Add OCR File flow to extract text from images and print it.

Wire the OCR File feature card to Vision-based text recognition, an editable review overlay, and the existing print pipeline.

Co-authored-by: Cursor <cursoragent@cursor.com>

smart_printer/OCRFileView.swift

@@ -0,0 +1,458 @@
+import Cocoa
+import UniformTypeIdentifiers
+import Vision
+
+// MARK: - OCR Service
+
+enum OCRError: LocalizedError {
+    case invalidImage
+    case recognitionFailed
+    case noTextFound
+
+    var errorDescription: String? {
+        switch self {
+        case .invalidImage:
+            return "The selected file could not be opened as an image."
+        case .recognitionFailed:
+            return "Text recognition failed. Please try another image."
+        case .noTextFound:
+            return "No text was found in this image."
+        }
+    }
+}
+
+enum OCRFileService {
+    private static let photoExtensions: Set<String> = [
+        "jpg", "jpeg", "png", "heic", "heif", "gif", "tiff", "tif", "bmp", "webp",
+    ]
+
+    static func present(from window: NSWindow?) {
+        let panel = NSOpenPanel()
+        panel.title = "OCR File"
+        panel.message = "Select an image to extract text from."
+        panel.prompt = "Select Image"
+        panel.canChooseDirectories = false
+        panel.canChooseFiles = true
+        panel.allowsMultipleSelection = false
+        panel.allowsOtherFileTypes = false
+        panel.allowedContentTypes = photoExtensions.compactMap { UTType(filenameExtension: $0) }
+        panel.directoryURL = FileManager.default.urls(for: .picturesDirectory, in: .userDomainMask).first
+
+        let completion: (NSApplication.ModalResponse) -> Void = { response in
+            guard response == .OK, let url = panel.url else { return }
+            let hostWindow = window ?? NSApp.keyWindow
+            guard let viewController = hostWindow?.contentViewController as? ViewController else { return }
+            viewController.presentOCRFile(imageURL: url)
+        }
+
+        if let window = window ?? NSApp.keyWindow {
+            panel.beginSheetModal(for: window, completionHandler: completion)
+        } else if panel.runModal() == .OK {
+            completion(.OK)
+        }
+    }
+
+    static func recognizeText(from url: URL) async throws -> String {
+        let accessed = url.startAccessingSecurityScopedResource()
+        defer {
+            if accessed { url.stopAccessingSecurityScopedResource() }
+        }
+
+        guard let image = NSImage(contentsOf: url),
+              let cgImage = image.cgImage(forProposedRect: nil, context: nil, hints: nil) else {
+            throw OCRError.invalidImage
+        }
+
+        return try await recognizeText(in: cgImage)
+    }
+
+    private static func recognizeText(in cgImage: CGImage) async throws -> String {
+        try await withCheckedThrowingContinuation { continuation in
+            let request = VNRecognizeTextRequest { request, error in
+                if let error {
+                    continuation.resume(throwing: error)
+                    return
+                }
+                guard let observations = request.results as? [VNRecognizedTextObservation] else {
+                    continuation.resume(throwing: OCRError.recognitionFailed)
+                    return
+                }
+
+                let lines = observations.compactMap { $0.topCandidates(1).first?.string }
+                let text = lines.joined(separator: "\n").trimmingCharacters(in: .whitespacesAndNewlines)
+                guard !text.isEmpty else {
+                    continuation.resume(throwing: OCRError.noTextFound)
+                    return
+                }
+                continuation.resume(returning: text)
+            }
+            request.recognitionLevel = .accurate
+            request.usesLanguageCorrection = true
+
+            let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+            do {
+                try handler.perform([request])
+            } catch {
+                continuation.resume(throwing: error)
+            }
+        }
+    }
+}
+
+// MARK: - Overlay
+
+final class OCRFileOverlayView: NSView, AppearanceRefreshable {
+    var onDismiss: (() -> Void)?
+
+    private let imageURL: URL
+    private let backButton = OCRFileToolbarButton(symbolName: "chevron.left", accessibilityLabel: "Back")
+    private let titleLabel = NSTextField(labelWithString: "OCR File")
+    private let textContainer = NSView()
