Skip to content

Install the Screensaver Worker

The ModelReins Companion is a screensaver that doubles as a fleet worker. When your machine is idle, it picks up jobs and runs them locally with Ollama. When you come back, it gets out of the way.

  • One installer, no shell config
  • Auto-pulls a small useful model (qwen2.5:7b, ~4 GB) on first run
  • Real OS screensaver registration — works with your existing power settings
  • Beautiful split-screen dashboard while it runs (your activity + the fleet)
  1. Grab the installer:
  2. Run it. Standard installer flow.
  3. On first launch, the setup window asks for:
    • ModelReins URLhttps://app.modelreins.com (default)
    • API token — paste from app.modelreins.com/settings/api-keys
    • Worker name — defaults to your hostname; change if you want

The companion checks for Ollama at 127.0.0.1:11434:

StateWhat happens
Ollama not installedCompanion runs in screensaver-only mode. Install Ollama from ollama.com, then restart the companion.
Ollama installed, no modelsCompanion auto-pulls qwen2.5:7b. ~4 GB, one-time, takes a few minutes. The worker comes online when it finishes.
Ollama installed with modelsCompanion picks the largest installed model. Worker comes online immediately.

When the screensaver activates (your OS idle threshold), you see:

Left pane — YOU

  • Your worker name + model
  • Status (idle / working / streaming)
  • Today’s jobs / tokens / cost / uptime
  • Active job card (prompt preview when running)
  • Recent jobs (your last 6)
  • Lifetime jobs + tokens

Right pane — FLEET

  • Carousels through ambient pages: fleet, jobs, metrics, leaderboard, audit, review queue, lessons learned, trends, providers, in memoriam, event log
  • Auto-rotates every 8 seconds, skips empty pages

Press Esc or move the mouse to dismiss.

After install, the companion registers itself with your OS screensaver system:

  • Windows: copies a .scr to System32 and sets the registry. Open Settings → Personalization → Lock Screen → Screen saver settings to confirm “ModelReins” is selected.

If you’d rather just run it in the foreground without OS integration, launch the app and it’ll show whenever the OS idle event fires.

The default is qwen2.5:7b because it’s the best per-byte intelligence in the small class. Want something else?

Terminal window
# Set the env var, then restart the companion
export MODELREINS_DEFAULT_MODEL=llama3.2:3b

Or pull a model manually with ollama pull <name> and the companion will pick it up.

Companion shows “screensaver-only” instead of “Ready” Ollama isn’t running. Install from ollama.com, confirm curl http://127.0.0.1:11434/api/tags works, restart the companion.

Auto-pull fails Disk space, network, or Ollama daemon issue. Run ollama pull qwen2.5:7b manually and restart the companion.

Worker shows offline in the saddle Check the companion’s tray icon — it should be there. If not, the app crashed. Logs are in your OS app logs directory.