What getsentry/sentry shipped

This week: agent chat messages are gated on Seer credits, single-run Autofix results and PR links now show in the Explorer chat, cron guides use the current monitor run command, and SCIM team and repo link endpoints validate input properly.

This week's changes: agent chat messages for token-billed organizations now return a 402 before reaching the backend when credits are insufficient, while dashboard generation runs are not gated. Autofix cards can be built from a single run, with Root Cause, Code Changes and Pull Requests from the run, and Start Analysis, Re-run and Open Autofix open single runs in the docked Explorer chat, which also shows single-run PR links. Cron onboarding guides now use the current monitor run command, and the SCIM team and project repository link endpoints now validate their request bodies instead of failing or drifting from the documented schema.

This week: endpoints can require a user, service accounts get their own principal on requests, the inbox issue preview shows a full header with activity and telemetry tabs, and a large index on ArtifactBundleIndex was dropped.

This week in getsentry/sentry: endpoints can now declare requires_user and return 403 for callers with no user, and a service account's request.user is a ServiceAccountPrincipal instead of AnonymousUser. The inbox issue preview now shows a full header with title, group actions, counts and activity and telemetry tabs, and inbox items have selection checkboxes with no bulk actions yet. The (url, artifact_bundle_id) index on ArtifactBundleIndex was dropped. Seer automation configs support create, read and update, the autofix GET response exposes failure_reason, metric monitors can connect to breached-metric automations, and uptime stats reject resolutions EAP can't serve with a 400.

This week: shared issue links stop reloading on 401s, metric alert emails link straight to issue open periods, inbox stack traces collapse again with Seer content, admin gets batch promo code creation, and native symbolication now covers flamegraph attachments behind a flag.

This week in the repository: shared issue links for native crashes no longer reload forever for logged-out viewers, and metric-alert emails now link directly to the issue's open period instead of the legacy incident lookup. In the inbox, stack traces collapse again when Seer content is present. Seer repo selection changes for projects on token-based billing, admin gains batch promo code creation, and native symbolication now covers version-1 flamegraph attachments behind the flamegraph-attachments flag. On the database side, a post-deployment migration builds new indexes on sentry_grouprelease ahead of the column swap, and two unused private endpoints were removed after 90 days without traffic.

This week: formulas now work in timeseries, literal-only equations are fixed, spans tables show the real timeout message, and project names default to the repository name in onboarding.

This week in Sentry: formulas are now available in timeseries views, and equations containing only literal values evaluate correctly. Spans tables show the backend's query timeout message instead of a generic error. SCM onboarding and create-project flows now default the project name to the repository name. Dropped-event reasons gained human-readable descriptions, and the grouprelease table got a shadow column for widening the release id.

This week: legacy login/registration/MFA views replaced by React auth, legacy option keys retired, dashboard resize debounce swapped to useDebouncedCallback, id swap lock order fixed, and double-bracketed URL values no longer crash issue detectors.

This week in the project: the legacy password login, registration and MFA views were replaced with React authentication, and legacy option command, channels and reporting keys were retired. On dashboards, the resize handler now uses useDebouncedCallback and saved span widgets use the filter-first _if syntax, while breakdown labels show the model icon plus all remaining group by values. Fixes include locking child tables before sentry_release in the id swap migration, handling double-bracketed URL values in issue detectors, returning zero performance score weights when no web vital has data, and keeping rule-only notifications out of the digest.

This week: autofix metrics with full sampling for better visibility, data exports now reject files outside retention, mobile span attributes propagate correctly, and per-monitor check-in rate limits for crons.

Shipping this week: Autofix now samples trigger metrics at full rate to reduce noise on low-volume steps. Data exports correctly reject requests outside event retention. Mobile device attributes propagate through child spans. Crons enforce per-monitor relay quotas. Notification rules from workflows assign correct environments. Investigations get smart default titles based on type.

This week: fixed billing display for usage-based pricing, patched a security issue with organization validation, and improved PR iteration reliability with Seer retry logic.

This week's Sentry updates brought critical fixes across three areas: billing now correctly displays usage-based line items in dollars, a security issue with organization selection in get_identity_or_404 was resolved, and PR iteration work is more resilient with improved retry logic for Seer lookups. We also continued consolidating legacy alert code and refactored evaluation artifacts for better type safety.

This week: suspect commit routing now works per project, not org-wide, fixing detection for 136K projects. Autofix improves with Seer RCA by default and fixability labels. Inbox combines personal and team filters.

Sentry shipped significant improvements this week: suspect commit detection now routes per project (fixing 136K projects across 50K organizations), autofix advances with Seer RCA by default and automatic fixability labeling, alert builder and detail pages gain semantic context awareness for better AI assistance, and logs now link span_id attributes to waterfalls. These changes improve error triage, resolution, and observability workflows.