proxies.live

内网加速节点 · 镜像 / 包 / 模型

把原始地址换成下面的代理地址即可。大文件首次拉取由服务器全速缓存、之后走本地，不会因为慢被掐断。登录凭据向管理员获取。

状态

—

服务在线

—

镜像缓存

—

缓冲层

—

Web 缓存

—

apt 缓存

—

磁盘

—

Docker 镜像加速

需 docker login

docker login hub.docker.proxies.live -u <用户名>
docker pull hub.docker.proxies.live/library/nginx

hubdocker.io

ghcrghcr.io

gcrgcr.io

k8s-gcrk8s.gcr.io

k8sregistry.k8s.io

quayquay.io

mcrmcr.microsoft.com

elasticdocker.elastic.co

nvcrnvcr.io

用法：把上游域名换成对应子域名，例如 ghcr.io/owner/img → ghcr.docker.proxies.live/owner/img。镜像浏览器：registry.docker.proxies.live

GitHub 加速

免登录

ghproxygithub.proxies.live

git clone https://github.proxies.live/https://github.com/<用户>/<仓库>
wget https://github.proxies.live/https://github.com/<用户>/<仓库>/releases/download/v1/x.zip

语言 / AI 包加速

免登录

# npm
npm config set registry https://npm.proxies.live
# pip
pip config set global.index-url https://pypi.proxies.live/index/
# conda
conda config --set channel_alias https://conda.proxies.live
# PyTorch / CUDA
pip install torch --index-url https://torch.proxies.live/whl/cu121
# Go
go env -w GOPROXY=https://goproxy.proxies.live,direct
# HuggingFace
export HF_ENDPOINT=https://hf.proxies.live

npmnpmjs

pypiPyPI

condaconda-forge

torchpytorch wheels

goproxyGo modules

hfHuggingFace

系统 / 通用

免登录

apt 缓存apt.proxies.liveUbuntu / NVIDIA CUDA …

# /etc/apt/sources.list 把上游域名前面加 https://apt.proxies.live/
deb https://apt.proxies.live/archive.ubuntu.com/ubuntu jammy main universe
deb https://apt.proxies.live/developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64 /

通用 / Helmcache.proxies.live直链文件 / chart

https://cache.proxies.live/https/<host>/<path>

通用缓存对会 302 跳转的上游（如 GitHub Releases）只透传不缓存，这类用 github.proxies.live。