMatch(host, "steamusercontent.com") ||
-        shExpMatch(host, "*.steamusercontent.com") ||
-        shExpMatch(host, "steamcontent.com") ||
-        shExpMatch(host, "*.steamcontent.com") ||
-        shExpMatch(host, "steamstatic.com") ||
-        shExpMatch(host, "*.steamstatic.com") ||
-        shExpMatch(host, "steam-chat.com") ||
-        shExpMatch(host, "*.steam-chat.com") ||
-        shExpMatch(host, "steam-api.com") ||
-        shExpMatch(host, "*.steam-api.com") ||
-        shExpMatch(host, "steamserver.net") ||
-        shExpMatch(host, "*.steamserver.net") ||
-        shExpMatch(host, "steamcdn-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamcdn-a.akamaihd.net") ||
-        shExpMatch(host, "steamcommunity-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamcommunity-a.akamaihd.net") ||
-        shExpMatch(host, "steamstore-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamstore-a.akamaihd.net") ||
-        shExpMatch(host, "steamuserimages-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamuserimages-a.akamaihd.net") ||
-        shExpMatch(host, "akamaihd.net") ||
-        shExpMatch(host, "*.akamaihd.net") ||
-        shExpMatch(host, "valvesoftware.com") ||
-        shExpMatch(host, "*.valvesoftware.com")) {
-        return proxy;
-    }
-
-    // Discord 全系
-    if (shExpMatch(host, "discord.com") ||
-        shExpMatch(host, "*.discord.com") ||
-        shExpMatch(host, "discordapp.com") ||
-        shExpMatch(host, "*.discordapp.com") ||
-        shExpMatch(host, "discord.gg") ||
-        shExpMatch(host, "*.discord.gg") ||
-        shExpMatch(host, "discord.media") ||
-        shExpMatch(host, "*.discord.media") ||
-        shExpMatch(host, "discordapp.net") ||
-        shExpMatch(host, "*.discordapp.net")) {
-        return proxy;
-    }
-
-    // Telegram 全系
-    if (shExpMatch(host, "telegram.org") ||
-        shExpMatch(host, "*.telegram.org") ||
-        shExpMatch(host, "t.me") ||
-        shExpMatch(host, "*.t.me") ||
-        shExpMatch(host, "telega.one") ||
-        shExpMatch(host, "*.telega.one") ||
-        shExpMatch(host, "telegram-cdn.org") ||
-        shExpMatch(host, "*.telegram-cdn.org")) {
-        return proxy;
-    }
-
-    // Reddit 全系
-    if (shExpMatch(host, "reddit.com") ||
-        shExpMatch(host, "*.reddit.com") ||
-        shExpMatch(host, "redd.it") ||
-        shExpMatch(host, "*.redd.it") ||
-        shExpMatch(host, "redditmedia.com") ||
-        shExpMatch(host, "*.redditmedia.com") ||
-        shExpMatch(host, "redditstatic.com") ||
-        shExpMatch(host, "*.redditstatic.com")) {
-        return proxy;
-    }
-
-    // LINE 全系
-    if (shExpMatch(host, "line.me") ||
-        shExpMatch(host, "*.line.me") ||
-        shExpMatch(host, "line-apps.com") ||
-        shExpMatch(host, "*.line-apps.com") ||
-        shExpMatch(host, "line-scdn.net") ||
-        shExpMatch(host, "*.line-scdn.net")) {
-        return proxy;
-    }
-
-    // Spotify 全系
-    if (shExpMatch(host, "spotify.com") ||
-        shExpMatch(host, "*.spotify.com") ||
-        shExpMatch(host, "spotifycdn.com") ||
-        shExpMatch(host, "*.spotifycdn.com") ||
-        shExpMatch(host, "scdn.co") ||
-        shExpMatch(host, "*.scdn.co")) {
-        return proxy;
-    }
-
-    // Netflix 全系
-    if (shExpMatch(host, "netflix.com") ||
-        shExpMatch(host, "*.netflix.com") ||
-        shExpMatch(host, "nflxvideo.net") ||
-        shExpMatch(host, "*.nflxvideo.net") ||
-        shExpMatch(host, "nflximg.net") ||
-        shExpMatch(host, "*.nflximg.net") ||
-        shExpMatch(host, "nflxext.com") ||
-        shExpMatch(host, "*.nflxext.com")) {
-        return proxy;
-    }
-
-    // Disney+ 全系
-    if (shExpMatch(host, "disneyplus.com") ||
-        shExpMatch(host, "*.disneyplus.com") ||
-        shExpMatch(host, "disney-plus.net") ||
-        shExpMatch(host, "*.disney-plus.net") ||
-        shExpMatch(host, "dssott.com") ||
-        shExpMatch(host, "*.dssott.com")) {
-        return proxy;
-    }
-
-    // TikTok 全系
-    if (shExpMatch(host, "tiktok.com") ||
-        shExpMatch(host, "*.tiktok.com") ||
-        shExpMatch(host, "tiktokcdn.com") ||
-        shExpMatch(host, "*.tiktokcdn.com") ||
-        shExpMatch(host, "tiktokv.com") ||
-        shExpMatch(host, "*.tiktokv.com")) {
-        return proxy;
-    }
-
-    // PlayStation 全系
-    if (shExpMatch(host, "playstation.com") ||
-        shExpMatch(host, "*.playstation.com") ||
-        shExpMatch(host, "sonyentertainmentnetwork.com") ||
-        shExpMatch(host, "*.sonyentertainmentnetwork.com")) {
-        return proxy;
-    }
-
-    // Nintendo 全系
-    if (shExpMatch(host, "nintendo.com") ||
-        shExpMatch(host, "*.nintendo.com") ||
-        shExpMatch(host, "nintendowifi.net") ||
-        shExpMatch(host, "*.nintendowifi.net")) {
-        return proxy;
-    }
-
-    // Epic Games 全系
-    if (shExpMatch(host, "epicgames.com") ||
-        shExpMatch(host, "*.epicgames.com") ||
-        shExpMatch(host, "unrealengine.com") ||
-        shExpMatch(host, "*.unrealengine.com")) {
-        return proxy;
-    }
-
-    // DuckDuckGo / Brave Search
-    if (shExpMatch(host, "duckduckgo.com") ||
-        shExpMatch(host, "*.duckduckgo.com") ||
-        shExpMatch(host, "search.brave.com") ||
-        shExpMatch(host, "*.search.brave.com")) {
-        return proxy;
-    }
-
-    // Proton 全系
-    if (shExpMatch(host, "proton.me") ||
-        shExpMatch(host, "*.proton.me") ||
-        shExpMatch(host, "protonmail.com") ||
-        shExpMatch(host, "*.protonmail.com") ||
-        shExpMatch(host, "protonvpn.com") ||
-        shExpMatch(host, "*.protonvpn.com")) {
-        return proxy;
-    }
-
-    // Stack Overflow / Stack Exchange 全系
-    if (shExpMatch(host, "stackoverflow.com") ||
-        shExpMatch(host, "*.stackoverflow.com") ||
-        shExpMatch(host, "stackexchange.com") ||
-        shExpMatch(host, "*.stackexchange.com") ||
-        shExpMatch(host, "superuser.com") ||
-        shExpMatch(host, "*.superuser.com") ||
-        shExpMatch(host, "serverfault.com") ||
-        shExpMatch(host, "*.serverfault.com")) {
-        return proxy;
-    }
-
-    // Quora / Medium
-    if (shExpMatch(host, "quora.com") ||
-        shExpMatch(host, "*.quora.com") ||
-        shExpMatch(host, "medium.com") ||
-        shExpMatch(host, "*.medium.com")) {
-        return proxy;
-    }
-
-    // Pixiv 全系
-    if (shExpMatch(host, "pixiv.net") ||
-        shExpMatch(host, "*.pixiv.net") ||
-        shExpMatch(host, "pximg.net") ||
-        shExpMatch(host, "*.pximg.net")) {
-        return proxy;
-    }
-
-    // ニコニコ 全系
-    if (shExpMatch(host, "nicovideo.jp") ||
-        shExpMatch(host, "*.nicovideo.jp") ||
-        shExpMatch(host, "nico.ms") ||
-        shExpMatch(host, "*.nico.ms")) {
-        return proxy;
-    }
-
-    // EA / Origin / Apex 全系
-    if (shExpMatch(host, "ea.com") ||
-        shExpMatch(host, "*.ea.com") ||
-        shExpMatch(host, "origin.com") ||
-        shExpMatch(host, "*.origin.com") ||
-        shExpMatch(host, "eaassets.com") ||
-        shExpMatch(host, "*.eaassets.com") ||
-        shExpMatch(host, "apexlegends.com") ||
-        shExpMatch(host, "*.apexlegends.com")) {
-        return proxy;
-    }
-
-    // Battle.net / Blizzard 全系
-    if (shExpMatch(host, "battle.net") ||
-        shExpMatch(host, "*.battle.net") ||
-        shExpMatch(host, "blizzard.com") ||
-        shExpMatch(host, "*.blizzard.com")) {
-        return proxy;
-    }
-
-    // Ubisoft 全系
-    if (shExpMatch(host, "ubisoft.com") ||
-        shExpMatch(host, "*.ubisoft.com") ||
-        shExpMatch(host, "ubi.com") ||
-        shExpMatch(host, "*.ubi.com")) {
-        return proxy;
-    }
-
-    // Riot Games / 英雄联盟 / Valorant 全系
-    if (shExpMatch(host, "riotgames.com") ||
-        shExpMatch(host, "*.riotgames.com") ||
-        shExpMatch(host, "leagueoflegends.com") ||
-        shExpMatch(host, "*.leagueoflegends.com") ||
-        shExpMatch(host, "valorant.com") ||
-        shExpMatch(host, "*.valorant.com")) {
-        return proxy;
-    }
-
-    // GOG / itch.io 全系
-    if (shExpMatch(host, "gog.com") ||
-        shExpMatch(host, "*.gog.com") ||
-        shExpMatch(host, "itch.io") ||
-        shExpMatch(host, "*.itch.io")) {
-        return proxy;
-    }
-
-    // Twitch 全系
-    if (shExpMatch(host, "twitch.tv") ||
-        shExpMatch(host, "*.twitch.tv") ||
-        shExpMatch(host, "ttvnw.net") ||
-        shExpMatch(host, "*.ttvnw.net") ||
-        shExpMatch(host, "twitchcdn.net") ||
-        shExpMatch(host, "*.twitchcdn.net")) {
-        return proxy;
-    }
-
-    // Crunchyroll 全系
-    if (shExpMatch(host, "crunchyroll.com") ||
-        shExpMatch(host, "*.crunchyroll.com")) {
-        return proxy;
-    }
-
-    // Amazon Prime Video 全系
-    if (shExpMatch(host, "primevideo.com") ||
-        shExpMatch(host, "*.primevideo.com")) {
-        return proxy;
-    }
-
-    // Apple TV+ 全系
-    if (shExpMatch(host, "tv.apple.com") ||
-        shExpMatch(host, "tvplus.apple.com")) {
-        return proxy;
-    }
-
-    // HBO Max / Hulu / Paramount+ 全系
-    if (shExpMatch(host, "max.com") ||
-        shExpMatch(host, "*.max.com") ||
-        shExpMatch(host, "hulu.com") ||
-        shExpMatch(host, "*.hulu.com") ||
-        shExpMatch(host, "paramountplus.com") ||
-        shExpMatch(host, "*.paramountplus.com")) {
-        return proxy;
-    }
-
-    // Slack 全系
-    if (shExpMatch(host, "slack.com") ||
-        shExpMatch(host, "*.slack.com") ||
-        shExpMatch(host, "slack-edge.com") ||
-        shExpMatch(host, "*.slack-edge.com") ||
-        shExpMatch(host, "slack-imgs.com") ||
-        shExpMatch(host, "*.slack-imgs.com")) {
-        return proxy;
-    }
-
-    // Zoom 全系
-    if (shExpMatch(host, "zoom.us") ||
-        shExpMatch(host, "*.zoom.us") ||
-        shExpMatch(host, "zoom.com") ||
-        shExpMatch(host, "*.zoom.com") ||
-        shExpMatch(host, "zoomgov.com") ||
-        shExpMatch(host, "*.zoomgov.com")) {
-        return proxy;
-    }
-
-    // Signal 全系
-    if (shExpMatch(host, "signal.org") ||
-        shExpMatch(host, "*.signal.org")) {
-        return proxy;
-    }
-
-    // Docker 全系
-    if (shExpMatch(host, "docker.com") ||
-        shExpMatch(host, "*.docker.com") ||
-        shExpMatch(host, "docker.io") ||
-        shExpMatch(host, "*.docker.io")) {
-        return proxy;
-    }
-
-    // GitLab 全系
-    if (shExpMatch(host, "gitlab.com") ||
-        shExpMatch(host, "*.gitlab.com") ||
-        shExpMatch(host, "gitlab.io") ||
-        shExpMatch(host, "*.gitlab.io")) {
-        return proxy;
-    }
-
-    // Vercel 全系
-    if (shExpMatch(host, "vercel.com") ||
-        shExpMatch(host, "*.vercel.com") ||
-        shExpMatch(host, "vercel.app") ||
-        shExpMatch(host, "*.vercel.app")) {
-        return proxy;
-    }
-
-    // Netlify 全系
-    if (shExpMatch(host, "netlify.com") ||
-        shExpMatch(host, "*.netlify.com") ||
-        shExpMatch(host, "netlify.app") ||
-        shExpMatch(host, "*.netlify.app")) {
-        return proxy;
-    }
-
-    // Perplexity / Poe / Midjourney 全系
-    if (shExpMatch(host, "perplexity.ai") ||
-        shExpMatch(host, "*.perplexity.ai") ||
-        shExpMatch(host, "poe.com") ||
-        shExpMatch(host, "*.poe.com") ||
-        shExpMatch(host, "midjourney.com") ||
-        shExpMatch(host, "*.midjourney.com")) {
-        return proxy;
-    }
-
-    // Figma / Canva 全系
-    if (shExpMatch(host, "figma.com") ||
-        shExpMatch(host, "*.figma.com") ||
-        shExpMatch(host, "canva.com") ||
-        shExpMatch(host, "*.canva.com")) {
-        return proxy;
-    }
-
-    // LinkedIn 全系
-    if (shExpMatch(host, "linkedin.com") ||
-        shExpMatch(host, "*.linkedin.com") ||
-        shExpMatch(host, "licdn.com") ||
-        shExpMatch(host, "*.licdn.com")) {
-        return proxy;
-    }
-
-    // Pinterest 全系
-    if (shExpMatch(host, "pinterest.com") ||
-        shExpMatch(host, "*.pinterest.com") ||
-        shExpMatch(host, "pinimg.com") ||
-        shExpMatch(host, "*.pinimg.com")) {
-        return proxy;
-    }
-
-    // Tumblr 全系
-    if (shExpMatch(host, "tumblr.com") ||
-        shExpMatch(host, "*.tumblr.com")) {
-        return proxy;
-    }
-
-    // Snapchat 全系
-    if (shExpMatch(host, "snapchat.com") ||
-        shExpMatch(host, "*.snapchat.com") ||
-        shExpMatch(host, "sc-cdn.net") ||
-        shExpMatch(host, "*.sc-cdn.net")) {
-        return proxy;
-    }
-
-    // Wikipedia / Wikimedia 全系
-    if (shExpMatch(host, "wikipedia.org") ||
-        shExpMatch(host, "*.wikipedia.org") ||
-        shExpMatch(host, "wikimedia.org") ||
-        shExpMatch(host, "*.wikimedia.org")) {
-        return proxy;
-    }
-
-    // PayPal 全系
-    if (shExpMatch(host, "paypal.com") ||
-        shExpMatch(host, "*.paypal.com") ||
-        shExpMatch(host, "paypalobjects.com") ||
-        shExpMatch(host, "*.paypalobjects.com")) {
-        return proxy;
-    }
-
-    // Patreon 全系
-    if (shExpMatch(host, "patreon.com") ||
-        shExpMatch(host, "*.patreon.com")) {
-        return proxy;
-    }
-
-    // 通用 CDN
-    if (shExpMatch(host, "cloudfront.net") ||
-        shExpMatch(host, "*.cloudfront.net") ||
-        shExpMatch(host, "fastly.net") ||
-        shExpMatch(host, "*.fastly.net") ||
-        shExpMatch(host, "fastlylb.net") ||
-        shExpMatch(host, "*.fastlylb.net") ||
-        shExpMatch(host, "jsdelivr.net") ||
-        shExpMatch(host, "*.jsdelivr.net") ||
-        shExpMatch(host, "unpkg.com") ||
-        shExpMatch(host, "*.unpkg.com") ||
-        shExpMatch(host, "statically.io") ||
-        shExpMatch(host, "*.statically.io") ||
-        shExpMatch(host, "cloudinary.com") ||
-        shExpMatch(host, "*.cloudinary.com") ||
-        shExpMatch(host, "imgur.com") ||
-        shExpMatch(host, "*.imgur.com")) {
-        return proxy;
-    }
-
-    // linux.do
-    if (shExpMatch(host, "linux.do") ||
-        shExpMatch(host, "*.linux.do")) {
-        return proxy;
-    }
-
-    return "DIRECT";
-}

