Changelog

Bug Fixes

• Dashboard Bar and Time Series panels now render correctly when groupBy data is missing, fixing crashes seen in panels created through LLM/MCP workflows.
• Unsupported ClickHouse and PromQL query types have been disabled for Anomaly Detection alerts.
• Several edge cases in the Alerts and Planned Downtime frontend flows, such as stale loading states, have been resolved.
• Operations on deleted service accounts now return a proper 404 Not Found HTTP response instead of 501 Not Implemented.

Maintenance

• Frontend internals were streamlined with remaining xstate usage replaced with plain React state and local dispatch helpers, light-mode specific styles cleaned up, and unused code deleted.
• The Trace Waterfall API was moved into the Trace Detail module, with improvements to handling of large traces, caching, and error resilience.
• Alerts and Planned Downtime frontend flows were migrated to generated clients, reducing custom API adapter code and improving consistency in error handling.
• Integration and E2E test infrastructure was consolidated under a shared pytest project with common fixtures, local backend-based E2E runs, and updated CI workflows.
• oxlint and oxfmt were added as frontend linter and formatter respectively, speeding up the developer feedback loop.
• The OpenTelemetry Collector was bumped to stable v0.144.3.

Bug Fixes

• Alert creation from dashboard panels now works correctly for non-builder query types.
• Styling regressions introduced during the AWS integration refactor have been fixed, including semantic token issues in affected UI surfaces.
• The HAVING filter in the Query Builder V2 now emits a valid NOT IN operator, fixing cases where selecting this operator produced malformed queries.

Maintenance

• Alerting and rule-management internals have been consolidated further, with AlertManager, Ruler, and planned maintenance handlers moved into signozapiserver and tighter OpenAPI validation added for rules and downtimes.
• The AlertManager backend has been extended to support templating and custom notifier flows, laying the groundwork for future alert customization.
• The frontend design token system has been moved toward semantic tokens, improving consistency across the UI and reducing styling drift.
• Frontend asset management has been hardened, with assets migrated to src/assets and new lint rules added to prevent unsupported asset references.
• Infrastructure Monitoring internals for Hosts and Kubernetes have been consolidated into shared list and detail components, improving consistency and making these views easier to maintain.
• Unused or deprecated internals have been cleaned up, including unused trace cache settings and other obsolete code paths.

Bug Fixes

• Routing policy expressions that reference undefined variables no longer fail validation. Variables can now be left undefined without blocking the expression from being saved.
• Queries using hasAll or hasAny filter operators now pass the value argument correctly, fixing cases where these filters returned incorrect results.
• Deleted service accounts now return a proper unauthenticated error code instead of a generic failure when their API key is used.

Maintenance

• The OpenTelemetry Collector has been upgraded to v0.144.3-rc.4.
• Cloud integration modules have been refactored with updated query parameter support and removal of deprecated underscore-prefixed attribute dashboards.
• SQLite connection max lifetime is now configurable, improving connection management under sustained load.
• Frontend onboarding data sources have been migrated from JSON to TypeScript, and asset imports now use the @ alias for cleaner path resolution.
• ClickHouse queries sent from dashboard panels are now obfuscated in telemetry to avoid leaking sensitive query content.

Bug Fixes

• The host list in Infrastructure Monitoring now correctly shows refresh status, and refresh interval queries no longer overlap with each other.
• Dormant metrics in the query range API now return a warning instead of an error, preventing unnecessary failures when querying metrics that haven't reported recently.
• Alert rules history no longer crashes when relativeTime is empty.
• Dashboard variables with special characters now fall back to the raw parameter if decodeURIComponent fails
• The having expression in the query builder now validates correctly before execution.
• Empty not() filter expressions are now handled gracefully instead of producing query errors.

Maintenance

• Cmd+click (or Ctrl+click) now works on all clickable items across the app, opening them in a new tab as expected.
• Deprecated v3/v4 rule format support and the v1 Metrics Explorer code have been removed.
• JSON Path indexing is now enabled for JSON log bodies, improving query performance on structured log data.

