Lab maintenance

Change one definition. Update the lab with control.

Software and system settings are changed in your private lab definition, built on the controller, checked on one workstation, and then deployed to the clients you select.

The operator remains the authority.Nothing reaches a workstation until the operator chooses the target and starts deployment.
UPDATE PATHExisting workstations
  1. 01
    Lab definitionOne declared change
  2. 02
    ControllerEvaluate and build
  3. 03
    Pilot workstationCheck before rollout
  4. 04
    Selected clientsDeploy over the LAN
Operator starts deployment
One prepared system follows a deliberate path from source to the room.
The update process

From a requested change to the room.

The running machines are not edited one by one. Nixorium prepares the declared system centrally and lets the operator expand a verified change from one target to the lab.

  1. 01
    Request

    Describe the result and choose the scope.

    State what should change and whether it applies to one workstation, a group, or the shared lab configuration. No workstation changes at this point.

    A bounded request
  2. 02
    Define

    Change the private lab definition once.

    Your packages, network details, desktop choices, and host-specific settings live in the private deployment generated for the lab.

    One source of truth
  3. 03
    Build

    Evaluate and build on the controller.

    The controller turns the requested definition into the system for the selected machines. Configuration or build errors stop here, before activation on a client.

    A prepared system
  4. 04
    Pilot

    Try the change on one workstation.

    Deploy the prepared system to a representative client and check the real result. An installed machine updates over the LAN; PXE installation is not repeated.

    A checked result
  5. 05
    Roll out

    Update the selected clients.

    The operator chooses the targets. Colmena reaches them over SSH and NixOS activates a new system generation that can be rolled back if necessary.

    An explicit deployment

Client Internet is not required. The controller provides the prepared system through the local cache, so installed clients can receive updates behind a captive portal or where policy blocks direct Internet access.

Installation and updating are different paths. PXE and disk installation are used for a new or recreated machine. Routine updates reach an installed client over SSH.

Agent Skill for your lab

Do not know Nix? Start with the change you need.

You can give a supported coding agent a plain request such as “Add VLC only to pc05.” The nixorium-maintainer skill is included in the generated private deployment and supplies the map and rules for configuring your lab. You do not need to know the Nix syntax or repository layout before asking.

It knows your lab structure.

The skill guides the agent through the private deployment where your machines, packages, network, and site choices are defined.

It keeps the change in scope.

A request for one workstation stays in that host override; a shared classroom change goes into the common client configuration.

It chooses relevant checks.

A one-PC package change and a change that affects the offline installer follow different validation paths.

This is Nixorium’s agentic layer: the agent inspects, edits, and validates the private lab configuration. You still review the intended result and decide whether and where to deploy it.

Read the Agent Skill Discoverable by Codex, OpenCode, Claude Code, and Pi.
Public frameworkShared Nixorium behaviorUsually left unchanged
Private lab deploymentYour machines and choicesWhere the pc05 request belongs
  1. 01Request“Add VLC only to pc05”
  2. 02ConsultRead the Nixorium Agent Skill
  3. 03LocateFind the pc05 site override
  4. 04CheckEvaluate and build pc05
  5. 05ResultPatch and checks ready for review
Validation receipt
  • The change stays scoped to pc05
  • The affected system evaluates and builds
  • The offline delivery path remains intact
Operator review
The skill turns project-specific rules into a concrete path the coding agent can follow.
Continue evaluating

See what happens behind the update.

Read how the controller builds systems, serves the local cache, reaches clients through Colmena, and leaves deployment under operator control.