v1.4.32

Kubernetes Logs At Break

• Read logs alongside changes and crashes in a focused workload Timeline.
• Drag the log-window bracket, or use its keyboard controls, to compare the
  incident interval with per-minute log volume and errors.
• Select Logs before crash to inspect available previous-container output
  for the selected Pod and termination.
• Filter by Pod, container mode, severity, and search; review repeated message
  patterns and load more matching logs without changing the window.
• Read source labels and coverage tips before interpreting missing messages.
• Use Ask chat to carry workload and time-window context into an
  investigation, or Open live logs to follow current workload output.
  Load earlier output without restarting the live read.

Captured History And Linked Sources

• Automatically capture bounded Kubernetes container logs at detected breaks.
  Available captures survive Pod deletion and Versus restarts for up to 24 hours,
  subject to capture, storage, and retention limits.
• Link one optional configured source per cluster: Loki, Elasticsearch,
  CloudWatch Logs, Graylog, Splunk, or SigNoz.
• Map workload fields, inspect Test link results, and save with the required
  settings-write permission. A successful preview does not guarantee complete
  history; missing logs, capture limits, and source coverage can leave gaps.

Per-Cluster Agent Preview

• Save logs, metrics, and traces learning toggles and namespace scope in the
  cluster's Agent tab. These preview settings have no runtime learning effect.
• Keep Kubernetes container-log capture available without linking an external
  source. Source linking is optional, not a requirement to investigate a break.

Clear Cluster Preparation

• Start collection for configured clusters at boot, including multi-cluster
  installations, without waiting for someone to open the explorer.
• Follow the familiar four preparation steps with one progress bar and live
  resource counts. Progress continues while remaining resource types load,
  rather than switching to “unknown”.
• Prepare only before the cluster's first usable publication. Page reloads,
  Versus restarts and credential renewal do not restart that experience.
• Keep investigations on screen during background refreshes, using only the
  existing top-bar sync indicator. Real failures, permission gaps and stale
  snapshots remain visible; snapshot age is not presented as startup duration.

Live Logs And Resource Details

• Follow Pod, Deployment and StatefulSet logs across reconnects without oversized
  resume requests interrupting the read.
• Open Logs directly from resource lists. Select a container, switch Current or
  Previous, filter output and load earlier lines without restarting the live read.
• Follow automatically while at the bottom, with an animated Live indicator and
  Jump to latest only when needed. Quiet, waiting and ended reads have clear states.
• Keep Overview live-log cards full width, including when no lines are available.
• Open resource details over the explorer without squeezing the list underneath.
  Hide empty crash summaries and pagers that have no additional page.
• Preserve available investigation history across OSS-to-Enterprise upgrades.

Kubernetes Issue Alerts

• Enable alerts per cluster from the Issues tab; alerts are off by default.
  Choose Critical or Warning and send notifications through configured alert
  channels using the existing incident flow.
• Open the issue's Timeline or investigate it with the DevOps Agent directly from
  a notification. Slack provides action buttons; other channels provide links.
• Enterprise alert-fatigue handling remains opt-in, not enabled by default.

Kubernetes Reader Permissions

• Include read-only serviceaccounts and secrets access in the shipped reader
  role. Secret reads use metadata-only requests; Secret values are not projected.
• Review the additional permissions before applying the updated role. Operators
  who do not want Secret metadata visible can omit that grant from their own role.
• Distinguish an unconfigured connector from unavailable credentials, and report
  unreadable resource types without blocking the rest of a readable cluster.

Enterprise

• Apply existing role, cluster-access, and MFA requirements to break-log reads,
  Agent settings, and linked-source tests.
• Allow owners and admins to save settings and use Test link; responders
  with infrastructure access can read permitted evidence and settings only.
• Audit settings changes and access denials.
• Preserve organization and allowed-cluster boundaries during boot collection,
  credential failures, first-publication recovery, and live-log resumes.

Logs at break, captured history, linked sources, and Agent preview settings are
shared OSS features, not paid-only capabilities. No automated remediation is
introduced.

What's Changed

• feat(kubernetes): add logs at break and per-cluster agent settings by @hoalongnatsu in #387
• feat(kubernetes): protect issue alerts and preserve SSO investigation… by @hoalongnatsu in #388
• feat(kubernetes): refresh resource details and stream workload logs by @hoalongnatsu in #389
• ci: fix Dockerfile for Kubernetes Harness by @hoalongnatsu in #390
• ci: fix Dockerfile for Kubernetes Harness by @hoalongnatsu in #391
• feat(kubernetes): simplify investigations and scale cluster views by @hoalongnatsu in #392
• feat(kubernetes): keep cluster preparation clear and live logs continuous by @hoalongnatsu in #393

Full Changelog: v1.4.31...v1.4.32