server_config/pac/proxy_wg.pac

@@ -1,648 +0,0 @@
-function FindProxyForURL(url, host) {
-    // RRPP 规则代理列表 — 仅以下域名走加密隧道
-    var proxy = "PROXY 127.0.0.1:1080";
-    // Google 全系 + Gemini + CDN
-    if (shExpMatch(host, "google.com") ||
-        shExpMatch(host, "*.google.com") ||
-        shExpMatch(host, "google.com.hk") ||
-        shExpMatch(host, "*.google.com.hk") ||
-        shExpMatch(host, "google.co.jp") ||
-        shExpMatch(host, "*.google.co.jp") ||
-        shExpMatch(host, "googleapis.com") ||
-        shExpMatch(host, "*.googleapis.com") ||
-        shExpMatch(host, "googleusercontent.com") ||
-        shExpMatch(host, "*.googleusercontent.com") ||
-        shExpMatch(host, "googletagmanager.com") ||
-        shExpMatch(host, "*.googletagmanager.com") ||
-        shExpMatch(host, "google-analytics.com") ||
-        shExpMatch(host, "*.google-analytics.com") ||
-        shExpMatch(host, "gstatic.com") ||
-        shExpMatch(host, "*.gstatic.com") ||
-        shExpMatch(host, "ggpht.com") ||
-        shExpMatch(host, "*.ggpht.com") ||
-        shExpMatch(host, "googlehosted.com") ||
-        shExpMatch(host, "*.googlehosted.com") ||
-        shExpMatch(host, "1e100.net") ||
-        shExpMatch(host, "*.1e100.net") ||
-        shExpMatch(host, "youtube.com") ||
-        shExpMatch(host, "*.youtube.com") ||
-        shExpMatch(host, "ytimg.com") ||
-        shExpMatch(host, "*.ytimg.com") ||
-        shExpMatch(host, "googlevideo.com") ||
-        shExpMatch(host, "*.googlevideo.com") ||
-        shExpMatch(host, "gemini.google.com") ||
-        shExpMatch(host, "*.gemini.google.com") ||
-        shExpMatch(host, "generativeai.googleapis.com") ||
-        shExpMatch(host, "*.generativeai.googleapis.com") ||
-        shExpMatch(host, "aistudio.google.com") ||
-        shExpMatch(host, "*.aistudio.google.com") ||
-        shExpMatch(host, "deepmind.google.com") ||
-        shExpMatch(host, "*.deepmind.google.com") ||
-        shExpMatch(host, "deepmind.com") ||
-        shExpMatch(host, "*.deepmind.com")) {
-        return proxy;
-    }
-
-    // OpenAI / ChatGPT 全系 + CDN
-    if (shExpMatch(host, "openai.com") ||
-        shExpMatch(host, "*.openai.com") ||
-        shExpMatch(host, "chatgpt.com") ||
-        shExpMatch(host, "*.chatgpt.com") ||
-        shExpMatch(host, "oaistatic.com") ||
-        shExpMatch(host, "*.oaistatic.com") ||
-        shExpMatch(host, "oaiusercontent.com") ||
-        shExpMatch(host, "*.oaiusercontent.com") ||
-        shExpMatch(host, "openaiusercontent.com") ||
-        shExpMatch(host, "*.openaiusercontent.com")) {
-        return proxy;
-    }
-
-    // Grok / xAI 全系 + CDN
-    if (shExpMatch(host, "grok.com") ||
-        shExpMatch(host, "*.grok.com") ||
-        shExpMatch(host, "x.com") ||
-        shExpMatch(host, "*.x.com") ||
-        shExpMatch(host, "twitter.com") ||
-        shExpMatch(host, "*.twitter.com") ||
-        shExpMatch(host, "twimg.com") ||
-        shExpMatch(host, "*.twimg.com") ||
-        shExpMatch(host, "x.ai") ||
-        shExpMatch(host, "*.x.ai") ||
-        shExpMatch(host, "xai.com") ||
-        shExpMatch(host, "*.xai.com")) {
-        return proxy;
-    }
-
-    // Meta 全系 + CDN
-    if (shExpMatch(host, "meta.com") ||
-        shExpMatch(host, "*.meta.com") ||
-        shExpMatch(host, "meta.ai") ||
-        shExpMatch(host, "*.meta.ai") ||
-        shExpMatch(host, "facebook.com") ||
-        shExpMatch(host, "*.facebook.com") ||
-        shExpMatch(host, "fb.com") ||
-        shExpMatch(host, "*.fb.com") ||
-        shExpMatch(host, "fbcdn.net") ||
-        shExpMatch(host, "*.fbcdn.net") ||
-        shExpMatch(host, "fbsbx.com") ||
-        shExpMatch(host, "*.fbsbx.com") ||
-        shExpMatch(host, "instagram.com") ||
-        shExpMatch(host, "*.instagram.com") ||
-        shExpMatch(host, "cdninstagram.com") ||
-        shExpMatch(host, "*.cdninstagram.com") ||
-        shExpMatch(host, "threads.net") ||
-        shExpMatch(host, "*.threads.net") ||
-        shExpMatch(host, "whatsapp.com") ||
-        shExpMatch(host, "*.whatsapp.com") ||
-        shExpMatch(host, "whatsapp.net") ||
-        shExpMatch(host, "*.whatsapp.net") ||
-        shExpMatch(host, "messenger.com") ||
-        shExpMatch(host, "*.messenger.com")) {
-        return proxy;
-    }
-
-    // Claude / Anthropic 全系 + CDN
-    if (shExpMatch(host, "anthropic.com") ||
-        shExpMatch(host, "*.anthropic.com") ||
-        shExpMatch(host, "claude.ai") ||
-        shExpMatch(host, "*.claude.ai") ||
-        shExpMatch(host, "claudeusercontent.com") ||
-        shExpMatch(host, "*.claudeusercontent.com")) {
-        return proxy;
-    }
-
-    // Dify 全系
-    if (shExpMatch(host, "dify.ai") ||
-        shExpMatch(host, "*.dify.ai") ||
-        shExpMatch(host, "udify.app") ||
-        shExpMatch(host, "*.udify.app")) {
-        return proxy;
-    }
-
-    // Cloudflare 全系
-    if (shExpMatch(host, "cloudflare.com") ||
-        shExpMatch(host, "*.cloudflare.com") ||
-        shExpMatch(host, "cloudflare.net") ||
-        shExpMatch(host, "*.cloudflare.net") ||
-        shExpMatch(host, "cloudflareinsights.com") ||
-        shExpMatch(host, "*.cloudflareinsights.com") ||
-        shExpMatch(host, "cloudflarestatus.com") ||
-        shExpMatch(host, "*.cloudflarestatus.com") ||
-        shExpMatch(host, "cloudflarecdn.com") ||
-        shExpMatch(host, "*.cloudflarecdn.com") ||
-        shExpMatch(host, "cf-cdn.com") ||
-        shExpMatch(host, "*.cf-cdn.com") ||
-        shExpMatch(host, "workers.dev") ||
-        shExpMatch(host, "*.workers.dev") ||
-        shExpMatch(host, "pages.dev") ||
-        shExpMatch(host, "*.pages.dev")) {
-        return proxy;
-    }
-
-    // GitHub 全系
-    if (shExpMatch(host, "github.com") ||
-        shExpMatch(host, "*.github.com") ||
-        shExpMatch(host, "githubusercontent.com") ||
-        shExpMatch(host, "*.githubusercontent.com") ||
-        shExpMatch(host, "githubassets.com") ||
-        shExpMatch(host, "*.githubassets.com") ||
-        shExpMatch(host, "github.io") ||
-        shExpMatch(host, "*.github.io") ||
-        shExpMatch(host, "githubcopilot.com") ||
-        shExpMatch(host, "*.githubcopilot.com") ||
-        shExpMatch(host, "ghcr.io") ||
-        shExpMatch(host, "*.ghcr.io")) {
-        return proxy;
-    }
-
-    // Hugging Face 全系 + CDN
-    if (shExpMatch(host, "huggingface.co") ||
-        shExpMatch(host, "*.huggingface.co") ||
-        shExpMatch(host, "hf.co") ||
-        shExpMatch(host, "*.hf.co") ||
-        shExpMatch(host, "huggingfacecdn.com") ||
-        shExpMatch(host, "*.huggingfacecdn.com")) {
-        return proxy;
-    }
-
-    // Microsoft Copilot 全系 + Bing + Azure CDN
-    if (shExpMatch(host, "microsoft.com") ||
-        shExpMatch(host, "*.microsoft.com") ||
-        shExpMatch(host, "bing.com") ||
-        shExpMatch(host, "*.bing.com") ||
-        shExpMatch(host, "bingapis.com") ||
-        shExpMatch(host, "*.bingapis.com") ||
-        shExpMatch(host, "bing.net") ||
-        shExpMatch(host, "*.bing.net") ||
-        shExpMatch(host, "live.com") ||
-        shExpMatch(host, "*.live.com") ||
-        shExpMatch(host, "microsoftonline.com") ||
-        shExpMatch(host, "*.microsoftonline.com") ||
-        shExpMatch(host, "microsoft365.com") ||
-        shExpMatch(host, "*.microsoft365.com") ||
-        shExpMatch(host, "office.com") ||
-        shExpMatch(host, "*.office.com") ||
-        shExpMatch(host, "office.net") ||
-        shExpMatch(host, "*.office.net") ||
-        shExpMatch(host, "azure.com") ||
