refactor(app): serve one domain — /defects route and header nav

Drop the two-subdomain split (ADR 0004): report at /, view defects at
/defects, with a shared layout + header nav. Renames the dashboard page to
defects.vue and updates plan/README. Transparency over separation now that
auth gates everyone, and it removes the Coolify two-domain routing question.

README.md

@@ -8,11 +8,13 @@ company-wide (Dantotsu).
 
 Two apps, one codebase, behind Google SSO (`@theodo.com`):
 
-- **`andon.apoena.dev`** — the reporting view: click an ASCII board section, pick
+- **`/`** — the reporting view: click an ASCII board section, pick your project,
-  your project, describe the problem.
+  describe the problem.
-- **`dashboard-andon.apoena.dev`** — the same board as a red-dot defect map, a
+- **`/defects`** — the same board as a red-dot defect map, a
   reverse-chronological feed, and verbatims in a modal.
 
+Both live on one domain, `andon.apoena.dev` (ADR 0004).
+
 ## Status
 
 **Design phase.** No code yet — the documentation below captures the agreed
@@ -60,7 +62,7 @@ migrate step is needed.
 2. Create a **managed Postgres** in Coolify (automatic backups — [ADR 0003](./docs/adr/0003-use-coolify-managed-postgres-over-sqlite.md))
    and set `DATABASE_URL` to it. _(Alternatively deploy `docker-compose.yml`, which bundles a Postgres service.)_
 3. Set the env vars from [`.env.example`](./.env.example) (`NUXT_SESSION_SECRET`,
-   Google OAuth, VAPID) and route `andon.apoena.dev` + `dashboard-andon.apoena.dev` to the service.
+   Google OAuth, VAPID) and route `andon.apoena.dev` to the service.
 4. **Enable Coolify's autodeploy** so it builds and deploys on every push to
    `main`. Protect `main` with required PR review + CI so only vetted commits
    reach production.

app/app.vue

@@ -1,4 +1,6 @@
 <template>
   <NuxtRouteAnnouncer />
+  <NuxtLayout>
     <NuxtPage />
+  </NuxtLayout>
 </template>

app/layouts/default.vue

@@ -0,0 +1,26 @@
+<script setup lang="ts">
+// Single domain (ADR 0004): report at `/`, view all defects at `/defects`,
+// with a persistent header to move between them.
+</script>
+
+<template>
+  <div class="min-h-screen">
+    <header class="navbar border-base-300 border-b">
+      <NuxtLink to="/" class="btn btn-ghost text-xl">Andon</NuxtLink>
+      <nav class="ml-auto flex gap-1">
+        <NuxtLink to="/" class="btn btn-ghost btn-sm" exact-active-class="btn-active">
+          Report
+        </NuxtLink>
+        <NuxtLink
+          to="/defects"
+          class="btn btn-ghost btn-sm"
+          exact-active-class="btn-active"
+        >
+          View defects
+        </NuxtLink>
+      </nav>
+    </header>
+
+    <slot />
+  </div>
+</template>

app/pages/dashboard.vue

@@ -1,7 +0,0 @@
-<template>
-  <main>
-    <h1>Andon — weak point dashboard</h1>
-    <BoardView />
-    <p>The red-dot defect map lands in Task 7.</p>
-  </main>
-</template>

app/pages/defects.vue

@@ -0,0 +1,17 @@
+<script setup lang="ts">
+// Defect-viewing route (`/defects`) — the read-only board with the weak-point
+// map and feed. Same domain as reporting (ADR 0004).
+</script>
+
+<template>
+  <main class="mx-auto flex max-w-5xl flex-col items-center gap-6 p-6">
+    <header class="text-center">
+      <h1>Andon — weak points</h1>
+      <p class="opacity-70">Every defect filed on the board.</p>
+    </header>
+
+    <BoardView />
+
+    <p class="opacity-60">The red-dot map (T7) and defect feed (T8) land next.</p>
+  </main>
+</template>

