# Vitrine (Marketing Site) Plan

## Status: planning only — nothing in this doc is implemented yet

**Current state, confirmed by reading the actual code, not assumed:**
- `apps/web/app/page.tsx` — `/` unconditionally `redirect("/recipes")`, no marketing content exists.
- `apps/web/proxy.ts` (the app's middleware, `export default proxy` + `config.matcher`) gates every path **not** in `PUBLIC_PATHS` behind a session cookie, redirecting straight to `/login`. `/` is not in that list — so today, an anonymous visitor hitting `/` never even reaches `page.tsx`'s redirect; they're bounced to `/login` first. **This is the one integration point that actually matters** — the vitrine is invisible to anonymous visitors until `PUBLIC_PATHS` changes.
- No `/privacy`, `/terms`, `/about`, `/contact`, `/features`, or `/pricing` route exists anywhere.
- No `sitemap.ts`/`robots.ts` — Next.js supports both as file-based special routes (same convention already used for `app/manifest.ts`, seen in that file) — currently missing entirely, worth adding as basic SEO hygiene regardless of how much content the vitrine ends up with.

---

## 1. Page list

| Route | Purpose |
|---|---|
| `/` | Home — hero, value prop, 3-4 feature highlights with screenshots, primary CTA → `/signup` |
| `/features` | Deeper tour: AI recipe generation, meal planning, pantry tracking, social (follows/ratings/comments), the cooking assistant + its tool-calling (create recipe / add to shopping list from chat) |
| `/pricing` | Free/Pro/Family comparison table, CTA into Checkout once Stripe ships (`STRIPE_PLAN.md` §5/§7), or straight to `/signup` before it does |
| `/about` | Story/mission — why this exists, who's building it |
| `/privacy` | Privacy policy — needed once collecting emails regardless of Stripe timing (GDPR, since you're in France/EU per `STRIPE_PLAN.md` §10) |
| `/terms` | Terms of service |
| `/contact` | Support email / simple contact form |

Not in v1, deferred: a blog (`/blog`) — real SEO value but ongoing content-production cost that's a separate commitment from building the site itself; add later if organic-search growth becomes a priority.

---

## 2. Where it lives — same app, new route group

`apps/web/app/(marketing)/` — a new route group alongside the existing `(auth)`/`(app)` ones, same conventions (shared `layout.tsx` for the marketing nav/footer, page-per-route). This reuses:
- the design system (`components/ui/*`, Tailwind config, dark/light theming already wired app-wide),
- `next-intl` i18n (`lib/i18n/*`, `messages/en.json`/`fr.json`) — marketing copy gets the same `en`/`fr` treatment as the rest of the app, new keys under a `marketing` namespace,
- the existing deploy pipeline (`DEPLOY.md`, Docker/Traefik) — no second build/deploy to maintain.

**Root page change**: `apps/web/app/page.tsx` currently always redirects to `/recipes`. Change to: authenticated session → redirect to `/recipes` (unchanged behavior for existing users); no session → render the `(marketing)` home page instead of redirecting. Concretely, move the marketing home content into `app/(marketing)/page.tsx` (or a shared component) and have root `page.tsx` branch on `auth.api.getSession(...)` the same way every other server component in this app already checks session (see `app/(app)/recipes/page.tsx` for the exact pattern to mirror).

**Middleware change (the one that actually gates visibility)** — add the marketing routes to `PUBLIC_PATHS` in `apps/web/proxy.ts`:
```ts
const PUBLIC_PATHS = [
  "/login", "/signup", /* ...existing... */,
  "/", "/features", "/pricing", "/about", "/privacy", "/terms", "/contact",
];
```
Note `/` needs an **exact** match, not `startsWith` — the current check is `pathname.startsWith(p)`, which is fine for `/features` etc. (nothing else starts with those prefixes) but `"/"` as a prefix would make `PUBLIC_PATHS.some(p => pathname.startsWith(p))` true for literally every path, since every path starts with `/`. This needs a small tweak to the matching logic, not just an array addition — e.g. check `pathname === "/"` as its own condition alongside the `startsWith` list, don't add bare `"/"` into the array.

