Skip to content
Sendora Cloud
Create account
Platform

Health

Backend health surface: latency histogram, throughput, recent 5xx, alert rules, Slack webhook.

Features

  • **API usage by endpoint + key** — top callers, slow endpoints, error rates per project.
  • **Delivery health per channel** — push delivered / opened / failed, email bounced / opened / clicked, SMS delivered / failed.
  • **Latency + error budgets** — p50 / p95 / p99 across modules.
  • **AI anomaly alerts** via email / Slack with per-org webhook routing — picks up traffic dips, send-rate spikes, error-rate excursions.
  • **Audit log** with actor / resource-type / date-range filters + expandable metadata viewer.
  • **Top endpoints (24h) + webhook tester + SSRF-guarded outbound probe** — debug integrations safely.
  • **Promote-failed + workflow-failed alerts** surface broken Automation runs before users notice.

Common use cases

  • Engineering health dashboards for the Sendora-backed parts of your stack.
  • Ops alerting on delivery degradation (sudden push-failure spike from Apple, email bounce-rate breach).
  • Compliance audit trail — auditor-ready filterable log of every actor action.

Key concepts

Alert-rule fields
name, metric, operator (gt/lt/gte/lte), threshold, windowMinutes, channelIds — plus, on the event_volume metric, an optional eventTypes array. windowMinutes doubles as the re-fire cooldown. An unknown/misspelled field is rejected (422), never silently ignored.
eventTypes filter (per-signal dead-man's switch)
Only on event_volume. WITHOUT it, event_volume counts every event type in the project, so `lt 1` fires only when ALL traffic stops — app analytics keep the count warm and hide a single dead job. WITH eventTypes: ["backup.healthy"], it counts only that type and fires when that heartbeat stops. Sent on any other metric it is rejected (422). API-only today; the dashboard alert builder exposes it for event_volume.
uptime is passive, not synthetic
uptime, error_rate, latency_p99 and throughput are derived from the API traffic Sendora RECEIVES (org-wide — API logs carry no project dimension), not from an outbound probe. If your box goes dark it stops sending and uptime can't see the silence — use an event_volume + eventTypes dead-man's switch for that. There is no synthetic/external HTTP check today.

Observability

Read health snapshot

FREE

Real-time uptime + p99 latency + error rate. Polls the same source the operator dashboard renders.

curl "https://api.sendoracloud.com/api/v1/orgs/<ORG_ID>/observability/health" \
  -H "x-api-key: sk_prod_…"

Define an alert rule

FREE

Fire a Slack (or webhook) alert when a metric breaches a threshold, evaluated every minute. Create the channel under Notifications first and reference it by id. Full field set: name, metric, operator, threshold, windowMinutes, channelIds, and — on the event_volume metric — an optional eventTypes filter. ⚠ eventTypes is what makes a per-signal dead-man's switch: WITHOUT it, event_volume counts EVERY event type in the project, so `lt 1` only fires when ALL traffic stops (app analytics keep the count warm and mask a single dead job); WITH `eventTypes: ["backup.healthy"]` it counts only that type, so it fires when that one heartbeat stops. eventTypes is accepted only on event_volume — sent on any other metric it is rejected (422), and an unknown/misspelled field is rejected too (422), so a typo can't arm a rule that looks scoped but isn't. Event metrics (event_volume) can be scoped to one project; API metrics (error_rate, latency_p99, throughput, uptime) are org-wide, because API logs carry no project dimension — and uptime is derived from received traffic, not a synthetic outbound probe. The AI anomaly detector runs separately and can raise alerts on its own.

# Per-signal dead-man's switch: page when the "backup.healthy" heartbeat
# specifically stops, even while other events keep the project busy.
curl -X POST https://api.sendoracloud.com/api/v1/orgs/<ORG_ID>/observability/alert-rules \
  -H "x-api-key: sk_prod_…" \
  -d '{
    "name": "Backups stopped",
    "metric": "event_volume",
    "operator": "lt",
    "threshold": 1,
    "windowMinutes": 360,
    "eventTypes": ["backup.healthy"],
    "channelIds": ["{{CHANNEL_ID}}"]
  }'

Related

Product overview
All modules