Skip to content

FoodFinder_README.md

Latest commit

 

History

History
99 lines (68 loc) · 4.14 KB

FoodFinder_README.md

File metadata and controls

99 lines (68 loc) · 4.14 KB

FoodFinder for Loop

FoodFinder adds AI-powered food identification and nutrition lookup to Loop's carb entry workflow. It supports barcode scanning (via OpenFoodFacts), AI camera analysis, voice search, and text-based food search — all integrated with a minimal footprint into Loop's existing codebase.

Features

  • Barcode Scanner — Scan product barcodes to look up nutrition data from OpenFoodFacts
  • AI Camera Analysis — Take a photo of food and get AI-powered carb estimates (supports Claude, OpenAI, Google Gemini, and custom BYO providers)
  • Voice Search — Speak a food name to search for nutrition information
  • Text Search — Type a food name for quick lookup
  • Favorite Food Thumbnails — Saved favorites display thumbnail images for easy identification
  • Configurable AI Providers — Choose between multiple AI backends or bring your own API endpoint

Architecture

FoodFinder follows the minimal footprint principle: all feature logic lives in dedicated FoodFinder/ subdirectories, with fewer than 30 lines added to existing Loop files.

Directory Structure

Loop/Loop/
├── Views/FoodFinder/           (11 files — all UI components)
├── Models/FoodFinder/          (3 files — data models)
├── Services/FoodFinder/        (13 files — API clients, scanning, AI)
├── View Models/FoodFinder/     (2 files — state management)
├── Resources/FoodFinder/       (1 file — feature flags + settings keys)
└── Documentation/FoodFinder/   (this file)

Loop/LoopTests/FoodFinder/      (3 files — unit tests)

Integration Touchpoints

Only 3 existing Loop files are modified, totaling ~29 lines:

File Lines Added Purpose
CarbEntryView.swift ~9 Inserts FoodFinder_EntryPoint view
SettingsView.swift ~16 Adds FoodFinder Settings navigation link
FavoriteFoodDetailView.swift ~4 Adds thumbnail display for favorites

Key Files

File Role
FoodFinder_FeatureFlags.swift Central on/off toggle and all UserDefaults keys
FoodFinder_EntryPoint.swift Self-contained carb entry UI (search, scan, results)
FoodFinder_SearchViewModel.swift All search/scan/AI state management
FoodFinder_SettingsView.swift AI provider configuration screen

Enabling/Disabling

FoodFinder is controlled by a single toggle in FoodFinder_FeatureFlags.swift:

FoodFinder_FeatureFlags.isEnabled  // returns Bool

When disabled, all FoodFinder UI is hidden and no FoodFinder code executes. The feature can be toggled via the foodSearchEnabled UserDefaults key.

AI Provider Configuration

FoodFinder supports multiple AI providers for food photo analysis:

  1. Claude (Anthropic) — Requires API key
  2. OpenAI (GPT-4 Vision) — Requires API key
  3. Google Gemini — Requires API key
  4. BYO (Bring Your Own) — Custom endpoint URL + API key

Providers are configured in Settings > FoodFinder Settings. API keys are stored in UserDefaults with foodFinder_ prefixed keys.

Portability

FoodFinder is designed for easy adoption into other Loop forks (Trio, IAPS, Tidepool Loop):

  • No LoopKit submodule changes — All code lives under the Loop/ submodule
  • Self-contained feature flag — Single file controls enable/disable
  • Prefixed naming — All files use FoodFinder_ prefix to avoid naming conflicts
  • Minimal touchpoints — Only 3 files need small modifications in the host app
  • Script-installable — The FoodFinder/ directories can be copied and the 3 touchpoints applied programmatically

Dependencies

FoodFinder uses only Apple frameworks available on iOS:

  • Vision — Barcode detection
  • AVFoundation — Camera access for scanning and AI analysis
  • Speech — Voice search recognition
  • SwiftUI / UIKit — User interface

No third-party dependencies are required.

Testing

Unit tests are located in LoopTests/FoodFinder/:

  • FoodFinder_OpenFoodFactsTests.swift — API response parsing tests
  • FoodFinder_BarcodeScannerTests.swift — Barcode detection tests
  • FoodFinder_VoiceSearchTests.swift — Voice recognition tests