# NVIDIA Personal AI Router

在同一网络中的兼容计算机之间路由彼此独立的本地推理请求。

> 这是 OpenNavo 上 NVIDIA Personal AI Router 页面的 Markdown 版本，内容与网页相同： <https://opennavo.com/zh/apps/nvidia-pair>

- **最新版本**: `1.0.0`
- **开发者**: NVIDIA
- **主页**: <https://github.com/NVIDIA/Personal-AI-Router>
- **分类**: [AI 工具](https://opennavo.com/zh/categories/ai)
- **安装**: `brew install --cask nvidia-pair`
- **近 30 天安装**: 15
- **Homebrew**: 可以安装，由 Homebrew 更新
- **下载地址**: <https://github.com/NVIDIA/Personal-AI-Router/releases/download/v1.0.0/NVPAIR-Setup-1.0.0-arm64.dmg>
- **SHA-256**: `46a7545ccaeec959e3321f9d200cc9a32c50207e09f277b4d96297394f461f1b`

SHA-256 只用来核对下载的文件是否完整，不代表安全认证。

## 安装与卸载内容

```json
[
  {
    "declaration": {
      "uninstall": [
        {
          "quit": "com.nvidia.nvpair"
        }
      ]
    },
    "phase": "uninstall",
    "sources": [],
    "type": "uninstall"
  },
  {
    "declaration": {
      "app": [
        "NVIDIA PAIR.app"
      ],
      "target": "/Applications/NVIDIA PAIR.app"
    },
    "phase": "install",
    "sources": [
      "NVIDIA PAIR.app"
    ],
    "target": "/Applications/NVIDIA PAIR.app",
    "type": "app"
  },
  {
    "declaration": {
      "zap": [
        {
          "trash": [
            "~/Library/Application Support/com.apple.sharedfilelist/com.apple.LSSharedFileList.ApplicationRecentDocuments/com.nvidia.nvpair.sfl*",
            "~/Library/Application Support/Nvidia Corporation/Personal AI Router",
            "~/Library/Logs/Personal AI Router",
            "~/Library/Preferences/com.nvidia.nvpair.plist",
            "~/Library/Saved Application State/com.nvidia.nvpair.savedState"
          ]
        }
      ]
    },
    "phase": "cleanup",
    "sources": [],
    "type": "zap"
  }
]
```

## 介绍

NVIDIA Personal AI Router (PAIR) 是一款 macOS 桌面应用，可在同一网络中的兼容计算机之间路由本地推理请求。它会发现参与的节点，并根据推理引擎和模型的可用情况以及当前负载，将每个独立请求发送到符合条件的节点。

- 为兼容 Ollama、OpenAI 和 Anthropic Messages API 的请求提供本地代理端点。
- 支持 Ollama、LM Studio 和 llama.cpp 推理引擎；节点必须运行兼容的引擎，并具备请求所需的模型，才能处理请求。
- 可将计算机配对组成集群，并在桌面界面中查看任务和节点活动。
- 每个请求只路由到一个节点；它不会合并 GPU 显存、将模型拆分到多台计算机，也不会把正在处理的请求分给多个节点。

使用 `brew install --cask nvidia-pair` 安装，然后从“应用程序”中打开 PAIR。此 cask 要求 macOS 13 或更高版本，并安装 `NVIDIA PAIR.app`。它会为 Apple 芯片 Mac 选择 arm64 磁盘映像，为 Intel Mac 选择 x64 磁盘映像。引擎和模型各自的要求取决于所选引擎及模型。当配置的客户端、模型来源、引擎和节点均位于本地时，提示和响应旨在留在本地网络中。

来源：[NVIDIA PAIR 项目](https://github.com/NVIDIA/Personal-AI-Router) · [Homebrew cask](https://raw.githubusercontent.com/Homebrew/homebrew-cask/HEAD/Casks/n/nvidia-pair.rb)

## 系统要求

- macOS 13 起 · 应用版本: 1.0.0 · Intel · <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 核查于 2026-11-22
  - <https://github.com/NVIDIA/Personal-AI-Router/releases/download/v1.0.0/NVPAIR-Setup-1.0.0-x64.dmg>
  - SHA-256: `d0366583df2b601a2f374d2d078b1048573a69c1d143c8eb828593dea67533fc`
- macOS 13 起 · 应用版本: 1.0.0 · Apple 芯片 · <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 核查于 2026-11-22
  - <https://github.com/NVIDIA/Personal-AI-Router/releases/download/v1.0.0/NVPAIR-Setup-1.0.0-arm64.dmg>
  - SHA-256: `46a7545ccaeec959e3321f9d200cc9a32c50207e09f277b4d96297394f461f1b`

