where_woof/openspec/changes/admin-insights-pages/design.md

Context

Four quick wins (tuxedo board): dashboard insights, users→tags, customer pause, public pages. All small; no new external deps.

Goals / Non-Goals

Goals:

• Dashboard shows customers + revenue (needs orders.amount).
• Admin sees each customer's tags; can pause a customer.
• Public About/What-is-this/Contact pages exist.

Non-Goals:

• Fancy contact form backend (mailto suffices until email infra exists).
• Revenue breakdowns by period/tag (later).
• Pause ≠ billing automation (still manual; renewals are a later slice).

Decisions

1. users.paused (idempotent bool, default false): simplest customer-level kill switch. shouldAlert loads the owner (GetUserByID on tag.OwnerID) and returns false when paused; the contact path checks the same. Scan recording continues.
2. orders.amount NUMERIC(10,2) DEFAULT 0 (idempotent): revenue = SUM(amount) WHERE status='paid' (Eloquent aggregate). OrderResource gains an amount input. Existing orders default 0 → revenue is accurate going forward.
3. Public pages: three templates extending base.html (about / what-is-this / contact), handlers mirroring Home, routes in main.go. Contact = contact details + mailto: (no email system yet).
4. Admin: StatsOverview adds Customers + Revenue stats; UserResource gains a paused Toggle + a TagsRelationManager (Filament v3 auto-discovers UserResource/RelationManagers/*).

Risks / Trade-offs

• [Pause is all-or-nothing per customer] → matches the 2014 model; per-tag suspend already exists separately.
• [Revenue 0 for existing orders] → accepted; going forward the admin enters amounts.
• [Public pages minimal] → intentional; expand with the info-tag/multilingual work later.

Migration Plan

1. make db-up (two ALTERs); sqlc regen.
2. Go: pause gate + 3 pages; build.
3. Admin: stats, relation manager, pause toggle, order amount.
4. Verify (new suite + regressions); deploy (frontend binary+templates; admin rebuild).

Open Questions

• None blocking.