+    private let textScrollView = NSScrollView()
+    private let textView = NSTextView()
+    private let loadingIndicator = NSProgressIndicator()
+    private let statusLabel = NSTextField(labelWithString: "Extracting text…")
+    private let printButton = OCRFilePrintButton()
+
+    init(imageURL: URL) {
+        self.imageURL = imageURL
+        super.init(frame: .zero)
+        translatesAutoresizingMaskIntoConstraints = false
+        setup()
+        refreshAppearance()
+        NotificationCenter.default.addObserver(
+            self,
+            selector: #selector(appearanceDidChange),
+            name: .appearanceDidChange,
+            object: nil
+        )
+        recognizeText()
+    }
+
+    deinit {
+        NotificationCenter.default.removeObserver(self)
+    }
+
+    @objc private func appearanceDidChange() {
+        refreshAppearance()
+    }
+
+    @available(*, unavailable)
+    required init?(coder: NSCoder) { nil }
+
+    func refreshAppearance() {
+        layer?.backgroundColor = AppTheme.background.cgColor
+        titleLabel.textColor = AppTheme.textPrimary
+        textContainer.layer?.backgroundColor = AppTheme.cardBackground.cgColor
+        textContainer.layer?.borderColor = AppTheme.paywallBorder.cgColor
+        statusLabel.textColor = AppTheme.textSecondary
+        backButton.refreshAppearance()
+        printButton.refreshAppearance()
+    }
+
+    func present(in parent: NSView) {
+        guard superview == nil else { return }
+        parent.addSubview(self)
+        NSLayoutConstraint.activate([
+            leadingAnchor.constraint(equalTo: parent.leadingAnchor),
+            trailingAnchor.constraint(equalTo: parent.trailingAnchor),
+            topAnchor.constraint(equalTo: parent.topAnchor),
+            bottomAnchor.constraint(equalTo: parent.bottomAnchor),
+        ])
+        alphaValue = 0
+        NSAnimationContext.runAnimationGroup { context in
+            context.duration = 0.2
+            animator().alphaValue = 1
+        }
+    }
+
+    func dismiss(animated: Bool = true) {
+        let remove = { [weak self] in
+            self?.removeFromSuperview()
+            self?.onDismiss?()
+        }
+        guard animated else {
+            remove()
+            return
+        }
+        NSAnimationContext.runAnimationGroup({ context in
+            context.duration = 0.15
+            animator().alphaValue = 0
+        }, completionHandler: remove)
+    }
+
+    private func setup() {
+        wantsLayer = true
+
+        titleLabel.font = AppTheme.semiboldFont(size: 18)
+        titleLabel.alignment = .center
+        titleLabel.translatesAutoresizingMaskIntoConstraints = false
+
+        textContainer.wantsLayer = true
+        textContainer.layer?.cornerRadius = AppTheme.contentPanelCornerRadius
+        textContainer.layer?.borderWidth = 1.5
+        textContainer.applyCardShadow()
+        textContainer.translatesAutoresizingMaskIntoConstraints = false
+
+        textScrollView.translatesAutoresizingMaskIntoConstraints = false
+        textScrollView.hasVerticalScroller = true
+        textScrollView.hasHorizontalScroller = false
+        textScrollView.autohidesScrollers = true
+        textScrollView.borderType = .noBorder
+        textScrollView.drawsBackground = false
+
+        textView.isRichText = false
+        textView.isEditable = true
+        textView.isSelectable = true
+        textView.drawsBackground = false
+        textView.textColor = AppTheme.textPrimary
+        textView.font = AppTheme.regularFont(size: 14)
+        textView.textContainerInset = NSSize(width: 16, height: 16)
+        textView.isAutomaticQuoteSubstitutionEnabled = true
+        textView.isAutomaticDashSubstitutionEnabled = true
+        textView.isAutomaticTextReplacementEnabled = true
+        textView.isHidden = true
+
+        textScrollView.documentView = textView
+
+        loadingIndicator.style = .spinning
+        loadingIndicator.controlSize = .regular
+        loadingIndicator.translatesAutoresizingMaskIntoConstraints = false
+        loadingIndicator.startAnimation(nil)
+
+        statusLabel.font = AppTheme.regularFont(size: 14)
+        statusLabel.alignment = .center
+        statusLabel.translatesAutoresizingMaskIntoConstraints = false
+
+        backButton.translatesAutoresizingMaskIntoConstraints = false
+        printButton.translatesAutoresizingMaskIntoConstraints = false
+        printButton.isEnabled = false
+
