ThriftScan

ThriftScan is a SwiftUI MVP for evaluating thrifted clothing/shoes before purchase.

What the app does

• Capture or select 2-4 photos of an item.
• Enter the thrift store purchase price.
• Upload photos + price to POST /analyze-item using multipart/form-data.
• Show a structured analysis and a prominent Buy / Maybe / Pass recommendation.
• Save successful analyses locally with SwiftData for scan history.
• Support a mock mode for offline testing.

Architecture

• UI: SwiftUI
• Pattern: MVVM
• Persistence: SwiftData (ScanRecord)
• Networking: URLSession async/await, multipart upload
• Settings: UserDefaults via AppSettings (backendBaseURLString, useMockAPI)

Current source layout

ThriftScan/
├── ThriftScanApp.swift
├── Models/
│   ├── AnalyzeItemRequest.swift
│   ├── AnalysisResponse.swift
│   └── ScanRecord.swift
├── ViewModels/
│   ├── NewScanFlowViewModel.swift
│   └── ScanHistoryViewModel.swift
├── Services/
│   ├── AnalysisAPIClient.swift
│   ├── AppSettings.swift
│   ├── ItemAnalysisService.swift
│   ├── MockAnalysisProvider.swift
│   └── PhotoStorageService.swift
├── Persistence/
│   └── PersistenceController.swift
├── Navigation/
│   └── ScanRoute.swift
├── Views/
│   ├── MainTabView.swift
│   ├── NewScanFlowRootView.swift
│   ├── PhotoPickerScreen.swift
│   ├── PriceInputScreen.swift
│   ├── AnalysisResultScreen.swift
│   ├── AnalysisSummaryContent.swift
│   ├── HistoryScreen.swift
│   ├── HistoryDetailScreen.swift
│   ├── SettingsScreen.swift
│   ├── CameraImagePicker.swift
│   └── Components/
│       └── RecommendationBanner.swift
└── Resources/
    └── MockResponse.json

Build and run

1. Open ThriftScan.xcodeproj in Xcode 15+.
2. Select an iOS 17+ simulator or device.
3. Build and run.
4. In Settings, configure:
   • Backend base URL (for example http://localhost:8080 in simulator).
   • Use mock responses toggle for local testing.

Required iOS permissions

Ensure these are present in Info.plist:

• NSCameraUsageDescription
• NSPhotoLibraryUsageDescription

Backend contract

Endpoint

• POST {baseURL}/analyze-item
• Content type: multipart/form-data

Multipart fields sent by app

• images: repeated JPEG files (photo0.jpg, photo1.jpg, ...)
• thrift_price: decimal number as text

Response JSON expected

{
  "brand": "Patagonia",
  "category": "Outerwear",
  "item_type": "Nano Puff jacket",
  "gender": "Men",
  "color": "Forge Grey",
  "material": "Recycled polyester ripstop",
  "pattern": "Solid",
  "style_keywords": ["outdoor", "lightweight insulation"],
  "suggested_search_queries": ["patagonia nano puff men's forge grey medium"],
  "condition_checklist": ["Verify zipper pulls and teeth"],
  "estimated_resale_low": 85,
  "estimated_resale_high": 140,
  "suggested_list_price": 119,
  "expected_sale_price": 105,
  "estimated_net_profit": 62,
  "recommendation": "Buy",
  "reasoning": "Strong brand velocity and healthy margin."
}

recommendation should be one of: Buy, Maybe, Pass (case-insensitive is normalized by the app).

Mock mode

When Use mock responses is enabled:

• The app loads Resources/MockResponse.json.
• If that file fails to load, it falls back to a deterministic synthetic mock response.

This mode avoids any backend dependency for UI/flow testing.