← Back to Skills Marketplace
smseow001

Exo Cluster

by SMS · GitHub ↗ · v1.0.1 · MIT-0
cross-platform ✓ Security Clean
49
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install exo-cluster
Description
Exo 分布式 AI 集群 - 将多台 Mac/PC/WSL2 设备整合为统一 GPU 集群,本地运行大模型。支持 DeepSeek、Qwen、LLaMA。
README (SKILL.md)

Exo 分布式 AI 集群

将闲置设备整合为统一算力集群


概述

Exo 可将多台 Mac、PC、手机等设备整合为一个 GPU 集群,在本地运行大模型。

项目 信息
Stars 44.5K
官网 https://exolabs.net/
GitHub exo-explore/exo

支持的模型

  • DeepSeek V3 / V2
  • Qwen3-235B
  • Kimi2-Thinking
  • LLaMA (MLX)
  • Mistral
  • LlaVA

触发词

搭建集群 / exo集群 / 本地部署大模型 / 分布式推理


环境准备

Mac (推荐)

# 1. 安装 Xcode
xcode-select --install

# 2. 安装 Homebrew
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# 3. 安装依赖
brew install uv node rust

# 4. 安装 Exo
git clone https://github.com/exo-explore/exo
cd exo

WSL2 / Windows

# 方式1: 用社区原生版
git clone https://github.com/tensorsofthewall/exo_windows
exo_windows

# 方式2: WSL2 中运行
git clone https://github.com/exo-explore/exo
cd exo
pip install -e .
python -m exo.main --role master

Linux + NVIDIA

# 安装 CUDA 和相关驱动
# 然后同样 clone exo
git clone https://github.com/exo-explore/exo
cd exo

快速开始

方式一:Nix(最简单)

# 如果有 Nix
nix run .#exo

方式二:手动安装

cd exo

# 启动主节点
python -m exo.main --role master --port 8080 --name "home-cluster"

# 其他设备加入(从节点)
python -m exo.main --role worker --master-addr 主节点IP:8080 --name "设备名"

# 低配设备轻量模式
python -m exo.main --role worker --master-addr 主节点IP:8080 --light-mode

WSL2 快速启动

# 1. 克隆
cd ~
git clone https://github.com/tensorsofthewall/exo_windows
exo_windows

# 2. 安装依赖
pip install -e .
# 或
pip install torch tinygrad

# 3. 启动
python -m exo.main --role master --port 8080 --name "wsl2-cluster"

# 浏览器打开 http://localhost:52415

查看和部署模型

查看可用模型

python -m exo.master.api --list-available-models

部署模型

# 部署 DeepSeek 7B
python -m exo.master.api --deploy-model deepseek-7b --min-nodes 2 --max-nodes 4

# 部署更大的模型
python -m exo.master.api --deploy-model qwen3-235b --min-nodes 2 --max-nodes 4

使用 API

Exo 提供多种 API 兼容:

OpenAI 风格

curl http://localhost:52415/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-7b",
    "messages": [{"role": "user", "content": "你好"}]
  }'

Claude 风格

curl http://localhost:52415/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-7b",
    "messages": [{"role": "user", "content": "你好"}]
  }'

Ollama 风格

curl http://localhost:52415/api/generate \
  -d '{"model": "deepseek-7b", "prompt": "你好"}'

集群管理

打开 Dashboard

http://主节点IP:52415

查看集群状态

python -m exo.main --status

硬件诊断

python -m exo.utils.info_gatherer.system_info --detailed

性能优化

推荐配置

设备数量 推荐模型 内存要求
1 台 7B ~16GB
2 台 14B ~32GB
4 台 70B ~128GB
4 台 Mac Studio 512GB 671B (8-bit) 2TB 统一内存

优化技巧

  1. 使用 RDMA over Thunderbolt(Mac 之间)
  2. 启用 Tensor Parallel
  3. 模型量化(8-bit / 4-bit)
  4. 调整任务分配策略

常见问题

Q: 设备发现不了?