+        backButton.onClick = { [weak self] in self?.dismiss() }
+        printButton.onClick = { [weak self] in self?.printText() }
+
+        addSubview(backButton)
+        addSubview(titleLabel)
+        addSubview(textContainer)
+        textContainer.addSubview(textScrollView)
+        textContainer.addSubview(loadingIndicator)
+        textContainer.addSubview(statusLabel)
+        addSubview(printButton)
+
+        NSLayoutConstraint.activate([
+            backButton.leadingAnchor.constraint(equalTo: leadingAnchor, constant: 24),
+            backButton.topAnchor.constraint(equalTo: topAnchor, constant: 20),
+            backButton.widthAnchor.constraint(equalToConstant: 40),
+            backButton.heightAnchor.constraint(equalToConstant: 40),
+
+            titleLabel.centerXAnchor.constraint(equalTo: centerXAnchor),
+            titleLabel.centerYAnchor.constraint(equalTo: backButton.centerYAnchor),
+
+            textContainer.leadingAnchor.constraint(equalTo: leadingAnchor, constant: AppTheme.contentPanelInset),
+            textContainer.trailingAnchor.constraint(equalTo: trailingAnchor, constant: -AppTheme.contentPanelInset),
+            textContainer.topAnchor.constraint(equalTo: backButton.bottomAnchor, constant: 28),
+            textContainer.bottomAnchor.constraint(equalTo: printButton.topAnchor, constant: -28),
+
+            textScrollView.leadingAnchor.constraint(equalTo: textContainer.leadingAnchor, constant: 12),
+            textScrollView.trailingAnchor.constraint(equalTo: textContainer.trailingAnchor, constant: -12),
+            textScrollView.topAnchor.constraint(equalTo: textContainer.topAnchor, constant: 12),
+            textScrollView.bottomAnchor.constraint(equalTo: textContainer.bottomAnchor, constant: -12),
+
+            loadingIndicator.centerXAnchor.constraint(equalTo: textContainer.centerXAnchor),
+            loadingIndicator.centerYAnchor.constraint(equalTo: textContainer.centerYAnchor, constant: -12),
+
+            statusLabel.topAnchor.constraint(equalTo: loadingIndicator.bottomAnchor, constant: 12),
+            statusLabel.leadingAnchor.constraint(equalTo: textContainer.leadingAnchor, constant: 24),
+            statusLabel.trailingAnchor.constraint(equalTo: textContainer.trailingAnchor, constant: -24),
+
+            printButton.centerXAnchor.constraint(equalTo: centerXAnchor),
+            printButton.bottomAnchor.constraint(equalTo: bottomAnchor, constant: -28),
+            printButton.widthAnchor.constraint(greaterThanOrEqualToConstant: 220),
+        ])
+    }
+
+    private func recognizeText() {
+        Task { @MainActor in
+            do {
+                let text = try await OCRFileService.recognizeText(from: imageURL)
+                showRecognizedText(text)
+            } catch {
+                showRecognitionError(error)
+            }
+        }
+    }
+
+    private func showRecognizedText(_ text: String) {
+        loadingIndicator.stopAnimation(nil)
+        loadingIndicator.isHidden = true
+        statusLabel.isHidden = true
+        textView.string = text
+        textView.isHidden = false
+        printButton.isEnabled = true
+        window?.makeFirstResponder(textView)
+    }
+
+    private func showRecognitionError(_ error: Error) {
+        loadingIndicator.stopAnimation(nil)
+        loadingIndicator.isHidden = true
+        statusLabel.isHidden = true
+
+        let alert = NSAlert()
+        alert.messageText = "OCR Failed"
+        alert.informativeText = error.localizedDescription
+        alert.alertStyle = .warning
+        alert.addButton(withTitle: "OK")
+        alert.beginSheetModal(for: window ?? NSApp.keyWindow ?? NSWindow()) { [weak self] _ in
+            self?.dismiss()
+        }
+    }
+
+    private func printText() {
+        PrintService.printText(textView.string, title: "OCR File", from: window)
+    }
+}
+
+// MARK: - Toolbar Button
+
+private final class OCRFileToolbarButton: NSControl, AppearanceRefreshable {
+    var onClick: (() -> Void)?
+
+    private let symbolName: String
+    private let iconView = NSImageView()
+
+    init(symbolName: String, accessibilityLabel: String) {
+        self.symbolName = symbolName
+        super.init(frame: .zero)
+        toolTip = accessibilityLabel
+
+        wantsLayer = true
+        layer?.cornerRadius = 12
+
+        if let image = NSImage(systemSymbolName: symbolName, accessibilityDescription: accessibilityLabel) {
