macos-vision 1.1.0 → 1.3.0

package/src/native/pdf-helper.swift

@@ -0,0 +1,122 @@
+import PDFKit
+import AppKit
+import Foundation
+
+// ─── Result struct ────────────────────────────────────────────────────────────
+
+struct PageResult: Codable {
+    let page: Int    // 0-based
+    let path: String
+}
+
+// ─── Helpers ──────────────────────────────────────────────────────────────────
+
+func fail(_ message: String) -> Never {
+    fputs("ERROR: \(message)\n", stderr)
+    exit(1)
+}
+
+func encodeJSON<T: Encodable>(_ value: T) -> String {
+    guard let data = try? JSONEncoder().encode(value),
+          let str = String(data: data, encoding: .utf8) else { return "[]" }
+    return str
+}
+
+// ─── Argument parsing ─────────────────────────────────────────────────────────
+
+let args = CommandLine.arguments
+guard args.count >= 2 else {
+    fail("Usage: pdf-helper <path-to-pdf>")
+}
+
+let pdfPath = args[1]
+let pdfURL = URL(fileURLWithPath: pdfPath)
+
+guard let pdf = PDFDocument(url: pdfURL) else {
+    fail("Cannot open PDF: \(pdfPath)")
+}
+
+let pageCount = pdf.pageCount
+guard pageCount > 0 else {
+    fail("PDF has no pages: \(pdfPath)")
+}
+
+// ─── Output directory: ~/.cache/macos-vision/{basename}-{uuid}/ ───────────────
+
+let basename = pdfURL.deletingPathExtension().lastPathComponent
+let uuid = UUID().uuidString.lowercased()
+let cacheBase = FileManager.default.homeDirectoryForCurrentUser
+    .appendingPathComponent(".cache/macos-vision")
+let outDir = cacheBase.appendingPathComponent("\(basename)-\(uuid)")
+
+do {
+    try FileManager.default.createDirectory(at: outDir, withIntermediateDirectories: true)
+} catch {
+    fail("Cannot create output directory \(outDir.path): \(error.localizedDescription)")
+}
+
+// ─── Rasterize each page at 300 DPI ──────────────────────────────────────────
+
+// PDF points are 72 pt/inch. Scale factor for 300 DPI = 300/72 ≈ 4.167
+let scale: CGFloat = 300.0 / 72.0
+
+var results: [PageResult] = []
+
+for pageIndex in 0..<pageCount {
+    guard let page = pdf.page(at: pageIndex) else {
+        fail("Cannot access page \(pageIndex) of \(pdfPath)")
+    }
+
+    let mediaBox = page.bounds(for: .mediaBox)
+    let width  = Int((mediaBox.width  * scale).rounded())
+    let height = Int((mediaBox.height * scale).rounded())
+
+    guard let bitmapRep = NSBitmapImageRep(
+        bitmapDataPlanes: nil,
+        pixelsWide: width,
+        pixelsHigh: height,
+        bitsPerSample: 8,
+        samplesPerPixel: 4,
+        hasAlpha: true,
+        isPlanar: false,
+        colorSpaceName: .calibratedRGB,
+        bytesPerRow: 0,
+        bitsPerPixel: 0
+    ) else {
+        fail("Cannot create bitmap for page \(pageIndex)")
+    }
+
+    guard let ctx = NSGraphicsContext(bitmapImageRep: bitmapRep) else {
+        fail("Cannot create graphics context for page \(pageIndex)")
+    }
+
+    // Fill white background (PDFs are transparent by default)
+    NSGraphicsContext.saveGraphicsState()
+    NSGraphicsContext.current = ctx
+    NSColor.white.setFill()
+    NSRect(x: 0, y: 0, width: width, height: height).fill()
+
+    ctx.cgContext.scaleBy(x: scale, y: scale)
+    page.draw(with: .mediaBox, to: ctx.cgContext)
+    NSGraphicsContext.restoreGraphicsState()
+
+    guard let pngData = bitmapRep.representation(using: .png, properties: [:]) else {
+        fail("Cannot encode page \(pageIndex) to PNG")
+    }
+
+    // Zero-pad page number to 3 digits: page-001.png, page-002.png, …
+    let filename = String(format: "%@-page-%03d.png", basename, pageIndex + 1)
+    let outPath = outDir.appendingPathComponent(filename)
+
+    do {
+        try pngData.write(to: outPath)
+    } catch {
+        fail("Cannot write \(outPath.path): \(error.localizedDescription)")
+    }
+
+    results.append(PageResult(page: pageIndex, path: outPath.path))
+}
+
+// ─── Output JSON ──────────────────────────────────────────────────────────────
+
+print(encodeJSON(results))