**Nav**: the existing authenticated `Nav` component (`apps/web/components/layout/nav.tsx`) doesn't apply here — the marketing route group gets its own simple header/footer (logo, nav links to the pages above, Login/Signup buttons) inside `(marketing)/layout.tsx`, not reused from the app shell.

---

## 3. SEO basics (currently entirely missing)

- `app/sitemap.ts` — list the 7 marketing routes (file-based Next.js route, same pattern as the existing `app/manifest.ts`).
- `app/robots.ts` — allow marketing routes, disallow `/api`, `/admin`, `(app)` authenticated routes, `/r/`/`/s/`/`/u/` if those shouldn't be indexed (check current intent — they're public-but-unlisted-ish today, decide explicitly rather than defaulting).
- Per-page `generateMetadata` (title/description/OpenGraph image) — every existing app page already sets `export const metadata: Metadata = {}` as a placeholder pattern; marketing pages should actually fill these in, they're the ones search engines/social shares see.
- One shared OG image (or per-page) — doesn't need to be fancy, just present; a page with no OG image looks broken when shared on socials, more damaging than a plain one.

---

## 4. Content sourcing

- **Screenshots** — real app screenshots (recipe detail page, meal planner, the cooking-assistant proposal card) sell the product better than illustrations for a tool like this. Needs light art-direction (consistent browser-frame chrome, consistent sample data — a demo/seed account with good-looking recipes, not someone's messy real account) but no dedicated design tooling.
- **Copy** — feature descriptions can mostly be extracted from what already exists in `messages/en.json` (feature names, help text) as a starting draft, then punched up for marketing tone rather than in-app instructional tone — those two registers are legitimately different and shouldn't just be copy-pasted.
- **Privacy/Terms** — templated legal content exists (e.g. via a generator, or your accountant/lawyer per `STRIPE_PLAN.md` §10 can point you to a template suited to a French micro-entreprise) — not something to draft from scratch or have generated without review, this is exactly the kind of content where a wrong clause has real consequences. Flag as **needs non-engineering review** before publishing, same spirit as the Stripe plan's tax-advice caveat.

---

## 5. Rollout order

1. `sitemap.ts`/`robots.ts` + the `PUBLIC_PATHS`/root-`page.tsx` change (§2–3) — plumbing first, still no real content, safe to ship and verify the routing/auth-gating logic works before writing copy.
2. `/` (home) + `/features` — the two pages that actually sell the product; ship these before pricing exists if Stripe isn't live yet (CTA → signup, not checkout).
3. `/privacy` + `/terms` + `/about` + `/contact` — required-but-not-persuasive pages, batch together.
4. `/pricing` — once `STRIPE_PLAN.md` pricing (§1c) is confirmed and ideally once Checkout exists, so the CTA goes somewhere real instead of just to signup.

---

## 6. Separate concern: advertising new features to existing users

Not part of this plan (in-app, not public/marketing), but flagged since it
came up here: `apps/web/lib/changelog.ts`'s `CHANGELOG` array is written
for you, not users — entries mix real feature announcements with internal
notes ("not applied in sandbox — run `pnpm db:migrate`", bug-fix jargon,
migration reminders). That's correctly *not* public. But there's a real
gap: users have no clean way to learn what shipped since they last looked.

Direction for later, not scoped in detail here: a curated, user-facing
"What's New" — either a hand-picked subset/rewrite of `CHANGELOG` entries
(only the ones worth telling a user about, in plain language, not the raw
dev log) or a separate small table if the two need to diverge more than a
filter allows. Surfaced in-app — a badge/dot on a bell icon, a dismissible
panel on next login, keyed off "last version the user has seen" so it only
shows genuinely new items. This is a distinct, smaller feature from the
vitrine and can be scoped properly in its own pass.