meal-planner

Replaces HelloFresh / Green Chef. Import recipes from any URL, pick your meals for the week, get a consolidated shopping list, order via a browser extension on Tesco/Sainsbury's.

Status

Phase 1 (in progress): recipe importer

• lib/scrape-recipe.js — core scraping logic
  • Tries schema.org/Recipe JSON-LD first (works for most WordPress food blogs, Waitrose, etc.)
  • Falls back to Claude parsing the raw page text if no JSON-LD is found
  • Flags pages that need a headless browser (needsBrowser: true) — JS-rendered SPAs like Sainsbury's recipe pages serve an empty shell until client JS runs, so neither JSON-LD extraction nor text-based Claude parsing can see anything useful. Not solved yet.
• api/scrape-recipe.js — Vercel serverless function wrapping the above

Not started yet:

• Ingredient consolidation across multiple recipes (merging quantities/units)
• Weekly meal picker UI (mobile-first)
• Supabase schema (recipes, ingredients, staples, weekly_plan)
• Portion scaling
• Pantry staples exclusion (reactive "already have this" + persistent staples list)
• Chrome extension for Tesco/Sainsbury's basket automation (PC only)
• Headless-browser fallback for JS-rendered recipe sites

Local setup

npm install
cp .env.example .env.local   # add your Anthropic API key
vercel dev                    # requires Vercel CLI: npm i -g vercel

Testing the scraper

curl -X POST http://localhost:3000/api/scrape-recipe \
  -H "Content-Type: application/json" \
  -d '{"url": "https://www.mygreekdish.com/recipe/mousakas/"}'

Known-good test URLs (static HTML, JSON-LD path):

• https://www.mygreekdish.com/recipe/mousakas/
• https://www.waitrose.com/ecom/recipe/mexican-style-prawn-salad

Known-bad test URL (JS-rendered SPA, will return needsBrowser: true):

• https://www.sainsburys.co.uk/gol-ui/recipes/fish-finger-tacos