+            let config = NSImage.SymbolConfiguration(pointSize: 15, weight: .semibold)
+            iconView.image = image.withSymbolConfiguration(config)
+        }
+        iconView.translatesAutoresizingMaskIntoConstraints = false
+
+        addSubview(iconView)
+        NSLayoutConstraint.activate([
+            iconView.centerXAnchor.constraint(equalTo: centerXAnchor),
+            iconView.centerYAnchor.constraint(equalTo: centerYAnchor),
+        ])
+        refreshAppearance()
+    }
+
+    @available(*, unavailable)
+    required init?(coder: NSCoder) { nil }
+
+    func refreshAppearance() {
+        layer?.backgroundColor = AppTheme.elevatedBackground.cgColor
+        layer?.borderColor = AppTheme.paywallBorder.cgColor
+        layer?.borderWidth = 1.5
+        iconView.contentTintColor = AppTheme.textPrimary
+    }
+
+    override func mouseUp(with event: NSEvent) {
+        guard bounds.contains(convert(event.locationInWindow, from: nil)) else { return }
+        onClick?()
+    }
+
+    override func resetCursorRects() {
+        addCursorRect(bounds, cursor: .pointingHand)
+    }
+}
+
+// MARK: - Print Button
+
+private final class OCRFilePrintButton: NSControl, AppearanceRefreshable {
+    var onClick: (() -> Void)?
+
+    private let titleLabel = NSTextField(labelWithString: "Print Text")
+    private let iconView = NSImageView()
+    private var hoverTracker: HoverTracker?
+    private var isHovered = false
+    private var isEnabledState = true
+
+    override var isEnabled: Bool {
+        get { isEnabledState }
+        set {
+            isEnabledState = newValue
+            alphaValue = newValue ? 1 : 0.45
+        }
+    }
+
+    init() {
+        super.init(frame: .zero)
+        wantsLayer = true
+        layer?.cornerRadius = 22
+
+        titleLabel.font = AppTheme.semiboldFont(size: 16)
+        titleLabel.textColor = .white
+        titleLabel.translatesAutoresizingMaskIntoConstraints = false
+
+        if let image = NSImage(systemSymbolName: "printer.fill", accessibilityDescription: "Print") {
+            let config = NSImage.SymbolConfiguration(pointSize: 14, weight: .semibold)
+            iconView.image = image.withSymbolConfiguration(config)
+        }
+        iconView.contentTintColor = .white
+        iconView.translatesAutoresizingMaskIntoConstraints = false
+
+        addSubview(titleLabel)
+        addSubview(iconView)
+
+        NSLayoutConstraint.activate([
+            heightAnchor.constraint(equalToConstant: 52),
+
+            titleLabel.leadingAnchor.constraint(equalTo: leadingAnchor, constant: 28),
+            titleLabel.centerYAnchor.constraint(equalTo: centerYAnchor),
+
+            iconView.leadingAnchor.constraint(equalTo: titleLabel.trailingAnchor, constant: 10),
+            iconView.centerYAnchor.constraint(equalTo: centerYAnchor),
+            iconView.trailingAnchor.constraint(equalTo: trailingAnchor, constant: -28),
+            iconView.widthAnchor.constraint(equalToConstant: 20),
+            iconView.heightAnchor.constraint(equalToConstant: 20),
+        ])
+
+        hoverTracker = HoverTracker(view: self) { [weak self] hovering in
+            self?.setHovered(hovering)
+        }
+        refreshAppearance()
+    }
+
+    @available(*, unavailable)
+    required init?(coder: NSCoder) { nil }
+
+    func refreshAppearance() {
+        let color = isHovered
+            ? AppTheme.blue.blended(withFraction: 0.15, of: .black) ?? AppTheme.blue
+            : AppTheme.blue
+        layer?.backgroundColor = color.cgColor
+        titleLabel.textColor = .white
+        iconView.contentTintColor = .white
+    }
+
+    private func setHovered(_ hovering: Bool) {
+        guard isEnabledState else { return }
+        isHovered = hovering
+        animateHover {
+            let color = hovering
+                ? AppTheme.blue.blended(withFraction: 0.15, of: .black) ?? AppTheme.blue
+                : AppTheme.blue
+            layer?.backgroundColor = color.cgColor
+            layer?.transform = hovering
+                ? CATransform3DMakeScale(1.03, 1.03, 1)
+                : CATransform3DIdentity
+        }
+    }
+
+    override func mouseUp(with event: NSEvent) {
+        guard isEnabledState else { return }
+        guard bounds.contains(convert(event.locationInWindow, from: nil)) else { return }
+        onClick?()
+    }
+
+    override func resetCursorRects() {
+        guard isEnabledState else { return }
+        addCursorRect(bounds, cursor: .pointingHand)
+    }
+}

