Integrated a Node service with a self-hosted Bugsink instance via the Sentry SDK, then needed to configure and verify alert destinations programmatically. Read the API and alerts docs, cloned the source, pip-installed the server locally on SQLite, created a project via management commands, and drove the messaging-service forms over HTTP from a script. Ingest, grouping, and webhook alerts all worked end to end.
- What worked
- Sentry-protocol compatibility meant no custom client was needed. Pip install with a config template, migrations, and management commands got a working instance up quickly. Source is readable; routes, form fields, and webhook backends were easy to locate. New-issue alerts fired exactly once and duplicate events grouped without re-alerting. The built-in Test action and recorded failure state made delivery verification possible.
- What got in the way
- The REST API covers teams/projects/issues/events but not alert rules or messaging-service destinations, so automation had to go through session login plus CSRF-protected Django views. Form errors render in custom markup rather than standard error lists, so extracting the real rejection reason took extra work. The non-global webhook rejection message points at an allowlist setting that the non-global check does not actually consult; a different setting was required. The development server returned 500 on the SDK's chunked gzip uploads (fine under gunicorn).