-        shExpMatch(host, "*.azure.com") ||
-        shExpMatch(host, "azureedge.net") ||
-        shExpMatch(host, "*.azureedge.net") ||
-        shExpMatch(host, "azurefd.net") ||
-        shExpMatch(host, "*.azurefd.net") ||
-        shExpMatch(host, "msn.com") ||
-        shExpMatch(host, "*.msn.com") ||
-        shExpMatch(host, "msecnd.net") ||
-        shExpMatch(host, "*.msecnd.net") ||
-        shExpMatch(host, "msftncsi.com") ||
-        shExpMatch(host, "*.msftncsi.com") ||
-        shExpMatch(host, "s-microsoft.com") ||
-        shExpMatch(host, "*.s-microsoft.com")) {
-        return proxy;
-    }
-
-    // Steam 全系 + CDN / 社区 / 下载
-    if (shExpMatch(host, "steampowered.com") ||
-        shExpMatch(host, "*.steampowered.com") ||
-        shExpMatch(host, "steamcommunity.com") ||
-        shExpMatch(host, "*.steamcommunity.com") ||
-        shExpMatch(host, "steam.com") ||
-        shExpMatch(host, "*.steam.com") ||
-        shExpMatch(host, "steamgames.com") ||
-        shExpMatch(host, "*.steamgames.com") ||
-        shExpMatch(host, "steamusercontent.com") ||
-        shExpMatch(host, "*.steamusercontent.com") ||
-        shExpMatch(host, "steamcontent.com") ||
-        shExpMatch(host, "*.steamcontent.com") ||
-        shExpMatch(host, "steamstatic.com") ||
-        shExpMatch(host, "*.steamstatic.com") ||
-        shExpMatch(host, "steam-chat.com") ||
-        shExpMatch(host, "*.steam-chat.com") ||
-        shExpMatch(host, "steam-api.com") ||
-        shExpMatch(host, "*.steam-api.com") ||
-        shExpMatch(host, "steamserver.net") ||
-        shExpMatch(host, "*.steamserver.net") ||
-        shExpMatch(host, "steamcdn-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamcdn-a.akamaihd.net") ||
-        shExpMatch(host, "steamcommunity-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamcommunity-a.akamaihd.net") ||
-        shExpMatch(host, "steamstore-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamstore-a.akamaihd.net") ||
-        shExpMatch(host, "steamuserimages-a.akamaihd.net") ||
-        shExpMatch(host, "*.steamuserimages-a.akamaihd.net") ||
-        shExpMatch(host, "akamaihd.net") ||
-        shExpMatch(host, "*.akamaihd.net") ||
-        shExpMatch(host, "valvesoftware.com") ||
-        shExpMatch(host, "*.valvesoftware.com")) {
-        return proxy;
-    }
-
-    // Discord 全系
-    if (shExpMatch(host, "discord.com") ||
-        shExpMatch(host, "*.discord.com") ||
-        shExpMatch(host, "discordapp.com") ||
-        shExpMatch(host, "*.discordapp.com") ||
-        shExpMatch(host, "discord.gg") ||
-        shExpMatch(host, "*.discord.gg") ||
-        shExpMatch(host, "discord.media") ||
-        shExpMatch(host, "*.discord.media") ||
-        shExpMatch(host, "discordapp.net") ||
-        shExpMatch(host, "*.discordapp.net")) {
-        return proxy;
-    }
-
-    // Telegram 全系
-    if (shExpMatch(host, "telegram.org") ||
-        shExpMatch(host, "*.telegram.org") ||
-        shExpMatch(host, "t.me") ||
-        shExpMatch(host, "*.t.me") ||
-        shExpMatch(host, "telega.one") ||
-        shExpMatch(host, "*.telega.one") ||
-        shExpMatch(host, "telegram-cdn.org") ||
-        shExpMatch(host, "*.telegram-cdn.org")) {
-        return proxy;
-    }
-
-    // Reddit 全系
-    if (shExpMatch(host, "reddit.com") ||
-        shExpMatch(host, "*.reddit.com") ||
-        shExpMatch(host, "redd.it") ||
-        shExpMatch(host, "*.redd.it") ||
-        shExpMatch(host, "redditmedia.com") ||
-        shExpMatch(host, "*.redditmedia.com") ||
-        shExpMatch(host, "redditstatic.com") ||
-        shExpMatch(host, "*.redditstatic.com")) {
-        return proxy;
-    }
-
-    // LINE 全系
-    if (shExpMatch(host, "line.me") ||
-        shExpMatch(host, "*.line.me") ||
-        shExpMatch(host, "line-apps.com") ||
-        shExpMatch(host, "*.line-apps.com") ||
-        shExpMatch(host, "line-scdn.net") ||
-        shExpMatch(host, "*.line-scdn.net")) {
-        return proxy;
-    }
-
-    // Spotify 全系
-    if (shExpMatch(host, "spotify.com") ||
-        shExpMatch(host, "*.spotify.com") ||
-        shExpMatch(host, "spotifycdn.com") ||
-        shExpMatch(host, "*.spotifycdn.com") ||
-        shExpMatch(host, "scdn.co") ||
-        shExpMatch(host, "*.scdn.co")) {
-        return proxy;
-    }
-
-    // Netflix 全系
-    if (shExpMatch(host, "netflix.com") ||
-        shExpMatch(host, "*.netflix.com") ||
-        shExpMatch(host, "nflxvideo.net") ||
-        shExpMatch(host, "*.nflxvideo.net") ||
-        shExpMatch(host, "nflximg.net") ||
-        shExpMatch(host, "*.nflximg.net") ||
-        shExpMatch(host, "nflxext.com") ||
-        shExpMatch(host, "*.nflxext.com")) {
-        return proxy;
-    }
-
-    // Disney+ 全系
-    if (shExpMatch(host, "disneyplus.com") ||
-        shExpMatch(host, "*.disneyplus.com") ||
-        shExpMatch(host, "disney-plus.net") ||
-        shExpMatch(host, "*.disney-plus.net") ||
-        shExpMatch(host, "dssott.com") ||
-        shExpMatch(host, "*.dssott.com")) {
-        return proxy;
-    }
-
-    // TikTok 全系
-    if (shExpMatch(host, "tiktok.com") ||
-        shExpMatch(host, "*.tiktok.com") ||
-        shExpMatch(host, "tiktokcdn.com") ||
-        shExpMatch(host, "*.tiktokcdn.com") ||
-        shExpMatch(host, "tiktokv.com") ||
-        shExpMatch(host, "*.tiktokv.com")) {
-        return proxy;
-    }
-
-    // PlayStation 全系
-    if (shExpMatch(host, "playstation.com") ||
-        shExpMatch(host, "*.playstation.com") ||
-        shExpMatch(host, "sonyentertainmentnetwork.com") ||
-        shExpMatch(host, "*.sonyentertainmentnetwork.com")) {
-        return proxy;
-    }
-
-    // Nintendo 全系
-    if (shExpMatch(host, "nintendo.com") ||
-        shExpMatch(host, "*.nintendo.com") ||
-        shExpMatch(host, "nintendowifi.net") ||
-        shExpMatch(host, "*.nintendowifi.net")) {
-        return proxy;
-    }
-
-    // Epic Games 全系
-    if (shExpMatch(host, "epicgames.com") ||
-        shExpMatch(host, "*.epicgames.com") ||
-        shExpMatch(host, "unrealengine.com") ||
-        shExpMatch(host, "*.unrealengine.com")) {
-        return proxy;
-    }
-
-    // DuckDuckGo / Brave Search
-    if (shExpMatch(host, "duckduckgo.com") ||
-        shExpMatch(host, "*.duckduckgo.com") ||
-        shExpMatch(host, "search.brave.com") ||
-        shExpMatch(host, "*.search.brave.com")) {
-        return proxy;
-    }
-
-    // Proton 全系
-    if (shExpMatch(host, "proton.me") ||
-        shExpMatch(host, "*.proton.me") ||
-        shExpMatch(host, "protonmail.com") ||
-        shExpMatch(host, "*.protonmail.com") ||
-        shExpMatch(host, "protonvpn.com") ||
-        shExpMatch(host, "*.protonvpn.com")) {
-        return proxy;
-    }
-
-    // Stack Overflow / Stack Exchange 全系
-    if (shExpMatch(host, "stackoverflow.com") ||
-        shExpMatch(host, "*.stackoverflow.com") ||
-        shExpMatch(host, "stackexchange.com") ||
-        shExpMatch(host, "*.stackexchange.com") ||
-        shExpMatch(host, "superuser.com") ||
-        shExpMatch(host, "*.superuser.com") ||
-        shExpMatch(host, "serverfault.com") ||
-        shExpMatch(host, "*.serverfault.com")) {
-        return proxy;
-    }
-
-    // Quora / Medium
-    if (shExpMatch(host, "quora.com") ||
-        shExpMatch(host, "*.quora.com") ||
-        shExpMatch(host, "medium.com") ||
-        shExpMatch(host, "*.medium.com")) {
-        return proxy;
-    }
-
-    // Pixiv 全系
-    if (shExpMatch(host, "pixiv.net") ||
-        shExpMatch(host, "*.pixiv.net") ||
