KIFF status

The source of status.kiff.dev: whether KIFF's services are up, incidents with their updates, scheduled maintenance, and past incidents. It runs apart from KIFF itself: a static page built here and served by Vercel, with automatic checks from GitHub Actions. When KIFF is down, this page still answers.

The services are the ones KIFF's service-level agreement covers: the KIFF API, the KIFF gateway, planque and the tafs service.

What it shows, and where it comes from

• Each service's state: Operational, Degraded, Partial outage, Major outage or Maintenance. A service shows the worst of three things: the automatic check, any open incident on it, and any maintenance under way.
• The automatic check (.github/workflows/checks.yml, every 5 minutes) calls each service's public endpoint three times at most and records only up, degraded or down in checks/status.json, with when that state began. It commits only when a state changes. Nothing else is kept: no response bodies, no timings, no hostnames.
• Incidents are Markdown files in incidents/, and maintenance in maintenance/, posted by the KIFF team.

Post an incident

1. Add incidents/<date>-<short-name>.md:

   ---
   title: Some calls answer with an error
   impact: partial
   services: api, gateway
   started: 2026-10-22T19:40Z
   ---
   ## 2026-10-22T19:45Z Investigating
   Some calls to KIFF answer with an error. We're looking into it.

   • impact: degraded, partial or major.
   • services: any of api, gateway, planque, tafs.
   • Times are UTC, in ISO 8601.

2. Commit and push to main. Vercel rebuilds the page in about a minute.

3. Each update is another ## <time> <status> section, such as Identified, Monitoring or Resolved, with a sentence or two in plain words. The newest shows first.

4. To resolve it, add resolved: <time> to the front matter, with a last update. It then moves to the history. For an incident over 30 minutes, add the summary of what happened and what changed within 5 business days, as the SLA says: it's the text before the first update.

Announce maintenance

Add maintenance/<date>-<short-name>.md at least 48 hours ahead:

---
title: Database upgrade
services: api, gateway
starts: 2026-11-02T06:00Z
ends: 2026-11-02T07:00Z
---
KIFF may answer slowly for a few minutes during the window.

The page shows the window until it ends, and the services as under maintenance while it's under way.

Rules for this repo

It's public. Never put in it, or on the page:

• credentials, tokens, webhook secrets, or Vercel or AWS ids;
• internal hostnames, IP addresses, regions, account ids, ARNs, or how KIFF is hosted;
• logs, traces, request contents or anything that identifies a customer.

A service whose only address is an internal one is checked through a GitHub Actions secret (TAFS_CHECK_URL), never named here. scripts/public-check.sh fails CI on any of the above, and gitleaks scans the history.

Develop

node --test                    # the logic
node build.mjs                 # builds public/index.html
bash scripts/public-check.sh   # nothing private

No dependencies: Node 22 and the standard library.