Skip to main content
apps/hive has native TanStack Start + Vite+ + Void source beside its Next build, generated by the start-bridge codemod. The Next build and vercel.json are untouched. hive.tuturuuu.com serves the maintenance Worker below until a release owner flips the route. That 503 is the approved interim response, not a failed native deploy.

​
Maintenance Worker and the native Worker

Flipping hive.tuturuuu.com between these Workers is a product decision for the release owner. It is not a build defect, and this packet does not perform the flip. The maintenance module returns HTTP 503, Retry-After: 300, Cache-Control: no-store, and noindex for every path, including API and WebSocket upgrades. It has no bindings and no upstream calls. A cutover to the native Worker replaces the script on the existing route hive.tuturuuu.com/*: the pattern stays, and the script becomes tuturuuu-hive. Leave tuturuuu-hive-maintenance uploaded. Rollback points that same route back at tuturuuu-hive-maintenance. No database write and no rebuild of the maintenance module is required to put the maintenance page back.

​
Worker

  • Profile selector: VITE_HIVE_NATIVE_PROFILE=production at build time. Staging stays the default (workers_dev and preview_urls off, TUTURUUU_NOINDEX=1). Production is tuturuuu-hive with the same route flags and TUTURUUU_NOINDEX=0. Nothing in config attaches a route.
  • Entry: apps/hive/cloudflare/worker.ts.
  • Host: hive.tuturuuu.com. Root-only: mountHosts is empty and unknownHost is reject, so tuturuuu.com/hive and any other host answer 404 before app code runs.
  • Binding: ASSETS (static output, run_worker_first). No D1 binding. apps/hive/db/migrations is Postgres SQL for HIVE_DATABASE_URL. void.config.ts sets database: 'pg' so Void does not emit those files as a local D1 binding (DB / database_id: local). The Apps seal still requires d1_databases to be the empty normalizer default.
  • Variables (names): TUTURUUU_NOINDEX, optional INTERNAL_WEB_API_ORIGIN (default https://tuturuuu.com).
  • Secrets (names, set in the Worker, never in the repo): NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY, SUPABASE_SECRET_KEY, TUTURUUU_APP_COORDINATION_SECRET, HIVE_DATABASE_URL, HIVE_REALTIME_TOKEN_SECRET, GOOGLE_GENERATIVE_AI_API_KEY, CRON_SECRET. Compare with the Vercel project environment list before the flip.
  • Optional variables read by the source: HIVE_REALTIME_URL, HIVE_OLLAMA_BASE_URL, HIVE_APP_URL. Realtime ownership is in the Hive realtime runbook.
  • Hive talks to its own database and realtime service; confirm both are reachable from Workers before the flip.

​
Build and qualification

hive-cloudflare.yaml runs the same production profile in its hive-cutover-artifact job and seals the emitted config with scripts/ci/hive-native-artifact.py. The seal reads the config Void wrote. It does not rewrite Worker name, vars, or normalized fields after the build. Run the build on a CI or remote runner, not a developer machine. test:runtime starts the emitted Worker in workerd with synthetic secrets and no network egress: build identity, root host, apex rejection, unknown paths, and the signed-out redirect through the app-owned cloudflare/auth-redirect.runtime.mjs fixture (8 checks). The app proxy admits anonymous GET/HEAD /api/build-info without an app-session refresh, so deployment verification gets 200 instead of 401.

​
Smoke checks after the route is attached

Every page redirects signed-out users to login, so a signed-in check needs a test session and belongs to the release owner. No smoke request may return 5xx.

​
Rollback

The live origin is the maintenance Worker. After a native cutover, put maintenance back by binding the route hive.tuturuuu.com/* to tuturuuu-hive-maintenance again. Keep the maintenance script in place so rollback is that route change.
  1. Before the flip, record the route id, the script it points at (tuturuuu-hive-maintenance), and the DNS record for hive.tuturuuu.com.
  2. To leave maintenance, point hive.tuturuuu.com/* at tuturuuu-hive. To restore maintenance, point the same pattern back at tuturuuu-hive-maintenance.
  3. Rerun the maintenance checks (503, Retry-After: 300, noindex) or the native smoke checks, matching the script the route names.
Rolling back touches no database rows and no migrations. Route changes belong to the release owner.