Bug Fixes

• Filtering traces by trace ID in the list view no longer returns duplicate spans or results from outside the selected time window. Previously, the trace_summary timerange override was applied incorrectly across query buckets, causing repeated and out-of-range results.
• Logs pipeline filters now correctly resolve manually entered attribute names that are not present in the filter suggestion dropdown. Previously, entering a field like ddtags Exists by hand would misclassify it as a column rather than an attribute, causing downstream processors such as regex parsers to fail silently.
• The normalize pipeline now runs first in the processing chain, ensuring consistent field formatting before other pipelines execute.
• PromQL queries that exceed the execution deadline now return a 504 Gateway Timeout instead of a misleading 400 Bad Request. The PromQL engine timeout is also now configurable via the prometheus.timeout setting (default: 2m).
• Charts on the Volume Details page in Infrastructure Monitoring now display proper legend labels instead of "undefined."
• Domain rows in External APIs now show a pointer cursor on hover, indicating they are clickable.
• Time series charts with isolated data points surrounded by null values on both sides render those points correctly instead of dropping them.
• Deleting an alert rule or planned maintenance entry that has associated dependencies no longer fails with a foreign key constraint error.
• Agent configurations no longer get stuck in the "In Progress" state indefinitely.
• Users with the Editor role can now access gateway APIs as expected.
• Checkout and billing errors now surface the actual downstream error message in the UI instead of a generic failure.

Maintenance

Patch release with internal fixes. No user-facing changes.

Bug Fixes

• The custom domain settings card (Cloud) no longer shows a confusing plan name label. The card now displays the workspace name and a copiable license key in the footer, and no longer renders empty space when no custom domain is configured.
• Querying a histogram metric that has not received data within the selected time range now returns a "not found" error with diagnostic details (including when the metric was last seen) instead of a generic "something went wrong" message.
• Switching a dashboard panel type from time series to table now loads the table data correctly. The right-side settings panel in edit mode is also scrollable again.
• Dashboard variables now retain their selected values when navigating back from the widget editor and when toggling between view and edit modes.
• Creating a new Slack notification channel no longer fails with a "no global Slack App URL set" error. This affected users who signed up before v0.115.0 and did not have an existing Slack channel configured.
• The default color for log severity_text entries (when no severity key is present) now uses a theme-appropriate background color, resolving poor contrast in dark mode.

Bug Fixes

• Quick filters now correctly highlight active selections when query keys include prefixes like resource.service.name.
• Opening the widget editor on dashboards imported from partially formed or LLM-generated JSON no longer crashes the threshold selector.
• Dashboard panel creation no longer fails when the underlying API response is unavailable; all panel types now handle missing data gracefully.
• Updating a dashboard variable no longer throws a crash caused by an attempt to modify read-only state.
• The filter dropdown on the trace waterfall view no longer overlaps the search bar when the browser is zoomed in.
• API calls made after a token refresh now correctly retry using the updated token across all pages.
• A planned downtime notification with a malformed description no longer crashes the Planned Downtime page.
• The app layout now uses the correct timer cleanup function, preventing potential issues in non-standard JavaScript environments.
• The selected y-axis unit in Logs, Meter, and Traces explorers is now preserved across reloads, back/forward navigation, and shared links.

Maintenance

• Dashboard panels with auto-refresh enabled no longer accumulate stale query cache entries in memory over time, preventing browser tab crashes on long-running dashboards.

Bug Fixes

• Editing the time range by typing directly into the custom time picker now applies your changes when closing the popover, instead of silently discarding them.
• PromQL-based dashboard panels no longer display internal attributes like __name__ and fingerprint in query results.
• Reverted the quick filters migration introduced in v0.114.0 to resolve issues it caused.

Maintenance

• Tooltip pinning has been reworked to support clicking on a data point across time series, bar, and histogram charts to keep the tooltip visible while inspecting values.

Bug Fixes

Dashboards

• Charts no longer leak memory due to canvas elements persisting across component re-mounts.
• Legend color picker now responds correctly under rapid interactions, and chart tooltips no longer leak memory on re-render.
• Dynamic variables that depend on textbox variables now load correctly even when the textbox value is empty.

Metrics Explorer

• Multiple queries using the same metric with no unit set are now correctly rendered as separate charts instead of being forced into one.
• Metric selection in the query builder now properly responds to blur and keyboard shortcuts (Cmd+Enter), fixing cases where aggregation options would not update after selecting a metric.
• Cmd+Enter now works in the summary tab, and key values can be expanded in full instead of loading five at a time.
• The most recently seen unit is now used instead of an arbitrary one, fixing incorrect unit labels on histogram count metrics.
• Errors are now surfaced to the user instead of failing silently.

Other

• Spatial aggregation queries are now validated before execution, preventing silently incorrect results from invalid configurations.

Maintenance

• SigNoz's frontend build system has been migrated from Webpack to Vite, reducing build times from 83 seconds to 10 seconds.
• Frontend bundle size has been further reduced by lazy-loading the syntax highlighter, moving antd-table-saveas-excel out of the main bundle, switching the Service Map to a lighter dependency, and caching the Intl number formatter.