Skip to main content

v1.13.0

Before You Upgrade​

All editions​

  • API tokens (breaking): Send your sk-nb-… API token directly as a Bearer token. POST /api/auth/token and POST /api/auth/revoke have been removed, so scripts and CLI versions that use the old exchange must be updated. Recreate any older token that stops working.
  • Relay MCP sessions: The relay now stores MCP sessions in Redis by default. Installs without Redis must set MCP_SESSION_STORE=memory and run a single relay pod.
  • Relay ingress: An ingress host with no paths now exposes only /register and /status. Custom paths are kept. If you relied on other public relay paths, list them explicitly.
  • Microsoft Teams: Teams messages now go through the bot instead of Microsoft Graph. On-prem installs must follow the Microsoft Teams setup guide and set MS_TEAMS_BOT_TENANT_ID and MS_TEAMS_ACTION_SIGNING_KEY.
  • Setup Center: The Setup Center is off by default. Set UI_ENABLE_SETUP_CENTER=true to turn it on.

Community edition​

  • Official database and broker images: PostgreSQL, RabbitMQ, and Redis now use official images instead of the Bitnami subcharts. Existing data is kept. Install fails with an error if your values use Bitnami options that the chart no longer supports.
  • No rollback to 1.12 after RabbitMQ 4 starts: This upgrade moves RabbitMQ from 3.13 to 4.2. To recover, upgrade to 1.13 again and run kubectl delete pod -n nudgebee rabbitmq-0. No data is lost.
  • helm upgrade --reuse-values is blocked: Upgrade with your values file (-f) or --reset-then-reuse-values.
  • First PostgreSQL start: PostgreSQL checks text indexes and refreshes collation versions before accepting connections. This takes seconds on a typical install.

What's Changed​

🔍 Troubleshoot​

  • feat: Pods killed by one full node are grouped into a single incident, and a later alert can lead its incident group
  • feat: Error-log spike investigations show which error patterns increased
  • feat: Service cards show how long a Service has had no endpoints
  • feat: Grafana Tempo traces appear in the Traces tab
  • feat: Triage score has been redesigned and moved to the first column
  • feat: Node-hostname alert titles now resolve to the node
  • fix: Alert charts and metric summaries show only the object that alerted
  • fix: Failed ECS deployment alerts name the cluster and service, and no longer merge into one alert
  • fix: Alerts close when their source reports them closed, and webhook alerts that fire again after closing are shown
  • fix: Alerts are grouped by timing only when they started together
  • fix: Noisy Neighbours no longer blames oversubscription when the node had memory to spare
  • fix: Log investigations use the requested time window and say when a read stopped at its line limit
  • fix: Cloud Logs and Cloud SQL metric query errors are now shown
  • fix: Utilization charts show network directions correctly (Community in this release)
  • fix: Parent span links stay within their trace (Community in this release)

📊 Dashboards​

  • feat: Dashboard variables, pie and histogram panels, drill-down data links, and dashboard generation with Nubi (Community in this release)
  • fix: Imported dashboards keep their panel providers and time settings

⚡ Optimization​

  • fix: Resource requests are never proposed or applied above their limit
  • fix: Container resource requests are counted once when several exporters publish them
  • fix: Auto-optimize pull request updates now land on the branch
  • fix: Elastic IP savings are shown on the address
  • fix: Cloud Posture rows show existing tickets and prevent duplicates
  • fix: Security & Compliance rows can all be clicked

🔄 Workflow​

  • feat: Allow automation network tasks to reach trusted private ranges with RUNBOOK_SERVER_TASK_ALLOWED_CIDRS
  • feat: PagerDuty custom fields, GitHub issue fields, and all ZenDuty keys are returned to automations
  • fix: Automations can now resize PVCs and report failures
  • fix: Script steps that can't run are caught, and the builder has what it needs to create steps that can
  • fix: Required task settings use their schema defaults
  • fix: Alerts raised by an automation close when the automation resolves them
  • fix: Deep run-history pages no longer repeat rows while runs are in progress

☸️ Kubernetes​

  • feat: Support SMTP relays without TLS or login
  • feat: Use one-click install commands from agent reminders
  • fix: Fresh installs register the bundled cluster without a restart and no longer open on Demo
  • fix: The vulnerability database refreshes instead of becoming unready after 7 days
  • fix: Built-in provider health is read per cluster, and a manual Disable stays in place
  • fix: Services recover from message broker restarts without a pod restart
  • fix: Services shut down cleanly during restarts

☁️ Cloud​

  • feat: EventBridge events for SNS, SQS, and load balancers
  • fix: Cloud alarms stay open until cleared, short and expression-based firings are recorded, and alarms link to the right resource
  • fix: Real-Time Events status reflects whether an account has received events
  • fix: Azure accounts stay connected when only cost data can't be read
  • fix: Resource age is measured from its cloud creation time
  • fix: GCP Recommender recommendations link to their target resources
  • fix: The nightly AWS account sync finishes within its time budget
  • fix: Cloud summary tiles count stopped instances and show 0 instead of -

🤖 AI​

  • feat: Nubi can work on another connected account, including Postgres, Redis, RabbitMQ, VM tools, and other integrations, and shows where each step ran
  • feat: Writes on another account are approved for that account
  • feat: Set per-account AI event analysis rules, including whether to reuse recent analysis and for how long
  • feat: Give an LLM fallback its own provider and credentials
  • feat: Fallback models support OAuth, extra headers, and other main-model options (Community in this release)
  • feat: Edit queued messages in place; message queuing is on by default
  • feat: Remediation is grounded in the investigation and states only recorded evidence; a remediation action can name the action it must follow
  • feat: Answer checks give guided feedback and send back root causes based only on an alert label
  • feat: Finished conversations show the approvals and questions from each turn
  • feat: Step details show the integration used, and you can copy a step's query and thought
  • feat: Code-fix PR owners are assigned when possible
  • feat: Memory reading, writing, and reranking can be turned off for a request (Enterprise)
  • fix: Secret-scanner false positives are down by 82% (Enterprise)
  • fix: Credentials in tool output are redacted under the egress filter settings
  • fix: Read-only commands are classified correctly, and database writes require approval
  • fix: Queued messages run after a failed turn
  • fix: Approvals resume with the original question and no longer replay completed batch calls
  • fix: Remediation success no longer resolves an event by itself
  • fix: Budget tracking now includes remediation and metric-filter LLM calls
  • fix: Faster Ask NudgeBee follow-up suggestions, and lower LLM cost through better prompt caching
  • fix: Routing rules that change nothing are rejected (Enterprise)

🔔 Notifications​

  • feat: See which rule suppressed a workflow notification
  • feat: Anomaly digests list resolved and returning anomalies
  • feat: Suppress alerts from Teams alert cards
  • fix: Spend anomaly endings send a resolved notification
  • fix: Security alert links open the right page for the account type
  • fix: Digest emails use your brand color for links and buttons

🎫 Tickets​

  • feat: ServiceNow tickets show an inline comment thread
  • fix: Ticket status sync works for every ticketing tool, and deleted tickets no longer stop it
  • fix: Temporary sync errors no longer disable a ticketing integration
  • fix: ServiceNow Resolved and Closed updates include resolution details, and Urgency is kept on ticket creation

🔌 Integrations​

  • feat: New Event Webhook for sending JSON alerts from any tool
  • feat: Utilization & Health for Splunk clusters and rightsizing data from Splunk Enterprise
  • feat: CubeAPM alert-rule warnings now appear on the Observability tab
  • fix: Splunk Enterprise trace search, index validation, and count queries are more reliable
  • fix: OpenObserve workload metrics now return correct results
  • fix: SigNoz metrics and alerts now work reliably
  • fix: Chronosphere span timestamps use UTC

🛡️ Admin & Platform​

  • feat: "View in audit log" opens the matching audit entry
  • feat: Knowledge Graph CSV and PNG downloads
  • fix: Integration secrets are no longer returned when you save an integration
  • fix: The shared password signs in only the configured admin
  • fix: API keys stay in the tenant where they were created when a user signs in or switches tenant
  • fix: Alert actions check account access
  • fix: Account feature flags are scoped to their tenant
  • fix: The hourly Knowledge Graph build no longer runs out of memory

🔒 Security (Community in this release)​

  • fix: Critical and high image vulnerabilities reduced across services and third-party images, with patched base images and dependencies
  • fix: Suspended users can no longer act through linked chat accounts
  • fix: Nubi no longer probes hosts you don't own
  • fix: The services-server action token is no longer logged