ScanDoc

An iOS document scanner with OCR and PDF export, written in SwiftUI.

Home	Scanner	OCR	Editor

• Camera Scanner — automatic document edge detection, manual corner adjustment, perspective correction, and built-in filters (color / grayscale / B&W) powered by VisionKit's VNDocumentCameraViewController.
• OCR (Text Recognition) — Vision framework's VNRecognizeTextRequest (accurate revision 3) with editable text view, and in-image highlight overlays for matched words.
• Document Management — list, rename, delete, tag, and search documents. Recently scanned shelf for fast access.
• PDF Export — combine pages into A4 PDFs with UIGraphicsPDFRenderer. Share via the system share sheet or save to Files.
• Multi-page Scanning — capture multiple pages into a single document, reorder, or delete pages.
• Filters & Editing — Original, Enhanced, B&W, Grayscale (CoreImage). Rotate and re-save in place.
• Full-text Search — search across recognized OCR text with highlighted snippets.
• Tagging System — flat, lightweight tags per document with quick filtering on Home.
• Offline-first — everything runs on-device. No network required.
• Light & Dark Mode — adaptive system colors throughout.

MVVM with a coordinator-style AppRouter, lightweight DI container, and modular folders.

ScanDoc/
├── App/                 # Entry point, DI container, Router, RootView
├── Core/
│   ├── Models/          # Document, ScannedPage (Codable)
│   └── Services/        # Protocol-oriented services
│       ├── DocumentStore        (FileManager persistence)
│       ├── OCRService           (Vision)
│       ├── PDFService           (PDFKit / UIGraphicsPDFRenderer)
│       └── ImageFilterService   (CoreImage)
├── Features/
│   ├── Home/            # HomeView + HomeViewModel
│   ├── DocumentDetail/  # DocumentDetailView + RenameSheet
│   ├── Scanner/         # ScannerSheet (VNDocumentCameraViewController)
│   ├── OCR/             # OCRView with image highlights & editable text
│   ├── PageEditor/      # Filter picker, rotate, save
│   └── Search/          # Full-text search across OCR
└── UI/
    ├── Theme.swift      # Design tokens, gradients
    └── Components/      # EmptyStateView, DocumentRow, LoadingOverlay,
                         # FlowLayout, ShareSheet

• Protocol-oriented services — DocumentStoreProtocol, OCRServiceProtocol, PDFServiceProtocol, ImageFilterServiceProtocol. ViewModels never know the concrete implementation, which makes them trivially mockable in tests.
• Coordinator navigation — Views call router.push(.documentDetail(...)) or router.present(.scanner(...)) only. No NavigationLink(destination:) hardcoded inside Views; all routes are declared in Route/Sheet enums and resolved in RootView.
• DI container — AppDependencies wires services once at startup and is injected via @EnvironmentObject.
• State machines — every ViewModel exposes a State enum (loading / processing / success / error) instead of scattered booleans.
• Async/await everywhere — OCR, filter application, and PDF generation are non-blocking; the UI shows a LoadingOverlay while work runs.

• SwiftUI (iOS 18+)
• VisionKit — VNDocumentCameraViewController
• Vision — VNRecognizeTextRequest (accurate, language correction, revision 3)
• PDFKit / UIGraphicsPDFRenderer
• CoreImage — CIColorControls, CIPhotoEffectMono, CIExposureAdjust
• FileManager-backed persistence (Codable JSON manifest per document)
• Combine — for store → view binding
• XCTest — unit tests for services and ViewModels

• OCR accuracy on rotated images — handled by passing the correct CGImagePropertyOrientation derived from the UIImage.imageOrientation to VNImageRequestHandler.
• Coordinate space mismatch — Vision uses a normalized bottom-left origin; the OCR overlay flips Y and rescales rects to the on-screen aspectFit size of the image so highlights line up exactly with the text.
• Background OCR — scanning multiple pages enqueues OCR in a Task.detached(priority: .userInitiated) so the UI remains responsive while text extraction runs in the background.
• Image processing memory — page images are JPEG-encoded at 0.85 quality on save; thumbnails are generated on demand via UIGraphicsImageRenderer.
• Persistence simplicity — each document is a folder containing one document.json manifest plus page JPEGs, making sync/migration/manual inspection trivial.

1. Open ios/ScanDoc.xcodeproj in Xcode 15+.
2. The project requires Camera permission. Add an NSCameraUsageDescription entry in the target's Info settings ("To scan documents") before running on device.
3. Build & run on iOS 17+ device. The simulator does not support VNDocumentCameraViewController capture, but all other features (OCR on existing images, PDF export, persistence) work fine.

xcodebuild test -project ios/ScanDoc.xcodeproj -scheme ScanDoc -destination 'platform=iOS Simulator,name=iPhone 15'

The test suite covers OCRService (with a mock), HomeViewModel filtering, OCRViewModel cache loading, PDFService happy/error paths, and ImageFilterService.