FreeGuard VPN

看起来更像 GUI 统计/订阅层面的累积，不一定是 TUN 或代理链路本身。我们 FreeGuard PC 这边处理同类长期运行问题时做了几个收口：主界面只保留当前上下行速率，不长期堆流量图历史；流量 WS/轮询在代理关闭、页面卸载和模式切换时主动 close/clear timer；后台状态事件做 coalescing/有界队列，只保留最新 snapshot 和必要状态边界，避免 WebView 事件队列积压。排查 Clash Verge 可以重点看 traffic graph、memory、connection summary 的订阅生命周期，是否重复订阅或缓存没有释放。

@freeguardteam issue 看描述只要长期运行都有泄露的，issue 已经有人发过，看样子是有几个月的老问题了。 AI 查了一下意思是代理本身没问题，是 GUI 轮询流量累积不释放的 bug。

clash verge 开启TUN 后长时间运行会内存泄露 被我碰上了 github.com/clash-verge-re…

@huihuapp THX.

太专业了

@N0Ke921 Before changing settings, I’d collect route evidence: timestamp, region, ISP, ping/packet loss to ME vs EU, and one alternate network or mobile hotspot test. If the alternate network is clean, it points to ISP/peering; if both are bad, it supports a server-side routing issue.

There have been numerous problems with Fortnite Middle East server since March. The problem concerns extremely high ping and unavailability of the game. Moreover, this issue is not an exception, and there are many people in this region who have been facing it, whereas Fortnite Status assures that everything works perfectly fine. However, the situation is further aggravated by the customer support service, as it does not offer any relevant advice. On the contrary, customer support gives different information, ranging from those who ignore the problem to the representatives who offer to switch to European servers despite being unaware of the problems related to Middle Eastern servers. There were even instances when the chat was stopped without addressing the question. Nevertheless, switching to the European server would not help players with participating in fights and tournaments. Now, since there is stability in the network in Dammam, it becomes important to assess whether the routing process is correct or needs to be modified so that the Middle Eastern traffic can pass through the servers at Dammam. @FortniteStatus @FNCompetitive @FortniteME the ME region deserves to be taken seriously.

@huihuapp 导入别的订阅链接不等于更安全，关键是远程配置来源是否可信、订阅 URL 是否含 token、出口 IP/DNS/WebRTC/IPv6 是否一致。不要公开贴订阅链接；先用临时配置测试，再看客户端规则命中和泄露检测结果。

用类似小火箭、clashx这种，导入别的订阅链接，能避免吗？

I’d split this into layers: visible IP/ASN, DNS resolver, WebRTC/STUN, then browser timezone/language. A VPN can route traffic correctly while DNS or browser signals still tell a different story. The cleanest test is to check all four in the same browser session before changing routes.

@ManmohitSandhu Well, in this case WiFI is not in charge of it. DNS leak is real and the spy pixel is a different story. VPN is not allowed in each country, but if you can use it, it's better to use.

Да, skip rule для IP MTProxy — правильная идея. Я бы ещё проверил два слоя: DNS не должен уходить через VPN, если Telegram идёт мимо него, и режим Rule/Config в Shadowrocket не должен ломать маршрут для остальных приложений. Если начнутся обрывы, сначала смотрел бы конфликт VPN + app proxy, а не менял все узлы.

На айфоне в телеге свой mtproxy. Т.к. VPN + mtpoxy - глючит, то в VPN-клиенте поставил mtproxy айпишник в Skip rule. В VPN shadowrocket настроены белые/черные списки, только нужный трафик через VPN. А если нужно экономить батарею, то VPN вырубаю, но телега будет работать. Кайф.

World Cup match-day route checklist: 1. Know which device is actually routed 2. Keep DNS on the same path as your VPN/proxy 3. Test WebRTC/IP leaks before kickoff 4. Avoid switching routes mid-match 5. Pick stability over peak speed For streaming, gaming, AI tools, and TV setups, the best route is the one that stays clean for the whole session.

