One robot working is a demo. Twenty robots working is a product. This guide covers the jump — shipping a versioned autonomy bundle across a fleet without holding your breath.
Bundle your stack
A deploy starts by packaging your nodes into a signed, versioned bundle:
rome bundle create --tag v2.4.1
The bundle is immutable and content-addressed, so what you test is exactly what ships.
Roll out in stages
Never deploy to everything at once. Start with a canary and widen only if it stays healthy:
rome deploy v2.4.1 --canary 10% # 10% of the fleet
rome deploy v2.4.1 --canary 50% # widen if healthy
rome deploy v2.4.1 --all # full rollout
Rome watches the intervention rate and fault count during each stage. If either crosses your threshold, the rollout pauses automatically.
Roll back instantly
Because bundles are immutable, rollback is just pointing the fleet at the last known-good version:
rome deploy rollback --to v2.4.0
There's no rebuild and no scramble — every robot reverts in seconds.
You've completed the path
You can now:
- Close a command/telemetry loop.
- Calibrate sensors for a clean world model.
- Ship behaviors to a whole fleet safely.
From here, dig into the blog for how we think about the harder autonomy problems, or jump into the feature tour.