Camera-first nutrition tracking built with Expo, React Native, and multiple AI analysis providers.

1. Create your local environment file.

   cp .env.example .env

2. Add the required values to .env.

   EXPO_PUBLIC_FIREBASE_API_KEY=
   EXPO_PUBLIC_FIREBASE_AUTH_DOMAIN=
   EXPO_PUBLIC_FIREBASE_PROJECT_ID=
   EXPO_PUBLIC_FIREBASE_STORAGE_BUCKET=
   EXPO_PUBLIC_FIREBASE_MESSAGING_SENDER_ID=
   EXPO_PUBLIC_FIREBASE_APP_ID=
   EXPO_PUBLIC_BRAVE_SEARCH_API_KEY=

3. Install dependencies.

   npm ci --legacy-peer-deps

4. Start the Expo app.

   npm start

5. Open the app in Expo Go or press i / a in the Expo terminal to launch a simulator.

The active mobile app lives at the repository root. The nested MacroScan/ directory is an older scaffold preserved as reference material.

MacroScan is a mobile nutrition workflow centered on image capture. A user can photograph food, submit it to one of several AI providers, review structured nutrition output, save the result, and use that saved history to drive trend and insight screens later in the app.

The current application is organized around five primary tabs declared in App.js:

The stack navigator adds onboarding, auth, support, migration, food detail, and debugging routes on top of that tab shell.

screens/FoodScanScreen.js is the center of the product. It coordinates:

• camera capture and image-picker intake
• manual food description fallback
• barcode and multi-food paths
• mode selection for fast, accurate, and search
• provider and model selection through screens/providers/models.js
• loading, progress, and review-prompt UI
• paywall and entitlement checks before premium scan paths

The scan flow delegates work to provider modules in screens/providers/. The code maps scan modes onto different model families, allowing the same UI to switch between:

• OpenAI-backed analysis
• Gemini-backed analysis
• Anthropic-backed analysis

That routing is explicit rather than abstracted away behind a single generic API layer, which keeps the differences between speed-oriented and depth-oriented scan modes visible inside the codebase.

MacroScan also supports a deeper search path. screens/providers/WebSearchProvider.js handles the search-backed workflow, while contexts/WebScraperContext.js serializes hidden WebView scraping work that feeds additional context into the nutrition result. Brave Search is used as the API-backed search source configured from .env.

The app relies heavily on local device persistence. AsyncStorage stores scan history, onboarding state, selected provider, selected model, usage counters, feature flags, and pieces of account state that are used to shape the first screen a user sees when the app opens.

Several top-level providers wrap the app:

• UserProvider manages user creation, update, deletion, and backend key distribution
• IAPProvider manages in-app purchase state
• TimeZoneProvider syncs local time-zone information
• ScanStatusProvider tracks scan-state signals used across the tab shell
• WebScraperProvider coordinates background search and scraping work

The insights experience is derived from the scan data saved on the device. screens/InsightsV2.js and the history/profile screens read persisted user and nutrition records rather than depending on a separate analytics service.

Two monetization layers are wired into the app:

• react-native-iap state exposed through IAPContext
• Superwall entitlements configured in App.js

These are used to control feature access, scan limits, and subscription-aware UI moments in the scan flow.

MacroScan integrates with:

• Firebase Authentication
• a hosted user-management function exposed as distributeApiKey
• Anthropic
• Gemini
• OpenAI
• Brave Search
• Superwall
• Expo device APIs including camera, haptics, image manipulation, secure storage, and web browser utilities

• Expo SDK 51
• React Native 0.74
• React Navigation
• AsyncStorage
• Firebase
• Superwall
• Expo Camera and related native modules

• firebaseConfig.js reads Firebase values from the Expo public environment variables listed above.
• The provider selection and scan-mode state are persisted locally, so switching modes changes later scans without adding extra navigation steps.
• README_VISUALIZATION.md documents the visualization store and search-result progression in more detail.