smart_printer/ViewController.swift

@@ -20,6 +20,7 @@ class ViewController: NSViewController {
     private var printTextOverlay: PrintTextOverlayView?
     private var printContactsOverlay: PrintContactsOverlayView?
     private var drawPrintOverlay: DrawPrintOverlayView?
+    private var ocrFileOverlay: OCRFileOverlayView?
 
     private var headerView: NSView!
     private var contentTopBelowHeader: NSLayoutConstraint!
@@ -86,6 +87,7 @@ class ViewController: NSViewController {
         printTextOverlay?.refreshAppearance()
         printContactsOverlay?.refreshAppearance()
         drawPrintOverlay?.refreshAppearance()
+        ocrFileOverlay?.refreshAppearance()
     }
 
     func presentPrintText() {
@@ -93,6 +95,7 @@ class ViewController: NSViewController {
         filePreviewOverlay?.dismiss(animated: false)
         printContactsOverlay?.dismiss(animated: false)
         drawPrintOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
 
         let overlay = PrintTextOverlayView()
         overlay.onDismiss = { [weak self] in
@@ -107,6 +110,7 @@ class ViewController: NSViewController {
         filePreviewOverlay?.dismiss(animated: false)
         printTextOverlay?.dismiss(animated: false)
         drawPrintOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
 
         if let overlay = printContactsOverlay {
             overlay.present(in: mainContentView)
@@ -118,11 +122,27 @@ class ViewController: NSViewController {
         overlay.present(in: mainContentView)
     }
 
+    func presentOCRFile(imageURL: URL) {
+        photoPreviewOverlay?.dismiss(animated: false)
+        filePreviewOverlay?.dismiss(animated: false)
+        printTextOverlay?.dismiss(animated: false)
+        printContactsOverlay?.dismiss(animated: false)
+        drawPrintOverlay?.dismiss(animated: false)
+
+        let overlay = OCRFileOverlayView(imageURL: imageURL)
+        overlay.onDismiss = { [weak self] in
+            self?.ocrFileOverlay = nil
+        }
+        ocrFileOverlay = overlay
+        overlay.present(in: mainContentView)
+    }
+
     func presentDrawPrint() {
         photoPreviewOverlay?.dismiss(animated: false)
         filePreviewOverlay?.dismiss(animated: false)
         printTextOverlay?.dismiss(animated: false)
         printContactsOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
 
         let overlay = DrawPrintOverlayView()
         overlay.onDismiss = { [weak self] in
@@ -135,6 +155,7 @@ class ViewController: NSViewController {
     func presentPhotoPreview(urls: [URL]) {
         photoPreviewOverlay?.dismiss(animated: false)
         filePreviewOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
 
         let overlay = PhotoPreviewOverlayView(urls: urls)
         overlay.onDismiss = { [weak self] in
@@ -147,6 +168,7 @@ class ViewController: NSViewController {
     func presentFilePreview(urls: [URL]) {
         photoPreviewOverlay?.dismiss(animated: false)
         filePreviewOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
 
         let overlay = FilePreviewOverlayView(urls: urls)
         overlay.onDismiss = { [weak self] in
@@ -238,6 +260,7 @@ class ViewController: NSViewController {
         printTextOverlay?.dismiss(animated: false)
         printContactsOverlay?.dismiss(animated: false)
         drawPrintOverlay?.dismiss(animated: false)
+        ocrFileOverlay?.dismiss(animated: false)
     }
 
     private func showDestination(_ destination: SidebarDestination) {
@@ -324,10 +347,10 @@ class ViewController: NSViewController {
     private func makeScanContentView() -> NSView {
         makeDestinationScrollView(
             title: "Scan",
-            features: [
+            features: featureCards(
                 FeatureCardData(title: "Scan File", subtitle: "Scan any document", iconKind: .scanFile),
-                FeatureCardData(title: "OCR File", subtitle: "Scan and print text from images", iconKind: .ocrFile),
-            ]
+                FeatureCardData(title: "OCR File", subtitle: "Scan and print text from images", iconKind: .ocrFile)
+            )
         )
     }
 
@@ -644,6 +667,10 @@ class ViewController: NSViewController {
                 updated.onActivate = { [weak self] in
                     DrawPrintService.present(from: self?.view.window)
                 }
+            } else if card.iconKind == .ocrFile {
+                updated.onActivate = { [weak self] in
+                    OCRFileService.present(from: self?.view.window)
+                }
             }
             return updated
         }