github zerx-lab/FluxDown v0.5.4

3 hours ago
📦 下载与安装说明(服务器 / CLI)· 文件校验见 SHA256SUMS.txt

📦 Headless 服务器:直接下载运行(免 Docker)

平台 资产
Windows x64 / ARM64 FluxDown-Server-0.5.4-windows-{x64,arm64}.zip
macOS Intel / Apple Silicon FluxDown-Server-0.5.4-macos-{x64,arm64}.tar.gz
Linux x64 / ARM64(musl 静态链接,任意发行版可用) FluxDown-Server-0.5.4-linux-{x64,arm64}.tar.gz
OpenWrt 软路由(x86_64 / aarch64 各子架构) fluxdown-server_0.5.4_<arch>.ipk + luci-app-fluxdown_0.5.4_all.ipk
QNAP NAS(QTS x86_64 / ARM64) FluxDown-Server-0.5.4-qnap-{x64,arm64}.qpkg
群晖 NAS(DSM 7 / DSM 6,x86_64 / ARM64) FluxDown-Server-0.5.4-synology-{dsm7,dsm6}-{x64,arm64}.spk

解压后运行 fluxdown-agent --server(Windows 为 fluxdown-agent.exe --server),浏览器访问 http://127.0.0.1:17800。
包内两个可执行文件必须放在同一目录:fluxdown-agent(含 Web 控制台,已编译进二进制)与 fluxdownd(下载 daemon,由 agent 自动拉起、退出时一并关停),无需附带任何静态资源目录。
首次启动请在浏览器里完成「初始化 FluxDown Server」向导设置访问密钥(无人值守部署可用 FLUXDOWN_TOKEN 预置)。

OpenWrt:opkg install fluxdown-server_*.ipk luci-app-fluxdown_*.ipk,LuCI「服务 → FluxDown」页管理,Web 界面在 http://路由器IP:17800。agent 二进制含完整 Web UI,另附 fluxdownd,需 overlay 空间充足的软路由设备。
QNAP:App Center → 手动安装 .qpkg(需在设置中允许安装未经验证的应用),装完点应用图标直达 Web 界面。
群晖:套件中心 → 手动安装 .spk(需在「设置 → 信任来源」允许任何发行者)。DSM 7.x 装 dsm7 包(以套件专属用户运行),DSM 6.x 装 dsm6 包;Intel/AMD 机型选 x64,ARM 机型(rtd1296/rtd1619b 等)选 arm64。装完在套件中心点「打开」直达 Web 界面(端口 17800)。

🐳 Headless 服务器:Docker(linux/amd64 · linux/arm64)

docker pull ghcr.io/zerx-lab/fluxdown-server:0.5.4
docker run -d -p 17800:17800 -v fluxdown-data:/data \
  ghcr.io/zerx-lab/fluxdown-server:0.5.4

Compose 部署见仓库 docker/docker-compose.yml。首次打开 Web 界面时按向导设置访问密钥,或用 -e FLUXDOWN_TOKEN=... 预置。

📦 命令行 CLI

平台 资产
Windows x64 / ARM64 FluxDown-CLI-0.5.4-windows-{x64,arm64}.zip
macOS Intel / Apple Silicon FluxDown-CLI-0.5.4-macos-{x64,arm64}.tar.gz
Linux x64 / ARM64(musl 静态链接,任意发行版可用) FluxDown-CLI-0.5.4-linux-{x64,arm64}.tar.gz

解压后得到 fluxdown(Windows 为 fluxdown.exe),放入 PATH 即可使用;也可从源码安装:
cargo install --git https://github.com/zerx-lab/FluxDown fluxdown_cli。

fluxdown ping                         # 探活(需 FluxDown App / server 运行中)
fluxdown add https://example.com/f.zip
fluxdown ls

服务地址经 --url 或 FLUXDOWN_URL(默认 http://127.0.0.1:17800),管理 token 经 --token 或 FLUXDOWN_TOKEN。

