Skip to main content

​
Context

Nothing in Stackra emits a metric, a trace or a structured log through a shared interface yet. Error reporting that captures URLs, user data or screenshots turns an incident tool into a privacy incident. The vendor choice is an implementation detail that should be swappable.

​
Decision

  1. The vendor sits behind configuration and one reporting function. Application code calls a reportError-style helper with value-free tags; it never imports a vendor SDK directly. Swapping vendors changes one module.
  2. Crash reporting is Sentry, one project per app, environment as a tag. The DSN is public configuration, not a secret; upload tokens are CI or build secrets (0009).
  3. No personal data. sendDefaultPii is off; no user object, no screenshots, no view hierarchy, no session replay. Breadcrumbs keep navigation only.
  4. No URL-bearing tracing. HTTP spans carry full URLs, which contain ids and query strings. Performance tracing stays off until a redaction step exists.
  5. Uptime and deploy events go to a hosted monitor; deploy events are log lines, not heartbeats.
  6. Open question, to be decided when the first service runs in production: the logging, metrics and health interface. This ADR stays Proposed until then.

​
Consequences

  • Reports are less detailed on purpose.
  • A vendor change is cheap.

​
Enforcement

Planned: a lint rule banning direct imports of the vendor SDK outside the reporting module; a test that the init options keep sendDefaultPii: false and tracing off. Nothing is implemented yet.

​
Supersedes and sources

Supersedes: none. Sources: academorix docs/adr/0008-sentry-mobile-errors; platform docs/adr/0006-observability-better-stack-and-sentry; basalt adr/0050-observability (the open question).