EXO: End-to-End Local AI Cluster Deployment Training Course

Introduction to EXO and Local AI Clustering

• Overview of the EXO framework and the exo-explore ecosystem.
• Comparison of centralized cloud inference versus distributed local inference.
• Architecture: libp2p device discovery, MLX backend, dashboard, and API layers.
• Hardware requirements: Apple Silicon (M3 Ultra, M4 Pro/Max), Thunderbolt 5, and shared storage.

Installing EXO on macOS

• Setting up Xcode, Metal Toolchain, and macOS prerequisites.
• Installing uv, Node.js, and the Rust nightly toolchain.
• Installing the pinned macmon fork for Apple Silicon monitoring.
• Cloning the repository and building the dashboard using npm.
• Running EXO from source and verifying access to the localhost:52415 dashboard.

Installing EXO on Linux

• Installing dependencies via apt or Homebrew on Linux.
• Configuring uv, Node.js 18+, and the Rust nightly toolchain.
• Building the dashboard and running EXO in CPU-only mode.
• Directory layout: XDG Base Directory paths for configuration, data, cache, and logs.

Automatic Device Discovery and Cluster Formation

• Understanding libp2p-based auto-discovery across local networks.
• Configuring custom namespaces using EXO_LIBP2P_NAMESPACE for cluster isolation.
• Verifying node membership within the dashboard's cluster view.
• Addressing discovery failures and network segmentation issues.

Enabling RDMA over Thunderbolt 5

• RDMA architecture and the reported 99 percent reduction in latency.
• Enabling RDMA in macOS Recovery mode using rdma_ctl.
• Cable requirements and port topology constraints specific to Mac Studio.
• Ensuring macOS versions match across all cluster nodes.
• Troubleshooting RDMA discovery and DHCP configuration.

Deploying Frontier Models

• Using the dashboard to load and shard DeepSeek v3.1, Qwen3-235B, and Llama family models.
• Previewing instance placements via the /instance/previews API endpoint.
• Creating model instances using pipeline or tensor-parallel sharding.
• Configuring custom model cards from the HuggingFace hub.

Monitoring and Troubleshooting

• Reading EXO logs and understanding distributed tracing mechanisms.
• Interpreting cluster health status in the dashboard's cluster view.
• Diagnosing worker node failures and reconnection behaviors.
• Utilizing EXO_TRACING_ENABLED for performance bottleneck analysis.

Cluster Maintenance and Updates

• Updating EXO binaries and procedures for dashboard rebuilds.
• Migrating model caches and managing pre-downloaded models over NFS.
• Gracefully removing nodes and rebalancing workloads.