A: 确保在同一局域网,使用有线或 5GHz WiFi

Q: 跑不动大模型?

A: 从 7B 开始测试,确认单设备能跑再增加节点

Q: 速度慢?

A: 使用 Thunderbolt 联网,或启用 RDMA


快速指令表

需求 命令
启动主节点 exo --role master
设备加入集群 exo --role worker --master-addr IP
部署模型 exo --deploy-model 模型名
查看模型列表 exo --list-models
查看状态 exo --status
打开管理界面 http://localhost:52415

与 OpenClaw 集成

可以配合使用:

  1. 在 Mac 上部署 Exo 集群
  2. 通过 API 调用模型
  3. OpenClaw 作为前端交互

示例:

# 让集群运行推理
curl -X POST http://mac-server:52415/v1/chat/completions \
  -d '{"model": "deepseek-7b", "messages": [{"role": "user", "content": "解释量子计算"}]}'

Exo | 分布式 AI 集群框架

Usage Guidance
Install only from trusted Exo repositories, review commands before running them, and secure the LAN-facing master/API ports. This skill appears purpose-aligned and instruction-only, but running an AI cluster still means you are installing external software and exposing local services.
Capability Analysis
Type: OpenClaw Skill Name: exo-cluster Version: 1.0.1 The skill bundle provides documentation and setup instructions for 'Exo', a legitimate open-source distributed AI cluster framework. The instructions in SKILL.md include standard installation procedures (e.g., Homebrew, git clone, pip install) and command-line usage for managing GPU clusters, with no evidence of malicious intent, data exfiltration, or prompt injection.
Capability Assessment
Purpose & Capability
The documented commands and APIs match the stated purpose of installing and operating a distributed local AI inference cluster.
Instruction Scope
The instructions are manual setup examples, not automatic execution, but they include starting master/worker services and using HTTP APIs, so users should understand the network exposure.
Install Mechanism
The skill has no bundled code or install spec, but its guide tells users to install package managers/dependencies and clone external repositories without pinned versions.
Credentials
Using local machines, GPUs, LAN addresses, and model-serving ports is proportionate for a cluster skill, but firewall/authentication guidance is not described.
Persistence & Privilege
Master and worker processes are intended to keep running while the cluster is in use; there is no evidence of hidden persistence, privilege escalation, or automatic startup.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install exo-cluster
  3. After installation, invoke the skill by name or use /exo-cluster
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
- 增加对 WSL2 和 Windows 的支持,补充相关环境准备及快速启动说明 - 关键词中加入 WSL2、Windows - 更新环境准备章节,详细覆盖原生 Windows 与 WSL2 用户的安装/启动流程 - 版本号从 1.0.0 升级至 1.0.1
v1.0.0
Initial release of exo-cluster. - Combine multiple Mac/PC devices into a unified local GPU cluster for running large AI models. - Supports popular models including DeepSeek, Qwen, LLaMA, Mistral, and LlaVA. - Provides step-by-step environment setup for macOS and Linux with NVIDIA GPUs. - Easy cluster management, deployment, and monitoring through CLI and web dashboard. - Offers OpenAI, Claude, and Ollama-style API compatibility. - Includes tips for performance optimization and troubleshooting common issues.
Metadata
Slug exo-cluster
Version 1.0.1
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 2
Frequently Asked Questions

What is Exo Cluster?

Exo 分布式 AI 集群 - 将多台 Mac/PC/WSL2 设备整合为统一 GPU 集群,本地运行大模型。支持 DeepSeek、Qwen、LLaMA。 It is an AI Agent Skill for Claude Code / OpenClaw, with 49 downloads so far.

How do I install Exo Cluster?

Run "/install exo-cluster" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is Exo Cluster free?

Yes, Exo Cluster is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does Exo Cluster support?

Exo Cluster is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created Exo Cluster?

It is built and maintained by SMS (@smseow001); the current version is v1.0.1.

💬 Comments