package/src/native/vision-helper.swift

@@ -0,0 +1,241 @@
+import Vision
+import AppKit
+import Foundation
+
+// ─── Result structs ──────────────────────────────────────────────────────────
+
+struct OCRResult: Codable {
+    let t: String
+    let x: Double; let y: Double; let w: Double; let h: Double
+    let confidence: Float
+}
+
+struct FaceResult: Codable {
+    let x: Double; let y: Double; let w: Double; let h: Double
+    let confidence: Float
+}
+
+struct BarcodeResult: Codable {
+    let type: String
+    let value: String
+    let x: Double; let y: Double; let w: Double; let h: Double
+    let confidence: Float
+}
+
+struct RectangleResult: Codable {
+    let topLeft: [Double]; let topRight: [Double]
+    let bottomLeft: [Double]; let bottomRight: [Double]
+    let confidence: Float
+}
+
+struct DocumentResult: Codable {
+    let topLeft: [Double]; let topRight: [Double]
+    let bottomLeft: [Double]; let bottomRight: [Double]
+    let confidence: Float
+}
+
+struct ClassificationResult: Codable {
+    let identifier: String
+    let confidence: Float
+}
+
+// ─── Helpers ─────────────────────────────────────────────────────────────────
+
+// Vision: 0,0 = bottom-left. We flip Y so 0,0 = top-left (web standard).
+func flipY(_ y: Double, _ h: Double) -> Double { 1.0 - y - h }
+
+func pt(_ p: CGPoint) -> [Double] { [Double(p.x), 1.0 - Double(p.y)] }
+
+func encodeJSON<T: Encodable>(_ value: T) -> String {
+    guard let data = try? JSONEncoder().encode(value),
+          let str = String(data: data, encoding: .utf8) else { return "[]" }
+    return str
+}
+
+// ─── Argument parsing ─────────────────────────────────────────────────────────
+
+let args = CommandLine.arguments
+let isJsonMode   = args.contains("--json")
+let isFaces      = args.contains("--faces")
+let isBarcodes   = args.contains("--barcodes")
+let isRectangles = args.contains("--rectangles")
+let isDocument   = args.contains("--document")
+let isClassify   = args.contains("--classify")
+
+let fileArgs = args.filter { !$0.hasPrefix("--") && !$0.contains("vision-helper") }
+
+guard let imagePath = fileArgs.first else {
+    print("Usage: vision-helper [--json|--faces|--barcodes|--rectangles|--document|--classify] <path>")
+    exit(0)
+}
+
+guard let image = NSImage(contentsOf: URL(fileURLWithPath: imagePath)),
+      let cgImage = image.cgImage(forProposedRect: nil, context: nil, hints: nil) else {
+    fputs("ERROR: Cannot open file: \(imagePath)\n", stderr)
+    exit(1)
+}
+
+let handler = VNImageRequestHandler(cgImage: cgImage, options: [:])
+
+// ─── OCR (default + --json) ───────────────────────────────────────────────────
+
+if isJsonMode || (!isFaces && !isBarcodes && !isRectangles && !isDocument && !isClassify) {
+    var ocrResults: [OCRResult] = []
+    var rawText = ""
+
+    let request = VNRecognizeTextRequest { (req, _) in
+        guard let obs = req.results as? [VNRecognizedTextObservation] else { return }
+        for o in obs {
+            guard let c = o.topCandidates(1).first else { continue }
+            let box = o.boundingBox
+            if isJsonMode {
+                ocrResults.append(OCRResult(
+                    t: c.string,
+                    x: Double(box.origin.x),
+                    y: flipY(Double(box.origin.y), Double(box.size.height)),
+                    w: Double(box.size.width),
+                    h: Double(box.size.height),
+                    confidence: c.confidence
+                ))
+            } else {
+                rawText += c.string + "\n"
+            }
+        }
+    }
+    request.recognitionLevel = .accurate
+
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision OCR failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(isJsonMode ? encodeJSON(ocrResults) : rawText.trimmingCharacters(in: .whitespacesAndNewlines))
+    exit(0)
+}
+
+// ─── Faces ───────────────────────────────────────────────────────────────────
+
+if isFaces {