-        shExpMatch(host, "pximg.net") ||
-        shExpMatch(host, "*.pximg.net")) {
-        return proxy;
-    }
-
-    // ニコニコ 全系
-    if (shExpMatch(host, "nicovideo.jp") ||
-        shExpMatch(host, "*.nicovideo.jp") ||
-        shExpMatch(host, "nico.ms") ||
-        shExpMatch(host, "*.nico.ms")) {
-        return proxy;
-    }
-
-    // EA / Origin / Apex 全系
-    if (shExpMatch(host, "ea.com") ||
-        shExpMatch(host, "*.ea.com") ||
-        shExpMatch(host, "origin.com") ||
-        shExpMatch(host, "*.origin.com") ||
-        shExpMatch(host, "eaassets.com") ||
-        shExpMatch(host, "*.eaassets.com") ||
-        shExpMatch(host, "apexlegends.com") ||
-        shExpMatch(host, "*.apexlegends.com")) {
-        return proxy;
-    }
-
-    // Battle.net / Blizzard 全系
-    if (shExpMatch(host, "battle.net") ||
-        shExpMatch(host, "*.battle.net") ||
-        shExpMatch(host, "blizzard.com") ||
-        shExpMatch(host, "*.blizzard.com")) {
-        return proxy;
-    }
-
-    // Ubisoft 全系
-    if (shExpMatch(host, "ubisoft.com") ||
-        shExpMatch(host, "*.ubisoft.com") ||
-        shExpMatch(host, "ubi.com") ||
-        shExpMatch(host, "*.ubi.com")) {
-        return proxy;
-    }
-
-    // Riot Games / 英雄联盟 / Valorant 全系
-    if (shExpMatch(host, "riotgames.com") ||
-        shExpMatch(host, "*.riotgames.com") ||
-        shExpMatch(host, "leagueoflegends.com") ||
-        shExpMatch(host, "*.leagueoflegends.com") ||
-        shExpMatch(host, "valorant.com") ||
-        shExpMatch(host, "*.valorant.com")) {
-        return proxy;
-    }
-
-    // GOG / itch.io 全系
-    if (shExpMatch(host, "gog.com") ||
-        shExpMatch(host, "*.gog.com") ||
-        shExpMatch(host, "itch.io") ||
-        shExpMatch(host, "*.itch.io")) {
-        return proxy;
-    }
-
-    // Twitch 全系
-    if (shExpMatch(host, "twitch.tv") ||
-        shExpMatch(host, "*.twitch.tv") ||
-        shExpMatch(host, "ttvnw.net") ||
-        shExpMatch(host, "*.ttvnw.net") ||
-        shExpMatch(host, "twitchcdn.net") ||
-        shExpMatch(host, "*.twitchcdn.net")) {
-        return proxy;
-    }
-
-    // Crunchyroll 全系
-    if (shExpMatch(host, "crunchyroll.com") ||
-        shExpMatch(host, "*.crunchyroll.com")) {
-        return proxy;
-    }
-
-    // Amazon Prime Video 全系
-    if (shExpMatch(host, "primevideo.com") ||
-        shExpMatch(host, "*.primevideo.com")) {
-        return proxy;
-    }
-
-    // Apple TV+ 全系
-    if (shExpMatch(host, "tv.apple.com") ||
-        shExpMatch(host, "tvplus.apple.com")) {
-        return proxy;
-    }
-
-    // HBO Max / Hulu / Paramount+ 全系
-    if (shExpMatch(host, "max.com") ||
-        shExpMatch(host, "*.max.com") ||
-        shExpMatch(host, "hulu.com") ||
-        shExpMatch(host, "*.hulu.com") ||
-        shExpMatch(host, "paramountplus.com") ||
-        shExpMatch(host, "*.paramountplus.com")) {
-        return proxy;
-    }
-
-    // Slack 全系
-    if (shExpMatch(host, "slack.com") ||
-        shExpMatch(host, "*.slack.com") ||
-        shExpMatch(host, "slack-edge.com") ||
-        shExpMatch(host, "*.slack-edge.com") ||
-        shExpMatch(host, "slack-imgs.com") ||
-        shExpMatch(host, "*.slack-imgs.com")) {
-        return proxy;
-    }
-
-    // Zoom 全系
-    if (shExpMatch(host, "zoom.us") ||
-        shExpMatch(host, "*.zoom.us") ||
-        shExpMatch(host, "zoom.com") ||
-        shExpMatch(host, "*.zoom.com") ||
-        shExpMatch(host, "zoomgov.com") ||
-        shExpMatch(host, "*.zoomgov.com")) {
-        return proxy;
-    }
-
-    // Signal 全系
-    if (shExpMatch(host, "signal.org") ||
-        shExpMatch(host, "*.signal.org")) {
-        return proxy;
-    }
-
-    // Docker 全系
-    if (shExpMatch(host, "docker.com") ||
-        shExpMatch(host, "*.docker.com") ||
-        shExpMatch(host, "docker.io") ||
-        shExpMatch(host, "*.docker.io")) {
-        return proxy;
-    }
-
-    // GitLab 全系
-    if (shExpMatch(host, "gitlab.com") ||
-        shExpMatch(host, "*.gitlab.com") ||
-        shExpMatch(host, "gitlab.io") ||
-        shExpMatch(host, "*.gitlab.io")) {
-        return proxy;
-    }
-
-    // Vercel 全系
-    if (shExpMatch(host, "vercel.com") ||
-        shExpMatch(host, "*.vercel.com") ||
-        shExpMatch(host, "vercel.app") ||
-        shExpMatch(host, "*.vercel.app")) {
-        return proxy;
-    }
-
-    // Netlify 全系
-    if (shExpMatch(host, "netlify.com") ||
-        shExpMatch(host, "*.netlify.com") ||
-        shExpMatch(host, "netlify.app") ||
-        shExpMatch(host, "*.netlify.app")) {
-        return proxy;
-    }
-
-    // Perplexity / Poe / Midjourney 全系
-    if (shExpMatch(host, "perplexity.ai") ||
-        shExpMatch(host, "*.perplexity.ai") ||
-        shExpMatch(host, "poe.com") ||
-        shExpMatch(host, "*.poe.com") ||
-        shExpMatch(host, "midjourney.com") ||
-        shExpMatch(host, "*.midjourney.com")) {
-        return proxy;
-    }
-
-    // Figma / Canva 全系
-    if (shExpMatch(host, "figma.com") ||
-        shExpMatch(host, "*.figma.com") ||
-        shExpMatch(host, "canva.com") ||
-        shExpMatch(host, "*.canva.com")) {
-        return proxy;
-    }
-
-    // LinkedIn 全系
-    if (shExpMatch(host, "linkedin.com") ||
-        shExpMatch(host, "*.linkedin.com") ||
-        shExpMatch(host, "licdn.com") ||
-        shExpMatch(host, "*.licdn.com")) {
-        return proxy;
-    }
-
-    // Pinterest 全系
-    if (shExpMatch(host, "pinterest.com") ||
-        shExpMatch(host, "*.pinterest.com") ||
-        shExpMatch(host, "pinimg.com") ||
-        shExpMatch(host, "*.pinimg.com")) {
-        return proxy;
-    }
-
-    // Tumblr 全系
-    if (shExpMatch(host, "tumblr.com") ||
-        shExpMatch(host, "*.tumblr.com")) {
-        return proxy;
-    }
-
-    // Snapchat 全系
-    if (shExpMatch(host, "snapchat.com") ||
-        shExpMatch(host, "*.snapchat.com") ||
-        shExpMatch(host, "sc-cdn.net") ||
-        shExpMatch(host, "*.sc-cdn.net")) {
-        return proxy;
-    }
-
-    // Wikipedia / Wikimedia 全系
-    if (shExpMatch(host, "wikipedia.org") ||
-        shExpMatch(host, "*.wikipedia.org") ||
-        shExpMatch(host, "wikimedia.org") ||
-        shExpMatch(host, "*.wikimedia.org")) {
-        return proxy;
-    }
-
-    // PayPal 全系
-    if (shExpMatch(host, "paypal.com") ||
-        shExpMatch(host, "*.paypal.com") ||
-        shExpMatch(host, "paypalobjects.com") ||
-        shExpMatch(host, "*.paypalobjects.com")) {
-        return proxy;
-    }
-
-    // Patreon 全系
-    if (shExpMatch(host, "patreon.com") ||
-        shExpMatch(host, "*.patreon.com")) {
-        return proxy;
-    }
-
-    // 通用 CDN
-    if (shExpMatch(host, "cloudfront.net") ||
-        shExpMatch(host, "*.cloudfront.net") ||
-        shExpMatch(host, "fastly.net") ||
-        shExpMatch(host, "*.fastly.net") ||
-        shExpMatch(host, "fastlylb.net") ||
-        shExpMatch(host, "*.fastlylb.net") ||
-        shExpMatch(host, "jsdelivr.net") ||
-        shExpMatch(host, "*.jsdelivr.net") ||
-        shExpMatch(host, "unpkg.com") ||
-        shExpMatch(host, "*.unpkg.com") ||
-        shExpMatch(host, "statically.io") ||
-        shExpMatch(host, "*.statically.io") ||
-        shExpMatch(host, "cloudinary.com") ||
-        shExpMatch(host, "*.cloudinary.com") ||
-        shExpMatch(host, "imgur.com") ||
-        shExpMatch(host, "*.imgur.com")) {
-        return proxy;
-    }
-
-    // linux.do
-    if (shExpMatch(host, "linux.do") ||
-        shExpMatch(host, "*.linux.do")) {
-        return proxy;
-    }
-
-    return "DIRECT";
-}
-

