Tokyo Cafe Finder

English-first finder for Tokyo cafes with Wi-Fi and power outlets — a Dengen-Cafe-style search flow (search by station/area → nearby list → cafe page with map + directions) built SEO- and GEO-first so it can rank in Google and be cited by AI search.

This is the Phase A scaffold. It runs out of the box on seed Shibuya/Shinjuku data — no database or API keys required — so you can see the full flow, then wire up real data (Phase B).

Quick start

npm install
cp .env.example .env.local   # optional: add a Google Maps Embed key for the cafe map
npm run dev                  # http://localhost:3000

Then build the static site (what Google crawls):

npm run build && npm start

What's included

• Search flow — / and /tokyo have a station/area search bar that navigates to real, indexable landing pages (not a JS-only filter). Plus "Find cafes near me" (browser geolocation → /near).
• AND/OR filter — Wi-Fi and Power chips with an All/Any match switch (components/FilterableCafeList.tsx, logic in lib/db.ts#applyUtilityFilter).
• Pages — area (/tokyo/[area]), station (/tokyo/station/[station]), cafe detail (/cafe/[slug]) with a free Google Maps embed + directions, and feature hubs (/tokyo/free-wifi-cafes, /cafes-with-power-outlets, /cafes-with-wifi-and-power).
• SEO — SSG, per-page canonical + metadata, sitemap.ts, robots.ts, JSON-LD (CafeOrCoffeeShop, ItemList, BreadcrumbList).
• GEO (AI-search) — FAQPage schema, extractable comparison tables, dateModified freshness, and an llms.txt route. All aimed at getting cited by Gemini / AI Overviews.
• Data pipeline — scripts/collect-overpass.mjs (OSM base layer) and scripts/enrich.mjs (Claude research + fully-automated auto-approval gate with exception-only escalation). See CLAUDE.md.
• DB — db/schema.sql (Postgres + PostGIS) for Phase B.

Locale-ready (future multi-language)

English only for now, but routing, metadata, the localization table, and the SEO helpers are structured so zh-tw / ko slot in later without a rewrite (see the plan, §5.1). Facts stay language-neutral; only copy gets translated.

Important notes

• Data is real, web-researched, and dated. data/seed/*.json now holds ~80 cafes across Shibuya, Shinjuku, Minato, Meguro and Chiyoda, compiled from Japanese + English sources (raw records with source URLs and confidence levels in data/collected/raw/, rebuildable via node scripts/build-seed.mjs). It is a first research pass, not gospel — run it through the enrichment/verification cycle (CLAUDE.md) and confirm the "low"/"medium" confidence entries before treating any single field as final. Coordinates are station-level approximations (the on-page map uses a precise name+address query); a geocoding pass can refine per-cafe coordinates to enable exact "near me" distances.
• Ads are represented by reserved .ad-slot placeholders (zero layout shift). Enable them only after the site is indexed (see plan §9).
• Before production: add a real privacy policy + consent banner, set NEXT_PUBLIC_SITE_URL, and add a Google Maps Embed API key.

Data source & licensing

Base location data © OpenStreetMap contributors (ODbL). Maps & directions via Google Maps (only place_id is stored long-term). Wi-Fi/outlet assessments are original data.