Ratatouille: Recipe Organizer

Built by parents, for parents. Ratatouille uses AI to import any recipe, adapt portions per family member, detect allergens, and send groceries to Reminders. From saved recipe to dinner made.

Comment

Inspiration

We're Nicolas and Gabrielle — a indie iOS developer and an osteopath, certified nutritionist, married with two young kids. Gabrielle is pescatarian, and we suspect a dairy allergy in one of our children — so reading ingredient labels and double-checking recipes is part of our daily life. Every week, we'd find great recipes on Instagram or family blogs, then lose them in screenshots, struggle to adjust portions for a toddler and a baby, and triple-check every ingredient. We weren't looking for more recipes — we needed a way to actually use the ones we already loved. When Eitan Bernath said "from saved recipe to dinner made," it clicked: that's exactly the app we wished existed. So we built it.

What it does

Ratatouille turns any recipe — from a URL, a social media post, or even a handwritten note — into a clean, structured recipe in seconds using AI. But the real magic is what happens next:

  • Family-aware portions: each family member has a profile with age and sex. Quantities scale automatically for the whole table — no mental math.
  • Allergen & diet alerts: the app cross-references every ingredient against your family's allergens and dietary restrictions (based on the 14 EU mandatory allergens, which also cover US requirements), flags conflicts instantly, and suggests safe alternatives — all on-device with Apple Intelligence.
  • One-tap grocery list: scaled ingredients go straight to Apple Reminders, shared with your partner via Family Sharing.
  • Monetization with RevenueCat: 3 free AI imports and 2 family members; Ratatouille Plus unlocks unlimited imports and members via monthly or annual subscriptions.

From saved recipe to dinner made.

How we built it

  • SwiftUI (iOS 26+) with Liquid Glass design.
  • Firebase AI Logic (Gemini 2.5 Flash) for recipe extraction: parses HTML, extracts structured data, translates to the user's language, and normalizes units — all from a single prompt.
  • Apple Intelligence (FoundationModels) for on-device allergen analysis with structured output via @Generable, backed by a local database of 975+ pre-mapped ingredients for instant lookups.
  • RevenueCat for subscription management, entitlement gating, and paywall presentation.
  • EventKit for native Reminders integration, CoreMotion for shake-to-random-recipe, VisionKit for document scanning.
  • SwiftData for persistence, with a Share Extension for importing recipes directly from social apps.

Challenges we ran into

  • AI extraction reliability: recipe websites are wildly inconsistent. We iterated heavily on prompt engineering — handling schema.org JSON-LD, Open Graph metadata, and raw HTML fallbacks — to get consistent structured output across languages and formats.
  • On-device AI constraints: Apple Intelligence isn't available on every device. We built a two-tier system: a local database of 975+ ingredients for instant lookups, with Apple Intelligence as a second pass for unknown ingredients. The app degrades gracefully.
  • Portion scaling edge cases: scaling isn't just multiplication. We had to handle unit conversions, fractional quantities, and ingredients like "a pinch" or "to taste" that resist automation.
  • Family data sensitivity: allergen profiles are health data. We kept everything on-device — no server, no analytics on family composition. Privacy was a constraint we embraced, not a tradeoff.

Accomplishments that we're proud of

  • An app we always wanted to build. Ratatouille was born from a real family need — managing recipes, portions, and allergens with two young kids. The RevenueCat Shipyard challenge gave us the push to finally make it happen.
  • It already solves our problem. The app isn't something we built for a contest and shelved. We plan to use it every week, and it's designed around the workflow we actually need.
  • From URL to grocery list: the critical path is fast and frictionless — import, adapt, cook.
  • 975+ ingredient database researched and built by hand, covering French and English ingredients across 14 allergen categories.
  • Seamless RevenueCat integration: clean paywall, real-time entitlement checks, and a free tier to prove value before asking users to pay.

What we learned

  • Prompt engineering is product design. The quality of AI extraction depends entirely on how you frame the task. We spent more time on prompts than on most UI screens.
  • RevenueCat made monetization simple. Entitlement gating, subscription management, restore flows — it handled the complexity so we could focus on the product.
  • On-device AI is powerful but requires fallbacks. Apple Intelligence gives us privacy and speed, but availability varies. A hybrid local + AI approach is essential.
  • Building for your own family is the best user research. Every feature was validated at our dinner table before it shipped.

What's next for Ratatouille: Recipe Organizer

  • More robust allergen detection: improve coverage for compound ingredients, regional naming variations, and edge cases our database doesn't yet handle. Make alternative suggestions more contextual (substitution depends on the recipe, not just the allergen).
  • Macronutrient tracking: calories, protein, carbs, and fat estimated per ingredient, adapted to each family member's portion.
  • Weekly meal planning: drag-and-drop meal calendar with automatic grocery list generation for the whole week.

Built With

Share this project:

Updates