## NVIDIA Personal AI Router 的安全事实

这里只列可以核验的事实，并注明来源和日期，不是安全认证。要检查你下载的文件，可以用 OpenNavo 在本机扫描。

- <https://github.com/NVIDIA/Personal-AI-Router> · 是否开源: 是 — <https://github.com/NVIDIA/Personal-AI-Router> · 核查于 2026-11-13

## 卸载

- `brew uninstall --cask nvidia-pair`

普通卸载会删除应用本身，设置和缓存会保留。

### 深度清理（可选）

- `brew uninstall --zap --cask nvidia-pair`

`--zap` 还会删除 Homebrew 定义里列出的设置、缓存等文件，其中可能有其他应用共用的数据，删除后无法恢复；它也不保证没有任何残留。请先备份需要保留的数据；如要保留设置和缓存，请不要加 `--zap`。

普通卸载会删除应用本身，设置和缓存会保留。

- 文件和文件夹: `~/Library/Application Support/com.apple.sharedfilelist/com.apple.LSSharedFileList.ApplicationRecentDocuments/com.nvidia.nvpair.sfl*` — <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 以上条目来自 Homebrew 定义，核查于 2026-11-22；它们只说明定义声明会删除什么，不代表你的 Mac 上都有。
- 文件和文件夹: `~/Library/Application Support/Nvidia Corporation/Personal AI Router` — <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 以上条目来自 Homebrew 定义，核查于 2026-11-22；它们只说明定义声明会删除什么，不代表你的 Mac 上都有。
- 文件和文件夹: `~/Library/Logs/Personal AI Router` — <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 以上条目来自 Homebrew 定义，核查于 2026-11-22；它们只说明定义声明会删除什么，不代表你的 Mac 上都有。
- 文件和文件夹: `~/Library/Preferences/com.nvidia.nvpair.plist` — <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 以上条目来自 Homebrew 定义，核查于 2026-11-22；它们只说明定义声明会删除什么，不代表你的 Mac 上都有。
- 文件和文件夹: `~/Library/Saved Application State/com.nvidia.nvpair.savedState` — <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 以上条目来自 Homebrew 定义，核查于 2026-11-22；它们只说明定义声明会删除什么，不代表你的 Mac 上都有。

## 最新版本说明

### 1.0.0 (2026-10-13)

新增 llama.cpp、引擎启动选项、RTX Spark 支持及终端 Jobs。浏览器请求遵循引擎 CORS；macOS 更名为 NVIDIA PAIR.app，需手动删除旧 PAIR.app。旧版安装的 LM Studio 不会自动移除，卸载保留模型。RTX Spark 验证限 Windows on ARM。下附完整英文原文。

#### 经审阅的说明

新增 llama.cpp、引擎启动选项、RTX Spark 支持及终端 Jobs。浏览器请求遵循引擎 CORS；macOS 更名为 NVIDIA PAIR.app，需手动删除旧 PAIR.app。旧版安装的 LM Studio 不会自动移除，卸载保留模型。RTX Spark 验证限 Windows on ARM。下附完整英文原文。

#### 官方完整原文（英文）

This release brings llama.cpp and RTX Spark support to PAIR, and establishes
its public development and contribution workflow on GitHub.

##### What's New

- **llama.cpp support.** Install and manage llama.cpp directly from PAIR,
  download compatible GGUF models from Hugging Face, and route inference
  requests across paired devices. Refer to
  [Managing Engines](engine-lifecycle.mdx).
- **Engine launch options.** Customize the launch arguments and environment
  variables of PAIR-managed engines, on this device or a paired one. Tune
  settings such as context size where the engine supports them. Refer to
  [Engine Settings](engine-settings.mdx).
- **RTX Spark support.** PAIR's Windows on ARM build has been validated with
  llama.cpp and CUDA on RTX Spark N1X hardware.
- **Improved terminal experience.** The terminal interface has a redesigned
  layout organized around each device, so you can manage engines, download
  models, and pair devices from one place. A new **Jobs** tab follows inference
  work across paired devices and runs the Inference Demo, giving headless
  machines the same core tools as the desktop app. Refer to
  [Terminal Interface](terminal-interface.mdx).

##### Compatibility Notes

- **Browser clients.** PAIR's local endpoints now follow each engine's CORS
  policy instead of allowing every browser origin. A browser-based client may
  need the engine configured to allow it — for Ollama, set `OLLAMA_ORIGINS` in
  the engine's launch options.
- **The macOS app is now NVIDIA PAIR.** The new `NVIDIA PAIR.app` installs
  beside the old `PAIR.app`, so remove the old one to keep just one copy. Your
  settings, cluster membership, and downloaded models carry over.
- **LM Studio installed by an earlier PAIR.** It keeps working as before. PAIR
  now removes only engines it has a record of installing, and earlier versions
  kept no record, so uninstalling or resetting PAIR leaves that LM Studio in
  place. You can remove it from `~/.lmstudio` yourself if you no longer want it.

