Alpha

Under active development. Breaking changes expected. APIs, installers, and UI may shift between releases.

COMPUTE NODEIn Development

The lean, headless worker

A compute node is the lean sibling of the Workstation. It runs the same GPU engine with none of the interface: a headless worker that accepts reconstruction and inference jobs from the cluster master and returns results. It is the profile you add when one machine is not enough.

See the cluster View on GitHub

In Development

The compute-node profile name is reserved and the ados-compute engine is in place. The standalone headless worker is being built. Run heavy compute on the Workstation today, and add dedicated workers as this profile matures.

0UI

Headless

GPU

Only job

Slave

Cluster role

Lean

Sibling of Workstation

THE IDEA

A worker, not a console

The Workstation carries the GPU engine and all the UI and the ground link. The compute node keeps only the engine. No dashboard, no video, no ground link, just a machine that does the heavy math and hands the answer back.

What a worker does

Job from masterover the cluster contract

reconstruct or inferdataset attached

Run on the GPUheadless, no UI

splat / cloud / meshor vision inference

Result to masterback to the operator

world model or detections

A compute node has no operator interface. It receives a job over the cluster contract, runs it on its accelerator, and returns the result to the master.

NAMING

Workstation vs compute node

Both run the ados-compute engine. The Workstation is the operator's whole machine and typically the master. The compute node is a stripped-down worker that exists only to add capacity.

	Workstation	Compute node
Role	Operator console	Headless worker
GPU compute engine	✓	✓
Mission Control UI	✓	—
Ground link	✓	—
Cluster position	Master (typical)	Worker
Status	Available	In development