@sudobility/sudojo_ocr

Cross-platform OCR library for extracting Sudoku puzzles from images. Works in browser, React Native, and Node.js.

Installation

bun add @sudobility/sudojo_ocr

Usage

// Web
import { extractSudokuFromImage } from '@sudobility/sudojo_ocr';
import { createWebAdapter } from '@sudobility/sudojo_ocr/web';
import Tesseract from 'tesseract.js';

const adapter = createWebAdapter();
const result = await extractSudokuFromImage(adapter, imageFile, Tesseract);
console.log(result.puzzle);      // "004002008100400000..."
console.log(result.confidence);  // 85.5

// Node.js
import { createNodeAdapter } from '@sudobility/sudojo_ocr/node';
const adapter = await createNodeAdapter();
const result = await extractSudokuFromImage(adapter, imageBuffer, Tesseract);

API

• extractSudokuFromImage() -- Full OCR pipeline: image -> board detection -> cell extraction -> OCR -> digit parsing
• detectAndCropBoard() -- Detect and crop board, returns data URL
• extractCellImages() -- Extract 81 cell images as data URLs
• Platform adapters: createWebAdapter(), createNodeAdapter()

Development

bun run build        # Build ESM + CJS
bun run test         # Run Vitest
bun run typecheck    # TypeScript check
bun run lint         # ESLint
bun run verify       # Typecheck + lint + test + build

Related Packages

• sudojo_app -- Web application (uses web adapter)
• sudojo_bot -- Bot application (uses Node.js adapter)
• @sudobility/sudojo_types -- Solver types for hint data

License

BUSL-1.1