Skip to content

Desktop Console

The desktop console is organised to answer one question quickly: what is the lab doing right now, and should you trust it?

Maabarium guided setup modal showing Git readiness, optional CLI shell integration, and the Pull Recommended Models action.

Guided setup is part of the desktop console experience, not a separate installer screen. It prepares the machine and runtime state that the rest of the console depends on.

Before you spend time in the main console, the setup modal handles the machine and runtime checks that make the rest of the UI meaningful. That includes Git readiness, optional shell CLI integration, workspace selection, local Ollama readiness, and the in-app Pull Recommended Models action.

The modal now also keeps the choice between Guided and Advanced visible. Guided mode favors environment profiles and safer defaults, while Advanced mode exposes lower-level runtime strategy, provider, and research-search controls directly.

The sticky top bar gives you fast movement across:

  • overview metrics
  • console activity
  • workflow library
  • maintenance tools through a direct Maintenance jump target

It also keeps the active workflow, run controls, and workflow actions visible.

The overview section surfaces high-signal metrics first:

  • latest score
  • winner score, including retained-winner context when available
  • average iteration behaviour
  • token usage
  • CPU load and telemetry health

This is the fastest way to see whether the system is healthy before reading raw logs.

The console area is where you spend most of your time during an active run. Expect to use it for:

  • log review
  • proposal review
  • experiment history
  • research evidence
  • runtime signals, Git readiness state, and LoRA runtime metadata when relevant

During setup and early troubleshooting, the desktop flow also surfaces runtime actions directly in context, including Git installation help, Ollama install and start actions, recommended-model pulls, and optional shell CLI integration.

Run Analytics, Evidence, Workflow Library, and Maintenance can now be collapsed independently, and the desktop app restores that panel layout on restart so the console reopens in the shape you last used.

The workflow library is the control point for choosing, filtering, editing, and creating blueprints.

Use it when you need to:

  • switch to a different workflow quickly
  • search by language or workflow type
  • open the blueprint wizard
  • edit an existing workflow definition

When you open the Blueprint Wizard, the guided path now keeps the workflow identity block before goal selection, then walks through output, workspace, runtime, and final review. Choosing a goal updates the derived workflow defaults, but you stay on the same step until you press Next Step. Advanced controls remain available when you need direct access to the underlying blueprint fields.

The guided wizard also keeps the live summary rail and step actions pinned inside their own sticky scroll area, which makes long review passes easier without losing the current navigation controls.

The maintenance section sits below the workflow library and groups operational tasks that matter after initial setup:

  • readiness centre checks
  • the Run Setup re-entry path for reopening guided setup later
  • Git readiness and installer details when Git is missing
  • local runtime actions such as installing Ollama, starting Ollama, and pulling recommended local models
  • optional CLI link management for exposing maabarium in your shell
  • branch cleanup previews
  • persisted stack summaries

That placement is deliberate. You should not have to hunt through the main console to find maintenance tasks.

Edit page

Last updated: