Cloud Connection Notes

- API Keys: Permission scopes now display clear, descriptive names, such as "Create Budgets" or "Upload Container Metrics", instead of internal identifiers, making it easier to see exactly what access each key grants.

• API Keys: Kubernetes permissions are now split into a send-only Agent scope and a separate Clusters scope, so administrators can grant metrics ingestion access without exposing cluster management permissions.
• Explorer: Monthly axis labels across cost charts now read as Jun '26 instead of Jun 26, so the two-digit number is clearly a year rather than a day.
• Budgets: Budget plans can now span up to 120 months, up from 24, whether entered through the in-app month picker or a CSV upload, so multi-year committed spend no longer needs to be split into shorter segments.
• Explorer: Hovering the normalized trend line now shows the same detailed tooltip as the bars, with cost broken out by category alongside the date and total, giving consistent detail across raw and normalized views.
• Explorer: The monthly normalization control is now a labeled two-option selector with an info tooltip explaining the 30-day scaling, making the setting easier to find and understand.

- Explorer: Chart zoom now works at weekly and monthly granularity, not just daily and hourly, and the current range caption always matches what's actually on screen.

• Explorer: Kubernetes usage data no longer omits pods that started partway through an hour, so container cost and utilization figures are complete for clusters with frequent pod churn, such as autoscaled node groups and batch jobs.
• Explorer: Pod resource request and limit values no longer get cleared when a pod isn't sampled in a given update cycle, keeping Kubernetes cost allocation figures stable and accurate.
• Optimize: Status sync between CloudZero and Jira tickets linked to recommendations now works reliably in both directions, and any future mismatch surfaces immediately instead of failing silently.
• Anomalies: Tickets created from the ticket panel or an anomaly's details flyout now sync status correctly with Jira instead of silently failing to update.
• Optimize: The recommendations list no longer fails to load when a cached results page is unusually large, so recommendations return successfully every time.
• Dimension Studio: Hidden dimensions can now be selected as sources, conditions, and child references when building dimensions, matching what was already possible by editing the underlying definition directly.
• Explorer: On first load, the filter bar and filter panel now show both levels of a saved default view's group-by, matching the cost data already displayed.
• Budgets: Sorting the Spend, Budget, and Forecast columns now orders amounts numerically, so the largest and smallest values sort where you expect instead of by digit order.
• Budgets: Alert threshold chips in the budget drawer now show the dollar amount for a single month, matching what the alert actually evaluates against, instead of the full budget period total.
• Explorer: Limited-access users no longer see views they don't have permission to open, removing the error they previously hit when selecting one from the dropdown.
• AI Signals: Activity and context details on the Livestream page now persist after a reload and display exactly as reported, instead of disappearing until a new live event arrives.
• Dimension Studio: Selecting a dimension or tag that shares a display name with another dimension now correctly marks it as selected, instead of silently selecting the wrong one.
• CZ Labs: Organization admins with permission to manage organization settings can now control org-wide Labs features without also needing super-user access, matching the permissions they were already granted.
• Explorer: A view built on a dimension that has since been deleted now recovers automatically and reloads the default view instead of showing a dead-end error.
• Access Controls: Data access filters that reference a hidden dimension now show the dimension's name correctly instead of a blank placeholder in the filter picker.
• Optimize: Sorting the recommendations table by a grouped column no longer produces an error or breaks group expansion, and any sort setting that was already broken is corrected automatically on your next visit.
• AI Signals: Telemetry sent during brief service interruptions is no longer silently dropped, so token and cost data doesn't go missing during transient outages.
• AI Signals: Team attribution for LiteLLM deployments that propagate identity through baggage headers is now correct, so cost rolls up to the right team instead of showing unattributed.
• AI Signals: Model, token limit, and temperature details for LiteLLM requests now populate correctly, giving accurate visibility into AI usage that previously appeared blank.
• AI Signals: AI cost captured from the desktop collector for Codex and other OpenAI-based tools now attributes correctly to the active organization, even if you switch organizations mid-request, so spend lands on the right account.
• AI Signals: User attribution in telemetry data no longer falls into an unattributed bucket when a name or email is missing, since additional identity data is now used as a fallback.
• AI Signals: Telemetry processing for high-volume accounts no longer stalls or times out under peak load, so usage data continues to arrive without delay.
• Cloud Connections: An Oracle Cloud month with no available cost reports now completes successfully with a warning, instead of failing the entire connection sync.
• Explorer: Shared and saved chart links no longer carry an inactive monthly-normalization setting when a date range switches the view away from monthly granularity, keeping links accurate.
• Explorer: Dimension values containing periods, colons, or brackets in their names no longer merge together in cost charts, so bar totals and drill-downs stay accurate.