Development
Planned work lives in the issue tracker.
Tooling is pinned with mise; the everyday tasks:
mise run test # unit tests (regenerates manifests first)mise run test-integration # envtest apiserver: CRD schema + CEL rulesmise run test-sanity # upstream csi-test sanity suite, in-processmise run lint # golangci-lintmise run build # bin/miroirmise run manifests # CRD + RBAC generationmise run helm-test # helm-unittest against the chartmise run docs-serve # this docs site, live-reloadingThe end-to-end suite runs against a real Talos cluster booted under QEMU by
talosctl-cluster-action;
CI runs it on every PR, and test/e2e-qemu/README.md documents how to run it locally.
The docs site is Astro Starlight: pages live under
docs/src/content/docs/, the sidebar lives in docs/astro.config.mjs,
and mise run docs builds the site, failing on a broken internal
link. CI runs the same build on every PR.