Last9
Book demo
Nov 10th, 2026

Fullscreen dashboards and Metrics Query Library

Improvements Fixes

TL;DR

Dashboards get a fullscreen mode and a simpler panel editor, Metrics Explorer gets a Query Library, and the MCP server adds alert coverage and previews.

Fullscreen dashboards and Metrics Query Library

Improvements

  • AI Assistant: Tool activity in a chat shows Approved, Denied, Pending approval, or Expired next to a tool that needed approval. The chip stays in restored history, and a pending approval turns to Expired at its deadline.
  • Dashboards
    • Added a fullscreen mode. Open it from the new More menu or press F, and leave it with Exit, Esc, F, or the browser Back button. Entering fullscreen turns on auto-refresh at 1 minute if it was off, and a Live chip pauses and resumes it.
    • Simplified the panel editor. Its settings now sit in 5 collapsible sections: Chart Settings, Legend Settings, Query Options, Links, and Panel Settings. A section opens by default when it holds a value. Lookback and Time Shift are checked as you type, so a value such as 24 hours shows an error and blocks Save.
  • Metrics: Added a Query Library to Metrics Explorer, next to Run Query, with Recent and Saved queries, All, Team, and Personal tabs, and search. Save a query with Save. A saved query restores its query, mode, and cluster; a recent query opens in code mode. Editors and admins can save, edit, and delete queries; viewers cannot.
  • Alerting: The header of the alert rule editor reads Edit Alert Rule followed by the saved rule’s name.
  • Settings > Audit Trail: Creating, updating, or deleting a dashboard writes an audit entry with the dashboard ID and the user.
  • Agent Observability: The trace panel shows a cost the trace reports (labelled Reported) ahead of an estimate, and the estimate accounts for cached tokens. When cache pricing is unknown, cost shows as a range. Time to First Token appears only when the span reports it, apart from Response Time.
  • MCP
    • Added the get_alert_coverage tool, which lists what you are not alerting on for a service, environment, alert group, or the whole organization. It groups gaps as uncovered signals, rules that are all disabled, rules with no notification channel, and groups muted too long, and suggests the fix tool for each.
    • Added preview_dashboard, which checks an unsaved dashboard or a patch and returns resolved queries and per-panel results without saving it, and dashboard authoring prompts that guide layout and queries.
    • Added the create_kpi and get_entity_kpis tools to create and list the KPIs of an alert group, and an author-alert prompt that walks an agent through writing an alert rule as a draft.
    • get_alert_config shows five recent one-minute evaluations for each rule.
    • Breaking: get_trace_waterfall returns the trace service’s waterfall with credentials, query literals, URLs, emails, IPs, and user identifiers redacted. See the MCP server release notes for v0.22.0 and v0.23.0.

Fixes

  • AI Assistant: When the connection to a reply dropped, the chat showed “An unexpected error occurred. Please try again.” The chat now reloads the turn instead.
  • Discover Kubernetes: The Logs tab of Node, Deployment, StatefulSet, and Pod details, and the detail metrics, were empty for EKS clusters labelled by ARN rather than by short name. Logs tabs now scope to namespace, workload, and pod name, so same-named workloads in the same namespace of another cluster in that region can appear together.
  • Control Plane
    • Remapping: Saved remap keys that were missing from the current label list were hidden in the Map Logs and Map Traces editors and dropped when you saved the rule.
    • Sensitive Data: The Phone Number rule redacted space-separated numeric fields in logs, such as 200 116 2459 in an HAProxy line. It now matches North American numbers only when they use - or ., and adds +60 and +66 numbers.