@pilotokskr You can also try our anonymous Clash subscription. It supports three protocols: HY2, AnyTLS, and Trojan. You may start with the $3.99 weekly subscription to test the stability first, and then switch to a monthly or yearly plan if it works well for you. freeguardvpn.com/clash/

梯子死了 谁有抗打的梯子 推荐推荐

@JFGAi 这个点很实用：本地回环地址最好单独排除。可以把 127.0.0.1、localhost、::1 明确放进 skip-proxy/直连，远端 API 再走代理。否则本地 bridge 被虚拟网卡接管后，看起来像 API 报错，其实是本机回环路径被绕坏了。

💡 想在 Codex 中使用 DeepSeek API 却遭遇报错？分享一个避坑指南！ 如果用 ccswitch 本地路由桥接时，遇到 stream disconnected before completion: error sending request for url (http://127.0.0.1:15721/v1/responses) 报错，这是因为本地 127.0.0.1 流量被 Shadowrocket（小火箭）等代理软件的虚拟网卡拦截了。 🛠 终极解决办法： 1️⃣ 小火箭全局路由切换为【配置 (Config)】模式 2️⃣ 在配置文件的 skip-proxy（跳过代理）列表里显式添加 127.0.0.1 3️⃣ 重启代理和 Codex，本地接口立马畅通！🚀

I’d separate two checks: leak tests show whether WebRTC/DNS exposes another IP, while VPN/proxy detection often comes from IP reputation databases. Even with no WebRTC leak, a site may still flag the exit IP as hosting/VPN. Test visible IP, DNS, WebRTC, timezone/language, and avoid switching routes repeatedly while testing.

@thedeepakk12 Hmm , no i have checked there's no webRtc leak , I wonder how do they determine ip is real one or is it vpn ip . Cause once they determine my real ip there's no going back .

Account fingerprint is crazy on ometv even if u use vpn it's gonna detect your country based on your activity

FreeGuard VPN for every device: iOS · Android · Windows · macOS · Chrome · TV · Router · CLI · Clash-compatible workflows One setup path should not limit how you work, watch, play, or build.

@NewWind2019 确实，到了 Docker/npm/k8s/Codex 这种混合场景，System Proxy 很容易漏进程。我的排查会先分层：浏览器看系统代理够不够；终端/daemon 看 HTTP_PROXY/HTTPS_PROXY；不吃代理的 app 再看 TUN。开 TUN 后先核对 DNS/fake-ip 和 LAN 排除，不然分流规则会越修越乱。

要翻墙的东西越来越多，从浏览器到docker,npm,python,k8s，再到现在的codex，终于只能用tun 模式了，又要分流…… 好家伙，clash verge 150MB+，mihomo clash甚至要500MB，v2rayN小一点也得120MB+……

这两天折腾代理工具，感慨良多。想当年改一改host 文件就能翻墙，再后来goagent 一个小脚本，再后来得加个小工具找活着的google ip，终于薅完这些ip 了得上shadowsocks 。中间还有昙花一现的蓝灯……终究是v2ray 慢慢成了主流。 再后来clash,single-box,v2rayN这些工具登场，病魔好像也没战胜方教授

@someagi 补一个排查顺序：先确认 Codex 进程实际拿到的是当前 mixed/http 端口；再分清是桌面端 WebSocket 断，还是终端/curl 也断。如果 npm、Docker、daemon 类工具也不稳定，系统代理通常不够，TUN 或显式环境变量会更可靠。最后一起看 DNS/FakeIP，不然规则命中了也可能像断线。

Codex Desktop 老是连不上、反复重试 5 次浪费时间，根源是新版改用 WebSocket 后，桌面应用不会自动吃系统代理，必须手动设置 HTTP_PROXY 和 HTTPS_PROXY 环境变量。 文章详细拆解了原因：在 macOS 上需要查看当前代理端口，然后把变量写入 ~/.codex/.env，保存后重启应用就能解决。还提供了直接发给 Codex 的 Prompt，以及一个一键自动化脚本。 我判断这类本地 AI 编码工具的代理问题会越来越常见。随着桌面端模型越来越多，系统代理配置会成为刚需。未来工具大概率会内置更智能的代理检测和一键配置，或者直接支持系统代理自动读取，否则用户体验会持续被这些小坑拖累。

Roku is awkward because it usually cannot run a VPN app directly. The usual path is routing the network it uses: VPN on the router, a laptop/mobile hotspot sharing the VPN, or using a TV device that supports an app. The key check is scope: is the Roku traffic actually routed, not just your phone or laptop?

