OZ CLI: Remote Fleet Operations From a Terminal
OZ CLI ships as the terminal-native operations interface for the venue fleet. It is not a monitoring dashboard or a web portal. It's a command-line tool that gives operations engineers and AI agents direct, structured access to every venue in the network.
Why a CLI#
Every venue is an edge node running GPU compute, six robotic cameras, networking equipment, and dozens of services. When something needs attention (a gimbal not responding, a GPU running hot, a NATS topic going quiet) the person on call needs answers in seconds, not minutes spent navigating a UI.
OZ CLI connects to any venue through double-hop SSH tunneling. The operator's terminal becomes a direct window into venue state. Every command outputs structured JSON, so the same diagnostic a human reads in a terminal can be consumed by an automated monitoring pipeline, a Slack bot, or an AI agent performing triage.
Core commands#
oz health: one-line-per-category venue health status covering CPU, RAM, GPU, disk, network, and all running services. Green/yellow/red at a glance, structured data underneathoz diag: full 12-metric diagnostic profile per venue: GPU temperature and utilization, load average, disk usage per filesystem, internet speed test, NATS topic health, Memgraph graph database status, camera feed latency, and gimbal responsivenessoz scan: device discovery across the venue network with automatic identification of Jetson modules, Raspberry Pi units, Sony PTZ cameras, and gimbal controllers. Finds every device on the network and reports firmware versions, IP assignments, and connection statusoz subscribe: real-time subscription to any NATS topic at the venue. Stream tracking data, fusion output, gimbal telemetry, or service health events directly to your terminal. Invaluable for debugging perception pipeline issues without touching the venueoz cursor/oz claude: open a remote venue project directly in your development tool of choice. SSH tunnel established automatically, project context loaded, ready to inspect or modify venue-side code- Enclosure control:
oz dome open,oz dome close,oz dome statusper camera or in bulk across the venue. Physical hardware control from the terminal
Remote access architecture#
- Double-hop SSH tunneling to edge devices with automatic key deployment. Connect to the venue gateway, then hop to any internal device (Jetson, camera controller, gimbal) without manual key management
- All commands support
--jsonoutput for agent and automation integration. The same command produces human-readable table output by default and machine-readable JSON with a flag - Structured error responses with machine-readable error codes. When a command fails, the error output tells you exactly what failed, why, and what to check, both for humans reading it and agents parsing it
Fleet-scale operations#
OZ CLI is designed for managing many venues, not just one. The oz health command can target a single venue or sweep the entire fleet. Diagnostics aggregate across venues so an operations engineer can spot which node in a 50-venue network is the outlier, without opening 50 tabs.
Combined with the --json output mode, this turns OZ CLI into the programmatic interface that monitoring systems and AI agents use to maintain fleet health. The NOC does not need a custom dashboard for every diagnostic; the CLI is the dashboard.
Two audiences, one interface#
OZ CLI is built for operations engineers managing the venue fleet and for AI agents that need programmatic access to venue state. The --json flag on every command means the same diagnostic that a human reads in a terminal can be consumed by an automated monitoring pipeline or an AI assistant performing triage. When an alert fires at 3 AM, the on-call engineer and the automated response system use the same tool.