server_config/scripts/rrpp-stats

@@ -1,4 +0,0 @@
-#!/bin/bash
-# /usr/local/bin/rrpp-stats — returns traffic stats as JSON
-
-vnstat --json 2>/dev/null || echo '{"error":"vnstat not available"}'

server_config/scripts/sync-nginx-local-ips.sh

@@ -1,13 +1,15 @@
 #!/bin/bash
 # sync-nginx-local-ips.sh
 # 自动维护 nginx 本机公网 IP 直连规则（/etc/nginx/local_ip_rules.conf）：
-#   出站公网 IP（ipify）+ iam.aeolianchi.me 的 A 记录 = 本机当前公网 IP 集合。
+#   出站公网 IP（ipify）+ <节点域名> 的 A 记录 = 本机当前公网 IP 集合。
 #   公网 IP 变更时重写规则文件并 reload nginx（仅变更时 reload）。
 # 由 /etc/cron.d/sync-nginx-local-ips 每 5 分钟调用一次。
+#
+# 部署时把 DOMAIN 改为本节点域名。
 set -u
 PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
 RULES=/etc/nginx/local_ip_rules.conf
-DOMAIN=iam.aeolianchi.me
+DOMAIN=<节点域名>
 
 ips=""
 # 出站公网 IP

server_config/sing-box/config.json

@@ -1,6 +1,6 @@
 {
   "log": {
-    "level": "warn",
+    "level": "info",
     "timestamp": true
   },
   "inbounds": [
@@ -8,7 +8,7 @@
       "type": "vless",
       "tag": "vless-reality-in",
       "listen": "0.0.0.0",
-      "listen_port": 8443,
+      "listen_port": "<vless端口>",
       "users": [
         {
           "uuid": "<VLESS_UUID>",
@@ -17,11 +17,11 @@
       ],
       "tls": {
         "enabled": true,
-        "server_name": "www.apple.com",
+        "server_name": "<伪装SNI域名>",
         "reality": {
           "enabled": true,
           "handshake": {
-            "server": "www.apple.com",
+            "server": "<伪装SNI域名>",
             "server_port": 443
           },
           "private_key": "<REALITY_PRIVATE_KEY>",
@@ -30,6 +30,14 @@
           ]
         }
       }
+    },
+    {
+      "type": "shadowsocks",
+      "tag": "ss-in",
+      "listen": "0.0.0.0",
+      "listen_port": "<ss端口>",
+      "method": "2022-blake3-aes-256-gcm",
+      "password": "<SS_PSK_BASE64_32字节>"
     }
   ],
   "outbounds": [

server_config/systemd/mtg.service

@@ -0,0 +1,12 @@
+[Unit]
+Description=MTProto proxy (mtg) for Telegram
+After=network.target
+
+[Service]
+Type=simple
+ExecStart=/usr/local/bin/mtg simple-run --prefer-ip=prefer-ipv4 0.0.0.0:<mtproto端口> <MTPROTO_SECRET>
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target

server_config/systemd/nginx.service

@@ -0,0 +1,16 @@
+[Unit]
+Description=nginx (RRPP proxy, compiled with ngx_http_proxy_connect_module)
+After=network.target
+
+[Service]
+Type=forking
+PIDFile=/var/run/nginx.pid
+ExecStartPre=/usr/local/nginx/sbin/nginx -t
+ExecStart=/usr/local/nginx/sbin/nginx
+ExecReload=/usr/local/nginx/sbin/nginx -s reload
+ExecStop=/usr/local/nginx/sbin/nginx -s quit
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target

server_config/systemd/sing-box.service

@@ -0,0 +1,12 @@
+[Unit]
+Description=sing-box (VLESS+REALITY / Shadowsocks)
+After=network.target
+
+[Service]
+Type=simple
+ExecStart=/usr/local/bin/sing-box -D /var/lib/sing-box -C /etc/sing-box run
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target

server_config/traffic-monitor/config.json

@@ -1,41 +0,0 @@
-{
-  "vnstat_bin": "vnstat",
-  "interface": "eth0",
-  "data_dir": "/home/agent/data",
-  "alert_state_path": "/home/agent/alert_state.json",
-  "poll_interval_sec": 60,
-  "schedule_interval_sec": 30,
-  "http_host": "127.0.0.1",
-  "http_port": 18080,
-  "quota_bytes": 15000000000,
-  "alert_thresholds_pct": [
-    50,
-    80,
-    90,
-    95,
-    99,
-    100
-  ],
-  "effective_domain": "iam.aeolianchi.me",
-  "log_level": "INFO",
-  "smtp": {
-    "host": "mail.rrrosmontis.icu",
-    "port": 465,
-    "user": "<SMTP_USER>",
-    "password": "<SMTP_PASSWORD>",
-    "from_name": "Royirin Rosmontis Private Proxy JP 流量监控",
-    "from_addr": "<SMTP_FROM>",
-    "to_addr": "<SMTP_TO>",
-    "starttls": false,
-    "skip_cert_verify": false,
-    "timeout_sec": 30,
-    "ssl": true
-  },
-  "current_ip": "<PUBLIC_IP>",
-  "billing_cycle_start_day": 1,
-  "azure_usage_enabled": true,
-  "azure_usage_fetch_interval_sec": 21600,
-  "azure_usage_cache_path": "/home/agent/data/azure_usage.json",
-  "azure_meter_name_match": "Data Transfer Out",
-  "azure_free_limit_gb": 15
-}

