Skip to content
ELEMENT 31

Czar

Czar is Element 31's appliance: a sealed node installed in a facility's own rack that provides the inference pool for the robots on that floor.

· 2 min read

IN SHORT

Czar is Element 31's appliance: a sealed node installed in a facility's own rack that provides the inference pool for the robots on that floor. It runs a signed, read-only software image, attests what it is running before it joins a Grid, and carries a provenance record, so the site can say what is in the rack from evidence.

Czar is Element 31's appliance: a sealed node installed in a facility's own rack that provides the inference pool for the robots on that floor. One Czar serves a small floor; several Czars joined together form a Czar Grid.

Why it matters

Operators who want a pool in the building usually face a build project: servers, a serving stack, a security review, and a sustainment plan stitched together from several suppliers. Czar packages all of that as one node that arrives sealed and goes into a rack the site already owns.

Because it is sealed, the site does not inherit an open system to harden. The software image is signed and read-only. The enclosure is sealed. Each node attests what it is running before it joins a Grid or accepts a task, and its integration, sealing, and software load are on record.

Because it is in the building, the robots reach it over the facility network. Camera feeds, captures, task records, and adapted model weights stay on the operator's hardware.

How it works at Element 31

Czar runs Substrate, the serving system that handles inference, each robot's latency budget, memory, retrieval over the site's documents, adaptation on the site's own captures, evaluation, integrity verification, and governed connectivity. Agency Studio, the task layer, runs on the same node.

Posture is a choice per site, not a different appliance. The same Czar can run air-gapped, with updates carried in as signed bundles on media, or connected only to the systems the site already secures: the robot vendors' fleet software, the warehouse system, the PLCs, and nothing the site did not declare.

Czar is the only appliance. A Czar Grid is several Czars sharing one larger workload, still Czar. Czar Chassis is the same Grid shipped by a robot vendor or integrator under their own name, still Czar. There is no second machine to evaluate.

Commercially, a site subscribes to the configured Grid: hardware, platform license, and sustainment in one line, fixed for the term. More robots on the pool do not move it. How to buy describes the two routes.

Questions this raises

Is one Czar enough for a facility?
One Czar is a pool for a small floor. A whole facility joins several Czars into a Czar Grid, which the robots see as one pool. The Grid is still Czar, not a different product.
Does a Czar need an internet connection to work?
No. A Czar can run air-gapped, with no path out of the building and updates carried in as signed bundles on media. A site can instead connect it only to the systems it already secures, and both postures use the same appliance.
What runs on a Czar?
Substrate, the serving system, handles inference, latency budgets, memory, retrieval, adaptation, evaluation, integrity verification, and governed connectivity. Agency Studio, the task layer, runs on the same node and records every site task run.

KEEP READING

Related.