@Cory_Hays407 I have no idea how to use a VPN on a Roku tv so I guess we'll wait

REMINDER: #SmackDown will start at 2 PM Eastern today on Netflix for most international viewers. If you’re in the United States, you can watch with a VPN when it’s live or you can watch it on tape delay at 8 PM Eastern on USA Network

我的稳定代理经验：固定一条主路径。全局就 TUN，不要只靠 System Proxy；DNS 交给 Clash/Mihomo，重启后用 `scutil --dns` 查有没有被 DHCP 抢回；IPv6 要么完整进 TUN，要么先关掉；Codex/CLI 这类工具显式写 `HTTP_PROXY/HTTPS_PROXY`。 另外 Clash Verge 本身是 Tauri 客户端，频繁切换模式、反复改配置/订阅，有时会把本地 profile 状态搞乱。遇到“怎么看都对但就是不生效”，可以先删掉订阅 profile，重启客户端，再重新导入订阅和刷新节点。别同时开多个代理/VPN，路由和端口很容易互相覆盖。

@freeguardteam 确实是的，很麻烦，而且还有权限问题 像我之前Mac每次重启之后必定会回到114.114.114.114的DNS，从路由器到Clash各种锁定的方法试完了也没用。 然后又折腾IPv6遇到同样问题，最后不知道改了什么最终解决....

@Lonely__MH Codex + Custom Skills.

@freeguardteam 这是哪家AI写的？

🚀不少人遇到 Codex 每次打开或提问时总是先触发 5 次 “Reconnecting” 重连 这个问题。 这个问题的本质原因：WebSocket（WSS）协议在你的网络或代理环境下无法正常通通信。 解决方案有3种： 1. 修改 Codex 的 config.toml，在对应的 provider 配置里明确禁用 WebSocket 并指定 wire_api： wire_api = "responses"supports_websockets = false 2. 如果使用 Clash verge 的话，直接开启 TUN 模式。 开启后，它会从系统底层接管全局所有流量，强行打通 WebSocket 阻碍。 3. 直接把下面提示词 发给 Codex （具体见文章） ------ 帮我修复 Codex Desktop 一直 Reconnecting 的问题。 请定位我本机正在使用的代理端口和代理协议，然后创建或更新 ~/.codex/.env，写入以下代理配置。不要写死端口，请替换成实际端口；如果文件已经存在，保留其他配置。 HTTP_PROXY="http://127.0.0.1:" HTTPS_PROXY="http://127.0.0.1:" 写入后检查配置是否正确，并告诉我需要如何重启 Codex Desktop。 -----

If several teammates dropped at the same moment and saw the same packet-loss %, I’d treat it as a server/route event first. Save timestamp, region, ISP, and match ID, then test one alternate network. If alternate network is clean, it points to ISP path/peering. If not, Riot-side evidence is stronger.

@riotgames @VALORANT @RiotSupport All of my teammates got disconnected at the same time during a crutial moment, Some got packet loss spike with the same ammount of percentage. This has cost us several rounds and we need your feedback regarding this issue ASAP. It's unplayable.

ME servers are worth testing as a route issue, not just local settings. I’d compare: 1. same region on home ISP vs hotspot 2. packet loss during an actual match 3. one nearby region as a control If only ME is bad across networks, it’s likely server/peering. If hotspot fixes it, ISP routing is the suspect.

NEW SEASON AND MIDDLE EAST SERVERS ARE STILL BAD BTW @Fortnite @FNCompetitive @FortniteStatus