server_config/traffic-monitor/traffic-monitor.service

@@ -1,15 +0,0 @@
-[Unit]
-Description=Royirin Rosmontis Private Proxy JP Traffic Monitor
-After=network-online.target nginx.service
-Wants=network-online.target
-
-[Service]
-Type=simple
-WorkingDirectory=/home/agent
-ExecStart=/usr/bin/python3 /home/agent/traffic_monitor.py --config /home/agent/config.json
-Restart=always
-RestartSec=5
-User=root
-
-[Install]
-WantedBy=multi-user.target

server_config/traffic-monitor/traffic_monitor.py

@@ -1,666 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-"""
-Royirin Rosmontis Private Proxy JP 流量监控（Azure 门户口径版）
-- 主计量口径：Azure 门户免费服务页 "Data Transfer Out (GB)"（Standard Data Transfer Out - Free，MGN 路由）
-  —— 通过 VM 托管身份调用 Consumption Usage Details API 拉取，按自然月聚合，缓存至 data/azure_usage.json
-- 参考口径：vnstat -i eth0（实时原始出站，仅作参考展示）
-- 数据源（参考）：vnstat -i <interface> --json m/d（eth0 为 bond 主接口，覆盖 nginx 443 与 sing-box 8443 全部流量）
-- 按 年/月 聚合成 JSON 存入 data/YYYY/YYYY-MM.json（弃用 SQLite）
-- HTTP 服务: GET /vpn/status/traffic 返回当月用量 JSON（Azure 口径为主）
-- 阈值告警: 达到配额 50/80/90/95/99/100% 时发邮件（每档每月只发一次，跳档补发；基于 Azure 口径）
-- 月报: 每月 1 日 00:05 起发送上个月报告（失败重试；基于 Azure 口径）
-"""
-import argparse
-import json
-import logging
-import os
-import socket
-import smtplib
-import ssl
-import subprocess
-import sys
-import threading
-import time
-import urllib.parse
-import urllib.request
-from datetime import date, datetime, timedelta, timezone
-from email.message import EmailMessage
-from email.utils import formataddr
-from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
-
-log = logging.getLogger('traffic_monitor')
-
-GB = 1_000_000_000  # Azure 数据转移按十进制 GB 计费
-
-
-def utcnow():
-    return datetime.now(timezone.utc)
-
-
-def load_json(path, default):
-    try:
-        with open(path, 'r', encoding='utf-8') as f:
-            return json.load(f)
-    except FileNotFoundError:
-        return default
-    except (json.JSONDecodeError, OSError) as e:
-        log.warning('read %s failed: %s; using default', path, e)
-        return default
-
-
-def save_json(path, obj):
-    tmp = path + '.tmp'
-    with open(tmp, 'w', encoding='utf-8') as f:
-        json.dump(obj, f, ensure_ascii=False, indent=2)
-    os.replace(tmp, path)
-
-
-def gb_str(n_bytes):
-    return f'{n_bytes / GB:.2f} GB'
-
-
-def month_struct(year, month):
-    return {
-        'year': int(year),
-        'month': int(month),
-        'rx_bytes': 0,
-        'tx_bytes': 0,
-        'days': [],
-        'updated_at': utcnow().isoformat(),
-    }
-
-
-class Monitor:
-    def __init__(self, config):
-        self.config = config
-        self.lock = threading.RLock()
-        self.months = {}
-        self.alert_state = None
-
-        os.makedirs(self.config['data_dir'], exist_ok=True)
-        self._load_all_months()
-        self._load_alert_state()
-        # Azure 计费用量（Usage Details API，VM 托管身份鉴权）
-        self.azure_usage = None
-        self._azure_last_fetch = 0.0
-        cache_path = self.config.get('azure_usage_cache_path')
-        if cache_path and os.path.exists(cache_path):
-            self.azure_usage = load_json(cache_path, None)
-            log.info('azure usage cache loaded: %s', self.azure_usage)
-
-    # ---------- 基础 ----------
-    @property
-    def _config_path(self):
-        return os.environ.get('TRAFFIC_CONFIG', '/home/agent/config.json')
-
-    def _load_all_months(self):
-        root = self.config['data_dir']
-        if not os.path.isdir(root):
-            return
-        for year_dir in os.listdir(root):
-            ypath = os.path.join(root, year_dir)
-            if not os.path.isdir(ypath):
-                continue
-            for fn in os.listdir(ypath):
-                if not fn.endswith('.json'):
-                    continue
-                try:
-                    with open(os.path.join(ypath, fn), 'r', encoding='utf-8') as f:
-                        m = json.load(f)
-                    mk = f"{m['year']:04d}-{m['month']:02d}"
-                    self.months[mk] = m
-                except Exception as e:
-                    log.warning('load %s/%s failed: %s', ypath, fn, e)
-
-    def _load_alert_state(self):
-        self.alert_state = load_json(
-            self.config['alert_state_path'],
-            {'alert_thresholds_sent': {}, 'monthly_reports_sent': []})
-        self.alert_state.setdefault('alert_thresholds_sent', {})
-        self.alert_state.setdefault('monthly_reports_sent', [])
-
-    def _save_alert_state(self):
-        save_json(self.config['alert_state_path'], self.alert_state)
-
-    def _month_path(self, year, month):
-        return os.path.join(self.config['data_dir'], str(year), f'{year:04d}-{month:02d}.json')
-
-    def _persist_month(self, mk):
-        m = self.months.get(mk)
-        if m is None:
-            return
-        path = self._month_path(m['year'], m['month'])
-        os.makedirs(os.path.dirname(path), exist_ok=True)
-        save_json(path, m)
-
-    # ---------- vnstat 读取 ----------
-    def _vnstat(self, what):
-        bin_path = self.config.get('vnstat_bin', 'vnstat')
-        iface = self.config.get('interface', 'eth0')
-        cmd = [bin_path, '-i', iface, '--json', what]
-        try:
-            out = subprocess.run(cmd, capture_output=True, timeout=20, check=True)
-            return json.loads(out.stdout.decode('utf-8'))
-        except FileNotFoundError:
-            log.error('vnstat binary not found: %s', bin_path)
-            return None
-        except subprocess.TimeoutExpired:
-            log.error('vnstat --json %s timed out', what)
-            return None
-        except (subprocess.CalledProcessError, json.JSONDecodeError, OSError) as e:
-            log.error('vnstat --json %s failed: %s', what, e)
-            return None
-
-    def poll_vnstat(self):
-        """拉取 vnstat 月/日数据，更新并落盘。返回更新的月数。"""
-        monthly = self._vnstat('m')
-        daily = self._vnstat('d')
-        touched = set()
-
-        with self.lock:
-            if monthly:
-                try:
-                    for itf in monthly.get('interfaces', []):
-                        for item in itf.get('traffic', {}).get('month', []):
-                            d = item.get('date', {})
-                            mk = f"{int(d['year']):04d}-{int(d['month']):02d}"
-                            m = self.months.get(mk)
-                            if m is None:
-                                m = month_struct(d['year'], d['month'])
-                                self.months[mk] = m
-                            m['rx_bytes'] = int(item.get('rx', 0))
-                            m['tx_bytes'] = int(item.get('tx', 0))
-                            m['updated_at'] = utcnow().isoformat()
-                            touched.add(mk)
-                except Exception as e:
-                    log.exception('parse vnstat monthly failed: %s', e)
-
-            if daily:
-                try:
-                    by_month = {}
-                    for itf in daily.get('interfaces', []):
-                        for item in itf.get('traffic', {}).get('day', []):
-                            d = item.get('date', {})
-                            mk = f"{int(d['year']):04d}-{int(d['month']):02d}"
-                            by_month.setdefault(mk, []).append({
-                                'day': int(d['day']),
-                                'rx_bytes': int(item.get('rx', 0)),
-                                'tx_bytes': int(item.get('tx', 0)),
-                            })
-                    for mk, days in by_month.items():
-                        m = self.months.get(mk)
-                        if m is None:
-                            y, mon = mk.split('-')
-                            m = month_struct(int(y), int(mon))
-                            self.months[mk] = m
-                        days.sort(key=lambda x: x['day'])
-                        m['days'] = days
-                        m['updated_at'] = utcnow().isoformat()
-                        touched.add(mk)
-                except Exception as e:
-                    log.exception('parse vnstat daily failed: %s', e)
-
-            for mk in touched:
-                self._persist_month(mk)
-        return len(touched)
-
-    # ---------- 用量（按 Azure 计费周期） ----------
-    def current_month_key(self):
-        return utcnow().strftime('%Y-%m')
-
-    def cycle_start(self, now):
-        """返回当前计费周期的起始 datetime（每月 billing_cycle_start_day 00:00 UTC）。"""
-        day = int(self.config.get('billing_cycle_start_day', 1))
-        day = max(1, min(28, day))
-        if now.day >= day:
-            return now.replace(day=day, hour=0, minute=0, second=0, microsecond=0)
-        prev = (now.replace(day=1, hour=0, minute=0, second=0, microsecond=0)
-                - timedelta(days=1))
-        return prev.replace(day=day, hour=0, minute=0, second=0, microsecond=0)
-
-    def cycle_end(self, start):
-        """周期结束 = 下个月同一天的 00:00（不含）。"""
-        nxt = (start.replace(day=1) + timedelta(days=32)).replace(day=1)
-        return nxt.replace(day=start.day, hour=0, minute=0, second=0, microsecond=0)
-
-    def _day_totals(self):
-        """返回 [{date, rx, tx}]，date 为 UTC date。"""
-        out = []
-        for m in self.months.values():
-            for d in m.get('days', []):
-                out.append((date(m['year'], m['month'], d['day']),
-                            int(d.get('rx_bytes', 0)), int(d.get('tx_bytes', 0))))
