Seed reports errors and crashes to Sentry, a service that collects errors from running programs. This page is for a contributor who needs to find an error or to change what is sent. It does not cover the settings that a Sentry administrator changes in the Sentry web UI.
Three programs report: the web app, the notification service and the desktop app. Each one starts Sentry's client library when a DSN is set. A DSN is the address of a Sentry project. The build workflows upload source maps and the debug symbols of the daemon, so that stack traces are readable.
Projects
All projects are in the Sentry organisation with the slug mintter. A slug is the short name in a Sentry URL. The code names that slug and the project slugs, so they must match.
Project slug | What sends events | Enabled by |
|---|---|---|
| the web app, both the server and the browser, through |
|
| the notification service, through |
|
| the desktop app: main process, renderers, preload and native crashes |
|
The Go daemon has no project of its own. Native crashes of the desktop app and its daemon go to seed-electron, which is where the daemon's debug symbols are uploaded.
Without a DSN the Sentry library does nothing. So a local build reports nothing unless you set one.
Environments and releases
Reading an error
Open the issue in its project and pick the environment and release.
Click View Source on a stack frame. You should see the original TypeScript. If you see minified JavaScript, the release name that the program sends does not match the release name of the uploaded source maps. Both come from the same variable: SITE_SENTRY_RELEASE for the web app, which is the commit hash, and VITE_VERSION for the desktop app, which is the git tag. Check in the logs of the CI run that built it that both are set.
For a crash of the daemon in seed-electron, the Go stack should show function names and file paths. If it shows raw addresses, look in Settings โ Projects โ seed-electron โ Debug Files for a seed-daemon-<arch> entry with the build id of that release.
@sentry/vite-plugin and sentry-cli upload the source maps and the releases during the CI build. Both need the repository secret SENTRY_AUTH_TOKEN. When that token expires, events still arrive, but the upload of source maps stops, so traces show minified code.
Where each setting lives in code
Sentry setting | Code location |
|---|---|
Project slug |
|
Organisation slug | The same files ( |
Sentry start (web) |
|
Sentry start (notify) |
|
Sentry start (desktop) |
|
Release name (web, notify) |
|
Release name (desktop) | the |
Environment |
|
Sampling rates | in the Sentry start files above. |
What gets stripped after upload |
|
Secrets passed to builds |
|
Daemon symbol upload |
|
See also
Releasing, which produces the tagged builds.
The web image sets SITE_SENTRY_ENVIRONMENT=production in its Dockerfile. The server and the browser read that variable, then SENTRY_ENVIRONMENT.
The notification service reads NOTIFY_SENTRY_ENVIRONMENT, then SENTRY_ENVIRONMENT, then NODE_ENV, and falls back to production.
The main process of the desktop app reads SENTRY_ENVIRONMENT. Without it, a packaged app reports as production and an app run from source reports as development. The renderer reads VITE_SENTRY_ENVIRONMENT, then the Vite mode.
Logs and endpoints and health, for the logs and metrics of a running app environment.
Working on Seed, the list of contributor guides.
Do you like what you are reading? Subscribe to receive updates.
Unsubscribe anytime