v1.4.29

Kubernetes Explorer

The Kubernetes page is now an explorer for triage. The overview shows cluster
health, capacity use, grouped issues, top pod usage, and the last hour of
changes, with summary cards for topology, Helm, and traffic.

New tabs cover Issues, a change Timeline, resource Topology, Helm releases,
GitOps applications, and Argo Rollouts. Every resource opens in one drawer
with events, scrubbed logs, safe YAML, related resources, metrics, and a
diagnosis. The page updates live, and ⌘K finds any resource.

Secret, ConfigMap, and environment values are never shown. Partial or
unavailable data is labeled rather than shown as healthy.

AI SRE Cluster Understanding

Chat, Analyze, and the CLI read the same cluster evidence as the dashboard:
grouped issues, recent changes, relationships, workload logs, Helm, and
GitOps status. One diagnosis call gives the agent a workload's status,
warning events, worst-pod logs, recent changes, and neighbors.

Kubernetes changes now count as recent changes for incident correlation, even
without GitHub. Investigate opens Chat with the resource already attached.

Approval-Gated Actions

With a separately configured actor credential, restart, scale, CronJob
trigger or suspend, and node cordon are available as proposals. Each runs a
dry-run first, requires human approval, and is recorded on the timeline.
Delete, drain, and rollback produce a manual guide only. The reader
connection stays read-only.

Enterprise

• Service traffic shows request rate, errors, latency, and bytes between
  workloads, using Tempo service graphs or Istio, Beyla, and Caretta metrics.
  It shows the source and freshness, and states clearly when no flow source
  is connected.
• Existing infrastructure permissions and MFA now cover every new Kubernetes
  view, the live stream, and the agent tools. Approving actions requires
  admin or owner.

What's Changed

• feat(kubernetes): explorer v2 and cluster context for the AI SRE by @hoalongnatsu in #379
• fix(test): stabilize Go CI tests and speed up Kubernetes fixtures by @hoalongnatsu in #380
• perf(fakekube): speed up large Kubernetes fixture generation by @hoalongnatsu in #381

Full Changelog: v1.4.28...v1.4.29