-        return out
-
-    def cycle_usage(self):
-        """返回当前计费周期 {rx_bytes, tx_bytes, start, end}。"""
-        now = utcnow()
-        start = self.cycle_start(now)
-        end = self.cycle_end(start)
-        rx = tx = 0
-        with self.lock:
-            for dt, r, t in self._day_totals():
-                if start.date() <= dt <= now.date():
-                    rx += r
-                    tx += t
-        return {'rx_bytes': rx, 'tx_bytes': tx, 'start': start, 'end': end}
-
-    def tx_bytes_for(self, mk):
-        with self.lock:
-            m = self.months.get(mk, {})
-            return int(m.get('tx_bytes', 0))
-
-    def current_status(self):
-        """当前状态：以 Azure 门户口径（Standard Data Transfer Out - Free）为主计量。"""
-        now = utcnow()
-        quota_gb = float(self.config.get('azure_free_limit_gb', 15))
-        cu = self.cycle_usage()
-        az = self.azure_usage or {}
-        used_gb = az.get('portal_data_transfer_out_gb')
-        used_pct = round(used_gb / quota_gb * 100, 4) if (used_gb is not None and quota_gb) else None
-        with self.lock:
-            cal_mk = now.strftime('%Y-%m')
-            m = self.months.get(cal_mk)
-            cal_tx = int(m.get('tx_bytes', 0)) if m else 0
-            cal_rx = int(m.get('rx_bytes', 0)) if m else 0
-            vnstat_updated = (m or {}).get('updated_at')
-        return {
-            'year': now.year,
-            'month': now.month,
-            'meter': 'azure_portal',
-            'used_gb': used_gb,
-            'quota_gb': quota_gb,
-            'quota_used_pct': used_pct,
-            'billing_cycle_start_day': int(self.config.get('billing_cycle_start_day', 1)),
-            'cycle_start': cu['start'].strftime('%Y-%m-%d'),
-            'cycle_end': cu['end'].strftime('%Y-%m-%d'),
-            'azure_portal_data_transfer_out_gb': az.get('portal_data_transfer_out_gb'),
-            'azure_bandwidth_out_gb': az.get('bandwidth_out_gb'),
-            'azure_usage_cost_cny': az.get('cost_cny'),
-            'azure_usage_month': az.get('month'),
-            'azure_usage_updated_at': az.get('updated_at'),
-            'vnstat_calendar_month_tx_bytes': cal_tx,
-            'vnstat_calendar_month_rx_bytes': cal_rx,
-            'vnstat_updated_at': vnstat_updated,
-            'updated_at': now.isoformat(),
-        }
-
-    def _vnstat_calendar_tx_gb(self):
-        """参考值：vnstat 自然月出站（GB，十进制）。"""
-        now = utcnow()
-        with self.lock:
-            m = self.months.get(now.strftime('%Y-%m'), {})
-            return int(m.get('tx_bytes', 0)) / GB
-
-    # ---------- Azure 计费用量（Usage Details API） ----------
-    def _azure_token(self):
-        """通过 VM 托管身份从 IMDS 获取 management.azure.com 的访问令牌。"""
-        url = ('http://169.254.169.254/metadata/identity/oauth2/token'
-               '?api-version=2018-02-01&resource=https%3A%2F%2Fmanagement.azure.com%2F')
-        req = urllib.request.Request(url, headers={'Metadata': 'true'})
-        with urllib.request.urlopen(req, timeout=15) as resp:
-            return json.loads(resp.read().decode('utf-8'))['access_token']
-
-    def _azure_subscription_id(self):
-        url = 'http://169.254.169.254/metadata/instance/compute?api-version=2021-02-01'
-        req = urllib.request.Request(url, headers={'Metadata': 'true'})
-        with urllib.request.urlopen(req, timeout=15) as resp:
-            return json.loads(resp.read().decode('utf-8')).get('subscriptionId')
-
-    def fetch_azure_usage(self, month_offset=0):
-        """拉取指定月份 usageDetails，聚合与门户免费服务页同源的 Data Transfer Out 用量。
-        month_offset: 0=当月（写入缓存），-1=上月（用于月报，不覆盖缓存）。"""
-        sub = self._azure_subscription_id()
-        if not sub:
-            raise RuntimeError('cannot read subscriptionId from IMDS')
-        token = self._azure_token()
-        now = utcnow()
-        month_start = (now.replace(day=1, hour=0, minute=0, second=0, microsecond=0)
-                       + timedelta(days=month_offset * 32)).replace(day=1)
-        nxt = (month_start.replace(day=1) + timedelta(days=32)).replace(day=1)
-        flt = ("properties/usageStart ge '{0}T00:00:00Z' and "
-               "properties/usageStart lt '{1}T00:00:00Z'").format(
-                   month_start.strftime('%Y-%m-%d'), nxt.strftime('%Y-%m-%d'))
-        base = ('https://management.azure.com/subscriptions/{sub}/providers/'
-                'Microsoft.Consumption/usageDetails?api-version=2023-11-01'
-                '&metric=actualcost&$filter={f}')
-        url = base.format(sub=sub, f=urllib.parse.quote(flt, safe=''))
-        portal_gb = 0.0
-        bw_out_gb = 0.0
-        cost_cny = 0.0
-        rows = 0
-        pages = 0
-        while url:
-            req = urllib.request.Request(url, headers={'Authorization': 'Bearer ' + token})
-            with urllib.request.urlopen(req, timeout=30) as resp:
-                d = json.loads(resp.read().decode('utf-8'))
-            for item in d.get('value', []):
-                p = item.get('properties', {}) or {}
-                qty = float(p.get('quantity') or 0)
-                rows += 1
-                name = str(p.get('meterName') or '')
-                subcat = str(p.get('meterSubCategory') or '')
-                if str(p.get('meterCategory') or '') == 'Bandwidth' and 'Data Transfer Out' in name:
-                    bw_out_gb += qty
-                    # 门户免费服务页 "Data Transfer Out (GB)" = MGN 路由的出站免费计费表
-                    if 'MGN' in subcat or '- Free' in name:
-                        portal_gb += qty
-                    cost_cny += float(p.get('costInBillingCurrency') or 0)
-            pages += 1
-            url = d.get('nextLink')
-            if pages > 30:
-                log.warning('azure usage pagination limit reached')
-                break
-        result = {
-            'month': month_start.strftime('%Y-%m'),
-            'portal_data_transfer_out_gb': round(portal_gb, 4),
-            'bandwidth_out_gb': round(bw_out_gb, 4),
-            'cost_cny': round(cost_cny, 4),
-            'rows': rows,
-            'updated_at': utcnow().isoformat(),
-        }
-        self.azure_usage = result
-        if month_offset == 0:
-            path = self.config.get('azure_usage_cache_path')
-            if path:
-                save_json(path, result)
-        log.info('azure usage fetched (offset=%d): %s', month_offset, result)
-        return result
-
-    def check_azure_usage(self):
-        if not self.config.get('azure_usage_enabled', False):
-            return
-        interval = float(self.config.get('azure_usage_fetch_interval_sec', 21600))
-        now = time.time()
-        if now - self._azure_last_fetch < interval:
-            return
-        self._azure_last_fetch = now
-        try:
-            self.fetch_azure_usage()
-        except Exception as e:
-            log.warning('azure usage fetch failed: %s', e)
-
-    # ---------- IP / 邮件 ----------
-    def get_current_ip(self):
-        configured = self.config.get('current_ip')
-        if configured:
-            return str(configured)
-        try:
-            s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
-            try:
-                s.connect(('8.8.8.8', 80))
-                return s.getsockname()[0]
-            finally:
-                s.close()
-        except OSError:
-            return '0.0.0.0'
-
-    def send_email(self, subject, body):
-        cfg = self.config['smtp']
-        msg = EmailMessage()
-        msg['Subject'] = subject
-        msg['From'] = formataddr((cfg.get('from_name', ''), cfg.get('from_addr')))
-        msg['To'] = cfg.get('to_addr')
-        msg.set_content(body)
-
-        context = None
-        if cfg.get('skip_cert_verify'):
-            context = ssl._create_unverified_context()
-        host = cfg['host']
-        port = int(cfg.get('port', 465))
-        timeout = int(cfg.get('timeout_sec', 30))
-        if cfg.get('ssl', False):
-            smtp_cls = smtplib.SMTP_SSL
-            kwargs = {'context': context}
-        else:
-            smtp_cls = smtplib.SMTP
-            kwargs = {}
-        with smtp_cls(host, port, timeout=timeout, **kwargs) as smtp:
-            if not cfg.get('ssl', False):
-                smtp.ehlo()
-                if cfg.get('starttls', True):
-                    smtp.starttls(context=context)
-                    smtp.ehlo()
-            smtp.login(cfg['user'], cfg['password'])
-            smtp.send_message(msg)
-        log.info('email sent: %s', subject)
-
-    def build_alert_mail(self, mk, used_gb, quota_gb, pct, az):
-        ip = self.get_current_ip()
-        month = (az or {}).get('month', mk.replace('azure:', ''))
-        az_updated = (az or {}).get('updated_at', '')
-        vnstat_tx = self._vnstat_calendar_tx_gb()
-        body = (
-            'Royirin Rosmontis Private Proxy JP 流量监控\n'
-            '==========\n'
-            '计量口径：Azure 门户（Standard Data Transfer Out - Free）\n'
-            f'统计月份：{month}（自然月）\n'
-            f'当前用量：{used_gb:.2f} GB / {quota_gb:.0f} GB（{pct:.2f}%）\n'
-            f'AZURE 数据更新：{az_updated}\n'
-            f'参考(vnstat 自然月出站)：{vnstat_tx:.2f} GB\n'
-            f'当前 IP：{ip}\n'
-            f'生效域名：{self.config.get("effective_domain", "")}\n'
-            '==========\n'
-            f'已达阈值 {pct:.2f}% (Azure 门户口径)，请注意使用情况'
-        )
-        subject = f'[流量告警] {month} 已用 {pct:.2f}% ({used_gb:.2f}/{quota_gb:.0f} GB)'
