Launch-critical monitoring wiring — alerts go to tiaBaby_Bot via Telegram.
- src/lib/alert.ts: sendAlert(level, title, detail?, {fields, silent}) — HTML
formatted, IST timestamped, best-effort (never throws). Env: TELEGRAM_BOT_TOKEN,
TELEGRAM_CHAT_ID
- GET /api/healthz: public, no-auth liveness probe (200 ok / 503 down) for
Uptime Kuma + Dokploy healthcheck. No sensitive detail
- cron/backup: alert on failure (fatal), warn if dump < 1KB (empty), silent
success confirmation with file + size
- cron/monitor: error-spike rising-edge detection (last 1h > 5 and > 2x prior
hour — stateless, no re-alert on flat rate), DB/migrations/integration checks.
?test=1 sends a Telegram test ping
- cron/visitor-summary: polls Umami REST API (login -> stats/metrics/active),
posts visitor digest to Telegram. ?hours=N window (default 24)
- CLAUDE.md: new env vars + Monitoring & Alerting section
Health up/down flip detection is delegated to Uptime Kuma (pings /api/healthz);
this code covers what Kuma can't see from outside.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
