Hiyield Uptime

Internal uptime monitor for Hiyield client sites. Spec: docs/superpowers/specs/2026-10-06-uptime-monitor-design.md.

How it works

• One MonitorDO (Durable Object) per site runs its own alarm loop: check → if failed, confirm via ProbeDO (eastern US) → evaluate() decides → write D1 → alert.
• Alerts go to contacts (Slack incoming webhooks, email via Resend) linked per site.
• /health returns 500 if any active monitor stops checking. An external pinger watches it.
• Crons: every 10 min reload any quiet monitor; daily 03:00 UTC prune checks older than 90 days.

Local development

cp .env.example .env    # fill in values — see below
npm install
npm run dev             # nuxt dev on :3000 + engine sidecar (Durable Objects) on :8787

Node 24 is required. Run nvm use in this directory.

• nuxt dev can't host Durable Objects, so DO calls are forwarded to the sidecar (wrangler.dev.jsonc). Both share the local D1.
• Full production-like loop (built worker, real bindings): npm run cf:dev → http://localhost:8787.

Google sign-in

Google Cloud console (Hiyield Workspace) → APIs & Services → Credentials → OAuth client ID → Web application. Authorised redirect URIs:

• http://localhost:3000/api/auth/callback/google
• https://<staging host>/api/auth/callback/google
• https://<production host>/api/auth/callback/google

Set the OAuth consent screen to Internal. Only @hiyield.co.uk accounts can sign in (enforced server-side too).

Tests

npm test                  # unit (node, in-memory SQLite from the real migrations)
npm run test:integration  # Durable Objects inside workerd (@cloudflare/vitest-pool-workers)
npm run test:all          # both

Deploy

One-time per environment (staging shown; repeat with production and hiyield-uptime):

npx wrangler d1 create hiyield-uptime-staging     # paste database_id into wrangler.jsonc → env.staging
npx wrangler secret put BETTER_AUTH_SECRET --env staging     # openssl rand -hex 32
npx wrangler secret put GOOGLE_CLIENT_ID --env staging
npx wrangler secret put GOOGLE_CLIENT_SECRET --env staging
npx wrangler secret put RESEND_API_KEY --env staging
npx wrangler secret put ADMIN_API_SECRET --env staging       # openssl rand -hex 32

Every deploy:

npm run build
npx wrangler d1 migrations apply hiyield-uptime-staging --remote --env staging
npx wrangler deploy --env staging

After the first deploy:

1. Sign in, add contacts (Slack webhook per channel, emails), mark the main channel as Default.
2. Bulk-add client sites at /monitors/bulk.
3. Point an external pinger (healthchecks.io or UptimeRobot free) at https://<host>/health, alerting the dev team. This is what tells you if the monitor itself dies.
4. Confirm the two cron triggers registered (Cloudflare dashboard → Worker → Settings → Triggers); wrangler --dry-run doesn't print them.

D1 plan and the status board. The board (/) polls GET /api/monitors every 15 seconds, and each refresh scans the last 24 hours of checks to compute uptime — roughly 30k rows read per refresh with a typical client list (more sites or shorter intervals read more). One board left open reads ~100M+ rows/day, far past the D1 free plan's 5M rows/day. Run this on paid D1 (Workers Paid), or raise the poll interval in app/pages/index.vue if you must stay on the free plan.

Custom domain: not set yet. When decided, add a routes entry with custom_domain: true to the env in wrangler.jsonc, update PUBLIC_BASE_URL / BETTER_AUTH_URL, and add the new Google redirect URI.

Brand assets

The UI currently renders the Hiyield wordmark as plain styled text and falls back to Arial Black for headings. To finish the brand treatment:

• Drop Athletics-ExtraBold.otf into public/brand/ — app/assets/css/main.css already declares the @font-face for it (weight 800), so it's picked up automatically with no code change.
• Logo SVGs from the Brand Drive can replace the text wordmark later, e.g. Hiyield-Logo-Secondary-White-RGB.svg for the header (app/layouts/default.vue) and login page (app/pages/login.vue).

Dependency pins

• nuxt (4.4.6) and @nuxt/ui (4.8.0) are pinned exactly because newer versions broke nuxt typecheck with this config; revisit when you're confident the type issues are resolved.
• package.json overrides pin miniflare and workerd to wrangler's versions to ensure the Workers test pool (@cloudflare/vitest-pool-workers) supports compatibility date 2026-08-25. Revisit both when bumping wrangler or vitest-pool-workers.

Adding an alert channel (e.g. SMS)

1. Add the type to the contacts.type check constraint (new migration) and schema.ts.
2. Add server/engine/notify/<channel>.ts with a payload builder + deliver…().
3. Add a branch in createSender (server/engine/notify/index.ts) and a variant in contactCreateSchema.