The runner page

Live steps execute inside a runner container, never on your machine directly. The Runner page (the Runner tab in the top navigation) manages the two things that makes that work: the connection to a Docker-API-compatible container engine, and the runner image itself.

Container engine

The status bar LED shows engine health at all times, and the same probe backs this page. Overwire talks to any Docker-API-compatible engine (Docker Desktop, Colima, OrbStack, Rancher Desktop). Without one, mock runs keep working; only live execution needs the engine. The CLI’s doctor command runs the equivalent checks in the terminal.

Runner image

The image section shows whether the configured image is present locally, its compatibility version badge, size, creation date, and digest, with a Recheck button to re-probe. The image reference defaults to overwire/runner:ubuntu-24.04 and can point at any tag you have built or mirrored; Pull image fetches it with progress and a cancel control.

The version badge matters: the app and the image share a compatibility contract, and an older image still runs but may fall back on newer features. For example, JavaScript actions declaring using: node16 or node24 get the matching side-by-side Node runtime from a v2 image, and fall back to the image’s default Node with a step diagnostic on older images.

Auto-pull on missing image controls what happens when a live run starts without the image present: pull automatically, or fail fast if you only want locally built images.

Tool cache

The tool cache is a host directory bind-mounted into every job container as RUNNER_TOOL_CACHE, so tools installed by setup-* actions persist across runs instead of downloading every time. The host directory is configurable here; leave it empty for the default location. The CLI exposes the same cache through overwire cache tool-cache for listing and evicting entries.