+    var results: [FaceResult] = []
+    let request = VNDetectFaceRectanglesRequest { (req, _) in
+        guard let obs = req.results as? [VNFaceObservation] else { return }
+        for o in obs {
+            let box = o.boundingBox
+            results.append(FaceResult(
+                x: Double(box.origin.x),
+                y: flipY(Double(box.origin.y), Double(box.size.height)),
+                w: Double(box.size.width),
+                h: Double(box.size.height),
+                confidence: o.confidence
+            ))
+        }
+    }
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision face detection failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(encodeJSON(results))
+    exit(0)
+}
+
+// ─── Barcodes ────────────────────────────────────────────────────────────────
+
+if isBarcodes {
+    var results: [BarcodeResult] = []
+    let request = VNDetectBarcodesRequest { (req, _) in
+        guard let obs = req.results as? [VNBarcodeObservation] else { return }
+        for o in obs {
+            let box = o.boundingBox
+            results.append(BarcodeResult(
+                type: o.symbology.rawValue,
+                value: o.payloadStringValue ?? "",
+                x: Double(box.origin.x),
+                y: flipY(Double(box.origin.y), Double(box.size.height)),
+                w: Double(box.size.width),
+                h: Double(box.size.height),
+                confidence: o.confidence
+            ))
+        }
+    }
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision barcode detection failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(encodeJSON(results))
+    exit(0)
+}
+
+// ─── Rectangles ──────────────────────────────────────────────────────────────
+
+if isRectangles {
+    var results: [RectangleResult] = []
+    let request = VNDetectRectanglesRequest { (req, _) in
+        guard let obs = req.results as? [VNRectangleObservation] else { return }
+        for o in obs {
+            results.append(RectangleResult(
+                topLeft: pt(o.topLeft), topRight: pt(o.topRight),
+                bottomLeft: pt(o.bottomLeft), bottomRight: pt(o.bottomRight),
+                confidence: o.confidence
+            ))
+        }
+    }
+    (request as VNDetectRectanglesRequest).maximumObservations = 0
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision rectangle detection failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(encodeJSON(results))
+    exit(0)
+}
+
+// ─── Document ────────────────────────────────────────────────────────────────
+
+if isDocument {
+    var results: [DocumentResult] = []
+    let request = VNDetectDocumentSegmentationRequest { (req, _) in
+        guard let obs = req.results as? [VNRectangleObservation] else { return }
+        for o in obs {
+            results.append(DocumentResult(
+                topLeft: pt(o.topLeft), topRight: pt(o.topRight),
+                bottomLeft: pt(o.bottomLeft), bottomRight: pt(o.bottomRight),
+                confidence: o.confidence
+            ))
+        }
+    }
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision document detection failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(encodeJSON(results))
+    exit(0)
+}
+
+// ─── Classify ────────────────────────────────────────────────────────────────
+
+if isClassify {
+    var results: [ClassificationResult] = []
+    let request = VNClassifyImageRequest { (req, _) in
+        guard let obs = req.results as? [VNClassificationObservation] else { return }
+        let top = obs.filter { $0.confidence > 0.01 }.prefix(10)
+        for o in top {
+            results.append(ClassificationResult(identifier: o.identifier, confidence: o.confidence))
+        }
+    }
+    do {
+        try handler.perform([request])
+    } catch {
+        fputs("ERROR: Vision classification failed: \(error.localizedDescription)\n", stderr)
+        exit(1)
+    }
+    print(encodeJSON(results))
+    exit(0)
+}

