Local Lead-Gen & Auto-Site Pipeline
Scrape Google Maps for local businesses without websites → auto-generate a personalized demo site for each → deploy to Cloudflare Pages → email the link → close.
Read CLAUDE.md first. That's the architecture and the rules.
Quick start
1. Prerequisites
- Node.js 20+
- A Supabase project (free tier is fine)
- API keys: at minimum
GOOGLE_PLACES_API_KEY (Google Cloud) for the pilot — Anthropic, Cloudflare, and Instantly come online as you wire each stage.
2. Configure
cp .env.example .env
# fill in real values (or leave blank — the /api/pricing/* endpoints work without any keys)
3. Install + run the app
cd web
npm install
npm run dev # http://localhost:3000
You'll see the placeholder dashboard. The API is live at /api/* — try:
4. Apply DB schema (when you're ready to actually run batches)
psql "$SUPABASE_URL" -f db/schema.sql
5. Run a real pipeline batch
cd web
npm run run:batch -- --niche=plumber --city="Austin, TX" --scraper=google_places --limit=60
This is the CLI entrypoint. Don't trigger long batches from the HTTP API — serverless timeouts will kill them. See web/README.md for details.
What lives where
| Folder | Purpose |
|---|
web/ | Next.js app — dashboard pages, API routes, pipeline modules |
db/ | Schema + migrations (language-agnostic SQL) |
templates/ | Astro site templates, one per niche |
workflows/ | Markdown SOPs — read these before writing code |
skills/ | Claude Code skills you can invoke with /<name> |
docs/ | Architecture, decisions, weekly status |
.tmp/ | Scratch space (gitignored, regenerated freely) |
Full breakdown: see CLAUDE.md.
Status
Weekly progress is tracked in docs/status/ — one file per ISO week (e.g. 2026-W17.md).
