system-prompts-and-models-of-ai-tools/dealix/docs/ON_CALL.md

Dealix On-Call / Incident Contact

Effective: 2026-04-23 Review: after first paid deal (G5) — then set up a real rotation if >1 engineer

---

Primary contact

• Name: Sami Assiri
• Email: sami.assiri11@gmail.com
• Role: Founder / sole maintainer
• Timezone: Asia/Riyadh (UTC+3)
• Availability: 24/7 during Primitive Launch phase (first 14 days)

Until there's a second engineer or a paid on-call service, Sami is the escalation path. This is a conscious trade-off — it's cheaper than a pager rotation until we have revenue.

Secondary (automated)

• Agent: Perplexity agent with GitHub connector (api_credentials=["github"])
  • Can: read repo state, open PRs, merge PRs with admin override, read GitHub-side CI logs
  • Cannot: SSH to prod, read .env values, run k6/drills on the server
  • Contact method: via existing Perplexity chat thread

Escalation decision tree

incident
  ├─ can I see it in /health/deep ?
  │   ├─ yes + 5xx → page Sami immediately
  │   └─ no        → check /admin/dlq/stats, /admin/approvals/stats
  ├─ is revenue path affected (/checkout, /webhooks/moyasar) ?
  │   ├─ yes       → page Sami + open GitHub issue with label `P0`
  │   └─ no        → file GitHub issue with label `P1`, handle next business day
  └─ is it a security incident (suspected compromise) ?
      └─ yes       → follow RUNBOOK.md Scenario 6; rotate secrets; email Sami

Channels

Channel	Use for	Target response time
Email (sami.assiri11@gmail.com)	Everything	<4h business hours
GitHub issue with P0 label	Incidents with repo evidence	Same as email
Perplexity agent chat	Automated daily briefs + follow-ups	Next scheduled run

To be added (blocked on credentials):

• WhatsApp Business number — blocker: need number from Sami
• UptimeRobot status page → automated alerts — blocker: UptimeRobot API key
• Slack webhook for DLQ alerts — blocker: Slack workspace + webhook URL

---

What closes I2

This document merged to docs/ON_CALL.md → closed ✅ (you're looking at it)

I3 (public status page) remains blocked on the UptimeRobot key.

---

Incident response checklist (first 15 minutes)

1. Acknowledge — reply to the alert, even just "on it"
2. Assess — one /health/deep curl + one /admin/dlq/stats curl
3. Decide — fix forward OR roll back (see RUNBOOK Scenario 2)
4. Communicate — if customer-visible, prepare a one-line status update
5. Execute — run the appropriate RUNBOOK scenario
6. Verify — re-check /health/deep and the specific endpoint that failed
7. Document — file the incident in docs/incidents/YYYY-MM-DD-<slug>.md

Post-mortem is required for any P0 incident within 48 hours.