package/.husky/commit-msg

@@ -1,2 +0,0 @@
-#!/bin/sh
-npx --no -- commitlint --edit $1

package/.husky/pre-commit

@@ -1,3 +0,0 @@
-#!/bin/sh
-npx lint-staged
-npx tsc --noEmit

package/.prettierignore

@@ -1,4 +0,0 @@
-dist/
-node_modules/
-bin/
-*.md

package/.prettierrc.json

@@ -1,7 +0,0 @@
-{
-  "semi": true,
-  "singleQuote": true,
-  "trailingComma": "es5",
-  "printWidth": 100,
-  "tabWidth": 2
-}

package/.release-it.json

@@ -1,20 +0,0 @@
-{
-  "$schema": "https://unpkg.com/release-it/schema/release-it.json",
-  "plugins": {
-    "@release-it/conventional-changelog": {
-      "preset": "conventionalcommits",
-      "infile": "CHANGELOG.md"
-    }
-  },
-  "git": {
-    "commitMessage": "chore(release): v${version}",
-    "tagName": "v${version}"
-  },
-  "github": {
-    "release": false
-  },
-  "npm": {
-    "publish": true,
-    "tag": "latest"
-  }
-}

package/CHANGELOG.md

@@ -1,38 +0,0 @@
-# Changelog
-
-## [1.1.0](https://github.com/woladi/macos-vision/compare/v1.0.3...v1.1.0) (2026-04-09)
-
-### Features
-
-* add PDF support via sips rasterization ([a48bf17](https://github.com/woladi/macos-vision/commit/a48bf17579a6df11aed6eadbde4fa5041ccaa981))
-
-## [1.0.3](https://github.com/woladi/macos-vision/compare/v1.0.2...v1.0.3) (2026-04-08)
-
-### Reverts
-
-* remove socket.ignore field — worsens supply chain risk score ([a1827ad](https://github.com/woladi/macos-vision/commit/a1827ad489220ebb7a2e8c85632945fe969438db))
-
-## [1.0.2](https://github.com/woladi/macos-vision/compare/v1.0.1...v1.0.2) (2026-04-08)
-
-## [1.0.1](https://github.com/woladi/macos-vision/compare/v0.3.1...v1.0.1) (2026-04-08)
-
-## [0.3.1](https://github.com/woladi/macos-vision/compare/v0.3.0...v0.3.1) (2026-04-08)
-
-## [0.3.0](https://github.com/woladi/macos-vision/compare/v0.2.0...v0.3.0) (2026-04-08)
-
-### Features
-
-* add inferLayout() — unified reading-order LayoutBlock representation ([aec507e](https://github.com/woladi/macos-vision/commit/aec507eb7cf133ec1e56759c0945563a48d871ee))
-
-## [0.2.0](https://github.com/woladi/macos-vision/compare/v0.1.4...v0.2.0) (2026-04-08)
-
-### Features
-
-* add confidence to VisionBlock and Barcode ([a87df27](https://github.com/woladi/macos-vision/commit/a87df275e51dec4b57fbff6e3bffc4220b96b4d7))
-
-### Bug Fixes
-
-* correct mkdirSync, CLI error on missing file, execFile timeout, README scope ([1cef2c7](https://github.com/woladi/macos-vision/commit/1cef2c7078430c9182fcd39792cf0c002833203f))
-* replace try? with do/catch in Swift helper — surface Vision errors properly ([f287065](https://github.com/woladi/macos-vision/commit/f2870655225806070be3db462ea15923201fecbf))
-
-## 0.1.4 (2026-04-08)

package/commitlint.config.js

@@ -1 +0,0 @@
-export default { extends: ['@commitlint/config-conventional'] };

package/debug.js

@@ -1,37 +0,0 @@
-#!/usr/bin/env node
-import {
-  ocr,
-  detectFaces,
-  detectBarcodes,
-  detectRectangles,
-  detectDocument,
-  classify,
-  inferLayout,
-} from './dist/index.js';
-
-const imagePath = process.argv[2] || './test/fixtures/sample.png';
-console.log(`\n📸 Analyzing: ${imagePath}\n`);
-
-const [text, blocks, faces, barcodes, rects, doc, labels] = await Promise.all([
-  ocr(imagePath),
-  ocr(imagePath, { format: 'blocks' }),
-  detectFaces(imagePath),
-  detectBarcodes(imagePath),
-  detectRectangles(imagePath),
-  detectDocument(imagePath),
-  classify(imagePath),
-]);
-
-const sep = (title) => console.log('\n' + '─'.repeat(60) + '\n' + title + '\n');
-
-sep('📝  OCR text');          console.log(text);
-sep('📝  OCR blocks');       console.log(JSON.stringify(blocks, null, 2));
-sep('👤  Faces');            console.log(JSON.stringify(faces, null, 2));
-sep('🔲  Barcodes');         console.log(JSON.stringify(barcodes, null, 2));
-sep('📦  Rectangles');       console.log(JSON.stringify(rects, null, 2));
-sep('📄  Document');         console.log(JSON.stringify(doc, null, 2));
-sep('🏷️   Classification');  console.log(JSON.stringify(labels, null, 2));
-
-const layout = inferLayout({ textBlocks: blocks, faces, barcodes, rectangles: rects, document: doc });
-sep('🗂️   Layout (reading order)'); console.log(JSON.stringify(layout, null, 2));
-console.log('\n' + '─'.repeat(60) + '\n');

package/eslint.config.js

@@ -1,21 +0,0 @@
-import tseslint from 'typescript-eslint';
-import prettier from 'eslint-config-prettier';
-
-export default tseslint.config(
-  ...tseslint.configs.recommended,
-  prettier,
-  {
-    files: ['src/**/*.ts'],
-    languageOptions: {
-      parser: tseslint.parser,
-      parserOptions: {
-        project: true, // Szuka najbliższego tsconfig.json
-      },
-    },
-    rules: {
-      '@typescript-eslint/no-explicit-any': 'warn',
-      '@typescript-eslint/no-unused-vars': ['error', { argsIgnorePattern: '^_' }],
-    },
-  },
-  { ignores: ['dist/**', 'node_modules/**', 'bin/**'] }
-);