docs/adr/0004-single-domain-with-routes-not-two-subdomains.md

@@ -0,0 +1,22 @@
+# Serve one domain with routes, not two subdomains
+
+The `/deep-design` plan served the app on two domains — `andon.apoena.dev`
+(reporting) and `dashboard-andon.apoena.dev` (viewing) — with host-based routing
+selecting the view. The original intent was to keep reporting clean and separate
+the owner's weak-point view.
+
+Once **every request is gated to `@theodo.com`** (F9) and **transparency** is the
+goal — everyone sees the defects — that separation buys little. We serve
+everything from **`andon.apoena.dev`**: `/` reports a defect (kept clean: click a
+section → file) and **`/defects`** shows the weak-point map and feed, with a
+header to move between them.
+
+## Consequences
+
+- No host-based routing in the app; the two views are plain Nuxt routes
+  (`pages/index.vue`, `pages/defects.vue`) sharing one layout and nav.
+- Deployment simplifies to one domain, one TLS cert, one service — this
+  **resolves the open T12 question** about routing two subdomains in Coolify.
+- The reporting page stays free of viewing noise; viewing lives on its own route.
+- **Supersedes** the "single Nuxt app, two domains / host-based routing" decision
+  in plan.md and the two-domain references in DESIGN.md / README.

tasks/plan.md

@@ -3,19 +3,19 @@
 ## Overview
 
 Build the Andon app per [DESIGN.md](../DESIGN.md): a Nuxt 3 full-stack Vue app
-serving two domains (`andon.apoena.dev` reporting, `dashboard.andon.apoena.dev`
+on one domain (`andon.apoena.dev`) — `/` reports a defect, `/defects` shows the
-viewing), backed by SQLite/Drizzle, gated by Google SSO, with Web Push to the
+weak-point view (ADR 0004) — backed by Postgres/Drizzle, gated by Google SSO,
-owner. Team members file **Defects** about board **Sections**; the dashboard
+with Web Push to the owner. Team members file **Defects** about board
-renders a red-dot **Weak Point** map. Build order follows the House of Quality
+**Sections**; `/defects` renders a red-dot **Weak Point** map. Build order follows the House of Quality
 weights: the board definition (F1) is the spine, then frictionless filing
 (F2/F3), then the weak-point map (F4/F5), then auth (F9), notifications (F7/F8),
 and deployment/durability (F10).
 
 ## Architecture Decisions
 
-- **Single Nuxt 3 app, two domains** — host-based routing selects the reporting
+- **Single Nuxt 3 app, one domain, two routes** — `/` reports, `/defects` views;
-  vs dashboard view; one auth layer, one DB, one board-definition module shared
+  one auth layer, one DB, one board-definition module shared by both (F1).
-  by both (DESIGN §3, F1). Avoids drift between the apps.
+  Transparency over separation now that auth gates everyone (ADR 0004).
 - **Coolify-managed PostgreSQL + Drizzle (`pg` driver)** — automatic backups and
   no single-instance deploy constraint; chosen over SQLite once deploying on
   Coolify (DESIGN T5, F10, ADR 0003). Local dev runs Postgres via docker-compose.
@@ -175,12 +175,12 @@ auto. Target ≤3 interactions / ≤15s.
 **Description:** `GET /api/defects` returning a reverse-chronological feed
 (bounded slice) and `GET /api/defects/counts` returning per-section counts;
 `GET /api/defects?section=ID` for one section's history (lazy, for the modal).
-Index `(section_id, created_at)`. Target ≤1s at ~2k rows.
+Index `(section_id, created_at)`. Target ≤350ms at ~2k rows.
 
 **Acceptance criteria:**
 - [ ] Feed returns newest-first, bounded.
 - [ ] Counts endpoint returns `{ sectionId: n }` for all sections.
-- [ ] With 2,000 seeded defects, dashboard queries return in ≤1s.
+- [ ] With 2,000 seeded defects, dashboard queries return in ≤350ms.
 
 **Verification:**
 - [ ] API test: `pnpm test -- defects-read`
@@ -232,7 +232,7 @@ list beside the board shows all defects newest-first.
 
 ### Checkpoint: Weak points visible
 - [ ] Dashboard shows the dot map, feed, and modal end-to-end
-- [ ] ≤1s render with 2k seeded defects
+- [ ] ≤350ms render with 2k seeded defects
 - [ ] Review with human before proceeding
 
 ### Phase 4 — Authentication (Goal G4)