0.5.4 - 2026-10-06

🚀 新功能

  • 桌面客户端支持调整界面字体大小(10~24 px),设置页「外观」中可选择,也可选「跟随主题」;与界面缩放叠加生效,文字不会被裁切
  • 账户新增密码管理:可修改密码、设置密码(通过当前密码或邮箱验证码验证),登录框内提供「忘记密码」重置流程,安全页面新增对应入口;验证码发送后重开对话框会继续倒计时
  • 自定义主题库现在随账户同步,在多台设备间共享;旧版本地主题文件会自动迁移
  • Web 控制台:当服务器与浏览器在同一台机器上(本机回环访问)时,任务右键菜单提供「打开文件」「打开所在文件夹」,进行中的任务也可打开所在文件夹;远程或经反向代理访问时不会出现该菜单。如需在服务器模式下显式放行,可设置 FLUXDOWN_ALLOW_LOCAL_PLATFORM
  • yt-dlp 插件升级:优先选择 H.264,下载后检测实际文件编码,对 AV1/VP9/HEVC 等系统播放器(如 QuickTime)无法解码的视频自动转码为 H.264/AAC/MP4(受 preferMp4 开关控制,兼容的 H.264 仅重封装不重编码);选择器优先考虑分辨率,并扩充了 Instagram、Twitter(X)、TikTok、Reddit、Facebook 的接管支持
  • yt-dlp 输出上限可通过配置调整(默认 16 MiB),超限时给出明确错误提示而不是「Unexpected end of JSON input」

🐛 问题修复

  • 修复启用 CORS 时被强制生成访问令牌的问题:现在仅在局域网监听且开启接管或 aria2 接口时才要求令牌,单独开启 CORS 可以保持令牌为空
  • 修复 YouTube 等站点因 yt-dlp 返回的元数据过大(单个视频超过 4 MiB)而解析失败的问题
  • 修复 HLS 流中 #EXT-X-KEY:METHOD=NONE 之后的明文分片仍沿用上一段 AES-128 密钥解密、导致下载失败的问题
  • 优化设置页 User-Agent 一行的布局:说明与预设下拉按卡片宽度自动换行,自定义输入框占满整行

0.5.4 - 2026-10-06

🚀 Features

  • Desktop client: adjustable UI font size (10–24 px) under Settings → Appearance, with a "Follow theme" option; it stacks with UI scaling and text is no longer clipped
  • Account password management: change or set a password (verified by current password or email code), a "Forgot password" reset flow in the login dialog, and a new entry on the security page; reopened dialogs resume the verification-code countdown
  • The custom theme library now syncs with your account across devices; old local theme files are migrated automatically
  • Web console: when the server and browser are on the same machine (loopback access), the task context menu offers "Open file" and "Open folder", and in-progress tasks can also reveal their folder; the menu does not appear for remote or reverse-proxied access. Set FLUXDOWN_ALLOW_LOCAL_PLATFORM to opt in explicitly in server mode
  • yt-dlp plugin upgrade: prefers H.264, inspects the actually downloaded file, and automatically transcodes AV1/VP9/HEVC videos that system players (e.g. QuickTime) can't decode to H.264/AAC/MP4 (gated by preferMp4; compatible H.264 is only remuxed, not re-encoded); the format selector now prioritizes resolution, and Instagram, Twitter(X), TikTok, Reddit and Facebook are supported for takeover
  • The yt-dlp output cap is now configurable (default 16 MiB), and an over-cap result reports a clear error instead of "Unexpected end of JSON input"

🐛 Bug Fixes

  • Fixed an access token being force-generated when CORS is enabled: a token is now required only when listening on the LAN with takeover or the aria2 endpoint enabled, so enabling CORS alone may keep an empty token
  • Fixed yt-dlp resolution failing on sites such as YouTube because the returned metadata was too large (over 4 MiB for a single video)
  • Fixed HLS streams where plaintext segments following #EXT-X-KEY:METHOD=NONE were still decrypted with the previous AES-128 key, causing the download to fail
  • Improved the User-Agent row layout in Settings: the description and preset dropdown wrap to the card width, and the custom input fills the full row

Don't miss a new FluxDown release

NewReleases is sending notifications on new releases.