Image variants were written to the ephemeral release directory
and wiped on every deploy, causing 500 errors with 50s timeouts
as browsers waited for images that could never be served.
- Point image_cache_dir at /data/image_cache in prod
- Add Plug.Static to serve from the persistent volume
- Exclude /image_cache/ from broken URL tracking
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Move LiveDashboard to /admin/dashboard behind session auth (all envs)
- Add ErrorTracker at /admin/errors for auto-captured exceptions
- Add Oban job and LiveView metrics to telemetry module
- Add logger_json for structured JSON logs in production
- Enable os_mon for CPU/disk/memory in LiveDashboard OS Data tab
- Extend logger metadata with oban_worker and oban_queue fields
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