-        return subject, body
-
-    def build_monthly_mail(self, month, used_gb, quota_gb, pct, az):
-        ip = self.get_current_ip()
-        over = '是' if used_gb > quota_gb else '否'
-        az_updated = (az or {}).get('updated_at', '')
-        vnstat_tx = self._vnstat_calendar_tx_gb()
-        body = (
-            'Royirin Rosmontis Private Proxy JP 流量监控\n'
-            '月度报告\n'
-            '==========\n'
-            '计量口径：Azure 门户（Standard Data Transfer Out - Free）\n'
-            f'报告月份：{month}（自然月）\n'
-            f'当月用量：{used_gb:.2f} GB / {quota_gb:.0f} GB（{pct:.2f}%）\n'
-            f'AZURE 数据更新：{az_updated}\n'
-            f'参考(vnstat 自然月出站)：{vnstat_tx:.2f} GB\n'
-            f'当前 IP：{ip}\n'
-            f'生效域名：{self.config.get("effective_domain", "")}\n'
-            '==========\n'
-            f'上月使用 {over} 超限, 使用量为限额的 {pct:.2f}%'
-        )
-        subject = f'[流量月报] {month} 用量 {used_gb:.2f} GB ({pct:.2f}%)'
-        return subject, body
-
-    # ---------- 定时任务 ----------
-    def check_alerts(self):
-        """阈值告警：以 Azure 门户口径为准（无 Azure 数据时跳过，避免误报）。"""
-        quota_gb = float(self.config.get('azure_free_limit_gb', 15))
-        if quota_gb <= 0:
-            return
-        az = self.azure_usage or {}
-        used_gb = az.get('portal_data_transfer_out_gb')
-        if used_gb is None:
-            log.warning('alerts skipped: no azure usage data yet')
-            return
-        mk = 'azure:' + str(az.get('month') or utcnow().strftime('%Y-%m'))
-        pct = used_gb / quota_gb * 100
-        sent = self.alert_state['alert_thresholds_sent'].setdefault(mk, [])
-        if not sent:
-            # 新周期/口径切换：已越过的档位直接标记，避免切换瞬间补发一堆邮件
-            for t in sorted(self.config.get('alert_thresholds_pct', [])):
-                if pct >= t:
-                    sent.append(t)
-            self._save_alert_state()
-            log.info('seeded thresholds for %s at %.2f%%: %s', mk, pct, sent)
-        for t in sorted(self.config.get('alert_thresholds_pct', [])):
-            if pct >= t and t not in sent:
-                subject, body = self.build_alert_mail(mk, used_gb, quota_gb, pct, az)
-                try:
-                    self.send_email(subject, body)
-                    sent.append(t)
-                    self._save_alert_state()
-                    log.info('threshold %s%% alert sent for %s (used %.2f%%)', t, mk, pct)
-                except Exception as e:
-                    log.error('send threshold %s%% alert failed for %s: %s; will retry',
-                              t, mk, e)
-                    break
-
-    def check_monthly_report(self):
-        """月度报告：以 Azure 门户口径为准（上月数据，当月 1 号后发送）。"""
-        now = utcnow()
-        cur_start = self.cycle_start(now)
-        prev_start = (cur_start.replace(day=1) - timedelta(days=1)).replace(day=1)
-        prev_mk = 'azure:' + prev_start.strftime('%Y-%m')
-        if prev_mk in self.alert_state['monthly_reports_sent']:
-            return
-        # 口径切换迁移：旧键(cycle:)已发过则直接标记，避免重复发信
-        old_mk = 'cycle:' + prev_start.strftime('%Y-%m')
-        if old_mk in self.alert_state['monthly_reports_sent']:
-            self.alert_state['monthly_reports_sent'].append(prev_mk)
-            self._save_alert_state()
-            log.info('monthly report key migrated: %s already sent as %s', prev_mk, old_mk)
-            return
-        try:
-            az_prev = self.fetch_azure_usage(month_offset=-1)
-        except Exception as e:
-            log.error('fetch prev month azure usage failed: %s; will retry', e)
-            return
-        quota_gb = float(self.config.get('azure_free_limit_gb', 15))
-        used_gb = az_prev.get('portal_data_transfer_out_gb') or 0.0
-        pct = used_gb / quota_gb * 100 if quota_gb else 0.0
-        subject, body = self.build_monthly_mail(
-            az_prev.get('month', prev_start.strftime('%Y-%m')), used_gb, quota_gb, pct, az_prev)
-        try:
-            self.send_email(subject, body)
-            self.alert_state['monthly_reports_sent'].append(prev_mk)
-            self._save_alert_state()
-            log.info('monthly report sent for %s (used %.2f GB, %.2f%%)',
-                     prev_mk, used_gb, pct)
-        except Exception as e:
-            log.error('send monthly report for %s failed: %s; will retry', prev_mk, e)
-
-
-def build_http_handler(monitor):
-    class Handler(BaseHTTPRequestHandler):
-        def do_GET(self):
-            if self.path == '/vpn/status/traffic':
-                try:
-                    payload = json.dumps(monitor.current_status(), ensure_ascii=False)
-                    body = payload.encode('utf-8')
-                    self.send_response(200)
-                    self.send_header('Content-Type', 'application/json; charset=utf-8')
-                    self.send_header('Content-Length', str(len(body)))
-                    self.end_headers()
-                    self.wfile.write(body)
-                except Exception as e:
-                    log.exception('status endpoint error')
-                    body = b'{"error":"internal"}'
-                    self.send_response(500)
-                    self.send_header('Content-Type', 'application/json')
-                    self.send_header('Content-Length', str(len(body)))
-                    self.end_headers()
-                    self.wfile.write(body)
-            elif self.path in ('/', '/healthz'):
-                body = b'ok\n'
-                self.send_response(200)
-                self.send_header('Content-Type', 'text/plain')
-                self.send_header('Content-Length', str(len(body)))
-                self.end_headers()
-                self.wfile.write(body)
-            else:
-                body = b'{"error":"not found"}\n'
-                self.send_response(404)
-                self.send_header('Content-Type', 'application/json')
-                self.send_header('Content-Length', str(len(body)))
-                self.end_headers()
-                self.wfile.write(body)
-
-        def log_message(self, fmt, *args):
-            log.debug('http: ' + fmt, *args)
-
-    return Handler
-
-
-def run_service(config):
-    monitor = Monitor(config)
-
-    def poll_loop():
-        while True:
-            try:
-                n = monitor.poll_vnstat()
-                log.debug('vnstat poll updated %d month(s)', n)
-            except Exception:
-                log.exception('poll loop error')
-            time.sleep(int(config.get('poll_interval_sec', 60)))
-
-    def schedule_loop():
-        while True:
-            try:
-                monitor.check_azure_usage()  # 先刷新 Azure 口径数据，再评估告警
-                monitor.check_alerts()
-                monitor.check_monthly_report()
-            except Exception:
-                log.exception('schedule loop error')
-            time.sleep(int(config.get('schedule_interval_sec', 30)))
-
-    threading.Thread(target=poll_loop, daemon=True, name='vnstat-poll').start()
-    threading.Thread(target=schedule_loop, daemon=True, name='scheduler').start()
-
-    host = config.get('http_host', '127.0.0.1')
-    port = int(config.get('http_port', 18080))
-    server = ThreadingHTTPServer((host, port), build_http_handler(monitor))
-    log.info('HTTP server listening on http://%s:%d/vpn/status/traffic', host, port)
-    server.serve_forever()
-
-
-def main():
-    ap = argparse.ArgumentParser(description='vnstat traffic monitor')
-    ap.add_argument('--config', default=os.environ.get('TRAFFIC_CONFIG', '/home/agent/config.json'))
-    ap.add_argument('--once', action='store_true', help='拉取一次 vnstat 后退出')
-    ap.add_argument('--status', action='store_true', help='打印当前状态 JSON 后退出')
-    ap.add_argument('--test-email', action='store_true', help='发送一封测试邮件后退出')
-    args = ap.parse_args()
-
-    os.environ['TRAFFIC_CONFIG'] = args.config
-    with open(args.config, 'r', encoding='utf-8') as f:
-        config = json.load(f)
-
-    logging.basicConfig(
-        level=getattr(logging, config.get('log_level', 'INFO').upper(), logging.INFO),
-        format='%(asctime)s %(levelname)s %(name)s: %(message)s')
-
-    if args.once:
-        m = Monitor(config)
-        n = m.poll_vnstat()
-        print(f'updated {n} month(s)')
-        return 0
-    if args.status:
-        m = Monitor(config)
-        m.poll_vnstat()
-        print(json.dumps(m.current_status(), ensure_ascii=False, indent=2))
-        return 0
-    if args.test_email:
-        m = Monitor(config)
-        m.poll_vnstat()
-        quota_gb = float(config.get('azure_free_limit_gb', 15))
-        az = m.azure_usage or {}
-        used_gb = az.get('portal_data_transfer_out_gb')
-        if used_gb is None:
-            try:
-                m.fetch_azure_usage()
-                az = m.azure_usage or {}
-                used_gb = az.get('portal_data_transfer_out_gb')
-            except Exception as e:
-                print('azure usage fetch failed:', e)
-        mk = 'azure:' + str(az.get('month') or utcnow().strftime('%Y-%m'))
-        pct = used_gb / quota_gb * 100 if (used_gb is not None and quota_gb) else 0.0
-        subject, body = m.build_alert_mail(mk, used_gb or 0.0, quota_gb, pct, az)
-        print('sending test email:')
-        print('Subject:', subject)
-        print(body)
-        m.send_email(subject, body)
-        print('sent')
-        return 0
-    run_service(config)
-
-
-if __name__ == '__main__':
-    sys.exit(main())