[MrYuion]

Backoffice has sent nothing to Sentry since 2020. That year a cleanup removed Sentry.init and the ErrorHandler provider. Only an unused @sentry/browser dependency was left.

Changes

• src/app/common/sentry.ts loads @sentry/angular with a dynamic import after the user signs in, so the initial bundle stays small. SentryReportingErrorHandler still logs to the console. It queues up to 20 errors until Sentry starts.
• The DSN comes from the authority config.backoffice.sentry_dsn. When that is not set, the PlaceOS default DSN is used. '' or false turns Sentry off. Sentry is also off on localhost and in mock mode.
• Sentry gets errors, 10% of traces (100% on dev builds) and replays of sessions that have an error. Replays use Sentry's default masking of all text and media. Only the user ID and a domain tag are sent. The old code also sent the email address.
• Sentry's own window error listeners are turned off. Angular already sends those errors to the error handler, so each error would be sent twice.
• Production builds make hidden source maps. CI injects debug IDs and uploads the maps, then deletes them before deploy. Debug IDs change the script files, so config/ngsw-rehash.js updates the hashes in ngsw.json. Without this, the service worker rejects those files.

Setup

Add the SENTRY_AUTH_TOKEN, SENTRY_ORG and SENTRY_PROJECT repository secrets. Until all three are set, the upload step is skipped. See docs/sentry.md for where to find each value.

Testing

• Unit tests: 1081 pass, including the new sentry.spec.ts. Lint and type checks pass.
• In a local PlaceOS stack, with the localhost check removed for the test: the session, spans, one error event and the replay all got HTTP 200. The event had the correct release and the user ID only.
• After sentry-cli sourcemaps inject, ngsw-rehash.js gives the same hashes as Angular's ngsw-config. On a fresh build it changes nothing.
• The CI upload has not run yet, because the secrets are not set.

The Sentry chunk is about 120 kB gzip and loads after sign-in. Most of it is replay.

Reviewed with sol-loop: 2 rounds with gpt-6.1-sol on xhigh.

Made by Claude Opus 5.5 (claude-opus-5-5) in Claude Code, running in T3 Code.

🤖 Generated with Claude Code

[vercel]

Deployment failed for project backoffice with the following error:

Resource is limited - try again in 24 hours (more than 100, code: "api-deployments-free-per-day").

Learn More: https://vercel.com/placeos?upgradeToPro=build-rate-limit

[greptile-apps]

Confidence Score: 5/5

[High impact] The reviewed update appears safe to merge; it fixes the hidden startup warning without introducing a new issue.

Summary

Restores Sentry error reporting after sign-in, with domain settings, queued errors, traces, and masked replays.

• Backoffice sends app errors and selected session data to Sentry.
• Production errors can show their original code in Sentry.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Angular error handler] --> B[Log to console]
    A --> C[Queue up to 20 errors]
    D[User signs in] --> E{Sentry enabled?}
    E -->|No| F[Discard queued errors]
    E -->|Yes| G[Load and start Sentry]
    G --> H[Send queued and future errors]
    G -->|Startup fails| I[Always show console warning]

Loading

_{Reviews (3) · Last reviewed commit: "fix: always log when Sentry fails to sta..." · Reviewed by Greptile}