Release 0.8¶
2 September 2026 · Full changelog
Kubus 0.8 makes Kubernetes resources easier to understand and safer to change. You can now inspect and edit a complete object in a structured manifest view, see problems and relationships more clearly in the redesigned details drawer, and follow Helm changes as they happen.
Inspect and edit complete manifests¶
Every resource now has a Manifest tab with two views:
- Tree shows the object as structured fields instead of one large YAML document. It includes field descriptions, data types, readable names for list items, links to related resources and tools to filter or expand the tree.
- YAML provides the familiar raw editor. Tree and YAML use the same draft, so you can switch between them without losing your changes.
The tree is a full editor, not just a viewer. You can change values, add or remove fields, add or remove list items and replace a section with YAML. Before anything is written to the cluster, Kubus shows the complete YAML diff and runs a server-side dry run.
Drafts are kept separately for each resource and remain available after you close the drawer or browse somewhere else. If the live object changes while you are editing it, Kubus tries to replay your changes on top of the new version and reports anything that needs your attention.
To prevent accidental damage, status and identity fields cannot be edited. Secret values
stay masked in the tree, filters and review diff until you explicitly reveal them. Hidden
Secret YAML is read-only, and data, stringData and binaryData are all protected.
Custom resources also have a simpler Schema tab with a version picker and controls to expand or collapse the complete schema.
Resource details that show what is wrong¶
The details drawer has been reorganized around the information you normally need first:
- Every resource starts with a clear status and a compact health summary.
- Pods show one panel per container with its state, restart count, last exit code, image, CPU and memory use, ports, probes, environment, mounts and command.
- Deployments show rollout progress, current replica counts, old ReplicaSets, failing pods and the actual messages that explain a stalled or failed rollout.
- Services show their in-cluster DNS name, port mappings and live EndpointSlice data, including endpoint readiness and links to the backing pods. Missing selectors, unavailable endpoints and permission errors are called out clearly.
- Nodes show health, resource use, system information and scheduled pods in one place.
- Conditions and other long messages wrap instead of being cut off. Sections can be collapsed while still showing a short summary.
Related resources can be opened directly from the drawer. A back button returns you to the resource you started from.
Open details now follow the same live Kubernetes updates as the resource list. Status, related pods, ReplicaSets and endpoints update without reopening the drawer. If an object is deleted, Kubus marks it as deleted, hides actions that no longer apply and can recover if an object with the same name is created again.
Helm changes appear as they happen¶
Helm releases no longer wait for a 30-second refresh. Kubus watches Helm release records and refreshes the list and open release pages when a change comes from Kubus, the Helm CLI, a GitOps controller or another Kubus window.
- A Live or Polling badge shows how each cluster is being refreshed. If Kubernetes permissions do not allow the live watch, Kubus falls back to polling automatically.
- A problem in one cluster no longer hides releases from other connected clusters.
- The release list has text search, Needs attention and Updates available filters, a manual refresh button and per-cluster error messages.
- A release overview shows every Kubernetes object in the release, its rollout or readiness state and a link to its full resource details.
- Upgrade, rollback, diff and uninstall actions are grouped clearly on the release page. The row menu also provides quick access to opening or uninstalling a release.
- Kubus refreshes Helm data after cluster reconnects, kubeconfig changes and socket reconnects, so an old cache is not left on screen.
- Uninstall checks are repeated immediately before the operation starts. Protected clusters still require the release name to be confirmed.
The live watch reads only Helm record metadata. Full Helm release payloads are fetched only when Kubus actually needs the release contents.
Use your own debug container images¶
The new Settings → Debug containers page lets you manage the images offered when you start a pod debug session.
Each entry can have a display name, image reference, description and preferred security profile. Add private toolbox images, pin a different tag, or replace a built-in preset by using the same name. The existing option to enter a one-off custom image remains available.
The catalog is stored in the server's settings.json, so it is shared by Kubus windows
that use the same server.
Smaller improvements and fixes¶
- Topology nodes now stay fixed while the topology controls are locked and become movable only when interaction is enabled.
- Switching tabs prefers matching favourites, and the correct favourite category expands automatically so the active item remains visible.
- The namespace picker is clickable across the complete macOS title bar area.
- The desktop app now closes cleanly instead of occasionally hanging during shutdown.
- Cancelled log streams and open WebSocket connections are closed cleanly during shutdown.
- Settings has a new About page with the Kubus version, platform, licence, project links and support information.
- The documentation has been reorganized and expanded, with updated guides and screenshots for the manifest editor, resource details and Helm pages.
Faster and more reliable cluster connections¶
- Kubernetes watches can share HTTP/2 connections. This keeps resource lists live without opening a separate network connection for every watched resource type.
- Custom resource instances appear in global search as they change, and newly installed CRDs are picked up automatically.
- On connections that cannot use HTTP/2, Kubus uses bounded background scans so large clusters cannot create an unlimited number of watches.
- Watches now work when the Kubernetes API server is reached through a URL path prefix, such as a gateway or reverse proxy.
- Aborted requests are stopped earlier, reducing leaked streams and shutdown delays.
- Bundled dependencies have been updated, including Electron 44 and Kubernetes client 2.0.
Source builds now require Node.js 24.20.0 or newer and use pnpm 11.24.0.
Patches¶
0.8.1¶
3 September 2026 · Changes
- Cmd+C copies selected text in the macOS desktop app even when the window's web content is not reported as focused.