The [Upgrade Guide](upgrade-guide.mdx) walks through each of these.

##### Targeted Bug Fixes

- Uninstalling LM Studio from PAIR, or uninstalling PAIR with its data, no
  longer deletes the models you downloaded. They stay in `~/.lmstudio/models`,
  just as Ollama's stay in `~/.ollama` and llama.cpp's in `~/.llamacpp`.

##### Community Contributions

- Added Anthropic Messages API routing through Ollama and LM Studio, so
  compatible clients such as Claude Code can use PAIR's model-aware routing and
  failover. Thanks [@Ratherrayees](https://github.com/Ratherrayees)!
  ([#27](https://github.com/NVIDIA/Personal-AI-Router/pull/27))
- Made engine monitoring more efficient by reusing HTTP connections for health
  checks. Thanks [@Psych0h3ad](https://github.com/Psych0h3ad)!
  ([#37](https://github.com/NVIDIA/Personal-AI-Router/pull/37))
- Fixed unnecessary cluster trust refreshes: notifications are now sent only
  after new peer endorsements are saved. Thanks [@RecursiveIntell](https://github.com/RecursiveIntell)!
  ([#62](https://github.com/NVIDIA/Personal-AI-Router/pull/62))
- Fixed workload synchronization between nodes: broadcasts preserve event
  order, and a failed broker notification stays eligible for retry. Thanks
  [@mkalkere](https://github.com/mkalkere)! ([#80](https://github.com/NVIDIA/Personal-AI-Router/pull/80))
- Fixed scheduler priority updates being marked as delivered after a failed
  notification, so they are retried and the status stays accurate. Thanks
  [@sylvesterkaczmarek](https://github.com/sylvesterkaczmarek)!
  ([#122](https://github.com/NVIDIA/Personal-AI-Router/pull/122))
- Isolated the service integration and proxy failover tests from local PAIR
  settings, so tests no longer read or overwrite a developer's configuration.
  Thanks [@somtri](https://github.com/somtri)! ([#95](https://github.com/NVIDIA/Personal-AI-Router/pull/95))
- Corrected the release-intent examples to use the required services version
  key. Thanks [@somtri](https://github.com/somtri)!
  ([#120](https://github.com/NVIDIA/Personal-AI-Router/pull/120))
- Updated the contributor instructions to branch from and rebase onto
  `develop`. Thanks [@somtri](https://github.com/somtri)!
  ([#126](https://github.com/NVIDIA/Personal-AI-Router/pull/126))
- Fixed node cards showing only the first GPU on multi-GPU systems; they now
  show every GPU. Thanks [@CryptoJones](https://github.com/CryptoJones)!
  ([#34](https://github.com/NVIDIA/Personal-AI-Router/pull/34))

##### Special Thanks

Thank you to these community members for sharing ideas, submitting proposals
and fixes, investigating issues, and testing PAIR:

[@XxSilentStormIxX](https://github.com/XxSilentStormIxX), [@agenticocho](https://github.com/agenticocho), [@Mayimbe07](https://github.com/Mayimbe07), [@cruzanstx](https://github.com/cruzanstx), [@pettersandvand](https://github.com/pettersandvand),
[@cguldogan](https://github.com/cguldogan), [@mkalkere](https://github.com/mkalkere), [@CryptoJones](https://github.com/CryptoJones), [@woodsonl](https://github.com/woodsonl), [@stevbev](https://github.com/stevbev), [@habakan](https://github.com/habakan), [@arcaven](https://github.com/arcaven),
[@somtri](https://github.com/somtri), [@imluffy000](https://github.com/imluffy000), [@ndizazzo](https://github.com/ndizazzo), and [@fpvo](https://github.com/fpvo).

We are grateful for every contribution, and for the warm welcome from the local
AI community.

#### Sources

- https://formulae.brew.sh/api/cask/nvidia-pair.json
- https://github.com/NVIDIA/Personal-AI-Router/releases/tag/v1.0.0
- https://api.github.com/repos/NVIDIA/Personal-AI-Router/releases/tags/v1.0.0
- https://github.com/NVIDIA/Personal-AI-Router/releases/download/v1.0.0/NVPAIR-Setup-1.0.0-arm64.dmg

## 资料与来源

- <https://opennavo.com/zh/apps/nvidia-pair>
- <https://opennavo.com/zh/apps/nvidia-pair/versions>
- <https://github.com/homebrew/homebrew-cask/blob/HEAD/Casks/n/nvidia-pair.rb> · 核查于 2026-11-22
- <https://formulae.brew.sh/cask/nvidia-pair>
- <https://github.com/NVIDIA/Personal-AI-Router> · 核查于 2026-11-13
