NVIDIA/Personal-AI-Router

Router that virtually distributes inference across connected devices in the home.

What it solves

NVIDIA Personal AI Router (PAIR) addresses the challenge of managing multiple local computers on a single network to handle AI inference workloads. It allows users to distribute independent AI requests across several machines, preventing any single machine from becoming a bottleneck, which is particularly useful for concurrent workloads like multi-agent applications.

How it works

PAIR acts as a local inference router. It discovers participating nodes on the network, manages supported inference engines (such as Ollama and LM Studio), and provides OpenAI-compatible and Ollama-compatible proxy endpoints. When a request arrives, PAIR routes it to an eligible node based on engine availability, model availability, and current workload.

Who it’s for

Developers and AI enthusiasts who have multiple local machines (Windows, Linux, or macOS) and want to create a local cluster for AI inference without sending data to the cloud.

Highlights

  • Cross-Platform Compatibility: Supports Windows 11, Linux, and macOS (x64 and arm64).
  • Unified API: Provides proxy endpoints that are compatible with OpenAI and Ollama.
  • Easy Setup: Includes a desktop application for installation and management of engines and models.
  • Local-First Privacy: Designed to keep prompts and responses on the local network.
  • Dynamic Routing: Routes requests based on live GPU and memory usage signals.

Related

  • Project
    agentplugins/agent-plugins-spec

    A vendor-neutral standard for packaging reusable AI agent skills and MCP servers into portable, distributable plugins.

  • Project
    NVIDIA-Merlin/Transformers4Rec

    A PyTorch-based library that brings Hugging Face Transformer architectures to sequential and session-based recommendation systems, supporting rich tabular input features and GPU acceleration.

  • Project
    d-kimuson/claude-code-viewer

    A web-based client for Claude Code that provides a modern interface for managing sessions, searching conversation history, and monitoring tasks in real-time.

  • Project
    kuafuai/aipexbase

    A Backend-as-a-Service infrastructure that provides AI-native backend capabilities, allowing developers to build full-stack AI applications without writing backend code.