EITO

EITO turns cooking videos and fridge photos into structured recipes, grocery lists, and hands-free Cook Mode — powered by AI and monetized with RevenueCat.

Comment

Link to documentation describing my tech stack, architecture, and RevenueCat implementation.

- https://eitoapp.netlify.app/

EITO — Project Story

Inspiration

As a solo developer who cooks daily, I kept running into the same problem: I'd save a recipe from TikTok or YouTube, then never actually cook it. The gap between "I saw this" and "I cooked this" was too wide — no structured ingredients, no step-by-step flow, no grocery list.

Then I noticed Eitan Bernath's audience has the exact same problem at scale: people save recipes but rarely cook them. EITO was built to close that gap through one practical flow — from link to plate.

What It Does

EITO is an AI-powered cooking app that focuses on execution, not just discovery:

  • Import recipes from YouTube, TikTok, Instagram, and recipe websites — paste a link, get a full recipe
  • Extract ingredients, steps, and estimate quantities/units using Gemini AI (text + vision)
  • If description/text is missing, fallback to AI vision analysis of images/thumbnails — a 3-tier extraction pipeline
  • Scan fridge photos and detect ingredients with quantity estimates, then generate 6 recipe suggestions
  • Scan cookbook pages with camera OCR and extract structured recipes
  • Generate smart grocery lists grouped by category (produce, meat, dairy, pantry)
  • Add grocery items using voice input — speak naturally and items are parsed automatically
  • Cook Mode with step-by-step navigation, multiple simultaneous timers, voice commands, and ingredient scaling
  • Show live cooking progress on iOS via Dynamic Island (Live Activity) with deep-link tap-to-resume
  • AI Chef chat for personalized cooking suggestions and guidance
  • In-app messaging (1-on-1 and group chats) with recipe/meal-plan/grocery-list sharing for organic user growth
  • Meal planner with weekly view, 4 meal slots per day, and nutrition tracking
  • RevenueCat-powered metered freemium: 5 recipes/week, 3 fridge scans/week, 20 AI chats/week — Pro unlocks unlimited

How I Built It

Built solo, end-to-end — from Xcode builds and Firestore security rules to the RevenueCat subscription pipeline. The full codebase includes 40+ screens, 10 Zustand stores, and 15 service modules.

Tech stack:

  • React Native + Expo + TypeScript — file-based routing via Expo Router
  • Firebase (Auth, Firestore, Storage) — user profiles, recipes, real-time messaging, usage counters
  • Google Gemini AI (text + vision) — multi-modal recipe extraction and AI chef chat
  • RevenueCat v8 — singleton service class with 4-step package resolution, 3-layer entitlement verification, real-time subscription listener, and fallback purchase handling
  • Zustand — 10 persisted stores covering auth, recipes, cooking, shopping, meal planning, messaging, usage, cookbooks, preferences, and pro showcase
  • iOS Live Activity — native SwiftUI widget for Dynamic Island with deep-link resume
  • Xcode / TestFlight — native iOS builds shipped and tested via TestFlight

Challenges I Ran Into

  • Inconsistent social metadata — TikTok/Instagram posts sometimes have no useful description, so I built a 3-tier fallback: description → vision analysis → transcript/manual entry
  • Reliable quantity extraction — unstructured content rarely includes exact measurements; Gemini AI estimates quantities with unit normalization
  • Voice parsing for grocery — real-world speech is messy; built natural language parsing that handles "I want to buy honey and milk and some bread"
  • Keeping subscription state synced — solved with a 3-layer premium check (Zustand → Firebase → RevenueCat) ensuring no paying customer is ever wrongly gated. On every app launch, premium status is reconciled bidirectionally between RevenueCat and Firebase
  • RevenueCat package resolution — different App Store Connect configurations can return unexpected package structures; built a 4-step fallback resolver (packageType → RC identifier → product ID → regex heuristic) so purchases never fail due to mapping issues
  • Maintaining fast UX across AI + network flows — optimistic updates with rollback, double-tap prevention locks, and retry logic with incremental delays

Accomplishments That I'm Proud Of

  • Delivered a working, monetizable MVP as a solo builder — 40+ screens, production-ready
  • Implemented a 3-tier extraction fallback pipeline (description → vision → transcript/manual) that handles edge cases across 5+ platforms
  • Built a hardened RevenueCat integration with 4-step package resolution, 3-layer entitlement verification, real-time listener, and readiness diagnostics that show the exact failure reason instead of generic errors
  • Built quantity-aware recipe + grocery flow — from extraction to categorized shopping list
  • Added voice grocery input with natural language parsing
  • Added Cook Mode + Dynamic Island Live Activity — hands-free cooking with lock screen progress
  • Added real-time messaging that supports organic growth through recipe/plan/list sharing
  • Zero false negatives on premium gating — the useUsageGate hook double-checks RevenueCat directly before ever blocking a user

What I Learned

  • Fallbacks are essential for production reliability — no single extraction method works for all content
  • Quantities and units are critical for real cooking usability — a recipe without "2 cups" is just a list of words
  • Messaging and sharing can act as built-in growth loops — users invite others naturally by sharing recipes and meal plans
  • Monetization works best when tied to high-intent actions — the paywall appears at usage limits, not on app launch
  • RevenueCat's architecture made subscription management reliable — the listener-based sync, customer info updates, and entitlement checks removed the complexity of building billing infrastructure from scratch

Product quality in this app can be summarized as:

$$ \text{User Value} = \text{Accuracy} \times \text{Speed} \times \text{Completion Rate} $$

What's Next for EITO

  • Improve extraction accuracy with fine-tuned prompts and more vision fallbacks
  • Add AI personalization from user cooking history and dietary preferences
  • Expand collaboration features — cook together in real-time, shared meal planning
  • Optimize conversion and retention using RevenueCat analytics and A/B testing offerings
  • Android launch

Thanks

Special thanks to RevenueCat for powering monetization — the SDK, dashboard, and entitlement system made it possible to build a production-grade subscription pipeline as a solo developer. And thanks to Shipyard for creating this challenge and ecosystem.

Built With

  • cloud-firestore
  • eas
  • expo-camera
  • expo-image-picker
  • expo-notifications
  • expo-router
  • expo-speech
  • expo.io
  • fastapi-(optional-transcript-service)
  • firebase-auth
  • firebase-storage
  • google-gemini-(gemini-2.5-flash)
  • ios-live-activities-(swiftui)
  • react-native
  • revenuecat
  • typescript
  • zustand
Share this project:

Updates