@kreuzberg/node 4.2.14 → 4.3.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kreuzberg/node",
-  "version": "4.2.14",
+  "version": "4.3.0",
   "description": "Kreuzberg document intelligence - Node.js native bindings",
   "author": {
     "name": "Na'aman Hirschfeld",
@@ -71,8 +71,10 @@
     "targets": [
       "aarch64-apple-darwin",
       "aarch64-unknown-linux-gnu",
+      "aarch64-unknown-linux-musl",
       "x86_64-pc-windows-msvc",
-      "x86_64-unknown-linux-gnu"
+      "x86_64-unknown-linux-gnu",
+      "x86_64-unknown-linux-musl"
     ]
   },
   "engines": {
@@ -106,23 +108,23 @@
   "devDependencies": {
     "@napi-rs/cli": "^3.5.1",
     "@types/archiver": "^7.0.0",
-    "@types/node": "^25.2.1",
+    "@types/node": "^25.2.3",
     "@types/which": "^3.0.4",
     "@vitest/coverage-v8": "^4.0.18",
     "archiver": "^7.0.1",
-    "oxlint": "^1.43.0",
+    "oxlint": "^1.46.0",
     "tsup": "^8.5.1",
     "typescript": "^5.9.3",
     "vitest": "^4.0.18"
   },
   "dependencies": {
     "@emnapi/runtime": "1.8.1",
-    "which": "^6.0.0"
+    "which": "^6.0.1"
   },
   "optionalDependencies": {
-    "@kreuzberg/node-darwin-arm64": "4.2.14",
-    "@kreuzberg/node-linux-arm64-gnu": "4.2.14",
-    "@kreuzberg/node-linux-x64-gnu": "4.2.14",
-    "@kreuzberg/node-win32-x64-msvc": "4.2.14"
+    "@kreuzberg/node-darwin-arm64": "4.3.0",
+    "@kreuzberg/node-linux-arm64-gnu": "4.3.0",
+    "@kreuzberg/node-linux-x64-gnu": "4.3.0",
+    "@kreuzberg/node-win32-x64-msvc": "4.3.0"
   }
 }

package/dist/ocr/guten-ocr.d.mts

@@ -1,193 +0,0 @@
-import { OcrBackendProtocol } from '../types.mjs';
-
-/**
- * Guten OCR backend for document OCR processing.
- *
- * This module provides integration with @gutenye/ocr-node for optical character recognition.
- * Guten OCR uses PaddleOCR models via ONNX Runtime for high-performance text extraction.
- *
- * @module ocr/guten-ocr
- */
-
-/**
- * Guten OCR backend for OCR processing.
- *
- * This backend uses @gutenye/ocr-node for text extraction from images.
- * It uses PaddleOCR models via ONNX Runtime for efficient processing.
- *
- * ## Installation
- *
- * Install the optional dependency:
- * ```bash
- * npm install @gutenye/ocr-node
- * # or
- * pnpm add @gutenye/ocr-node
- * # or
- * bun add @gutenye/ocr-node
- * ```
- *
- * ## Usage
- *
- * ```typescript
- * import { GutenOcrBackend } from '@kreuzberg/node/ocr/guten-ocr';
- * import { registerOcrBackend, extractFile } from '@kreuzberg/node';
- *
- * // Create and register the backend
- * const backend = new GutenOcrBackend();
- * await backend.initialize();
- * registerOcrBackend(backend);
- *
- * // Extract with OCR enabled
- * const result = await extractFile('scanned.pdf', null, {
- *   ocr: { backend: 'guten-ocr', language: 'en' },
- * });
- * console.log(result.content);
- * ```
- *
- * ## Supported Languages
- *
- * Guten OCR supports multiple languages via different model configurations.
- * The default models support English ("en") and Chinese ("ch_sim", "ch_tra").
- *
- * @example
- * ```typescript
- * // Basic usage with default settings
- * const backend = new GutenOcrBackend();
- * await backend.initialize();
- *
- * // Custom model configuration
- * const customBackend = new GutenOcrBackend({
- *   models: {
- *     detectionPath: './models/detection.onnx',
- *     recognitionPath: './models/recognition.onnx',
- *     dictionaryPath: './models/dict.txt'
- *   }
- * });
- * await customBackend.initialize();
- * ```
- */
-declare class GutenOcrBackend implements OcrBackendProtocol {
-    private ocr;
-    private ocrModule;
-    private options?;
-    /**
-     * Create a new Guten OCR backend.
-     *
-     * @param options - Optional configuration for Guten OCR
-     * @param options.models - Custom model paths (default: uses bundled models)
-     * @param options.isDebug - Enable debug mode (default: false)
-     * @param options.debugOutputDir - Directory for debug output (default: undefined)
-     * @param options.onnxOptions - Custom ONNX Runtime options (default: undefined)
-     *
-     * @example
-     * ```typescript
-     * // Default configuration
-     * const backend = new GutenOcrBackend();
-     *
-     * // With debug enabled
-     * const debugBackend = new GutenOcrBackend({
-     *   isDebug: true,
-     *   debugOutputDir: './ocr_debug'
-     * });
-     * ```
-     */
-    constructor(options?: {
-        models?: {
-            detectionPath: string;
-            recognitionPath: string;
-            dictionaryPath: string;
-        };
-        isDebug?: boolean;
-        debugOutputDir?: string;
-        onnxOptions?: unknown;
-    });
-    /**
-     * Get the backend name.
-     *
-     * @returns Backend name ("guten-ocr")
-     */
-    name(): string;
-    /**
-     * Get list of supported language codes.
-     *
-     * Guten OCR supports multiple languages depending on the model configuration.
-     * The default models support English and Chinese.
-     *
-     * @returns Array of ISO 639-1/2 language codes
-     */
-    supportedLanguages(): string[];
-    /**
-     * Initialize the OCR backend.
-     *
-     * This method loads the Guten OCR module and creates an OCR instance.
-     * Call this before using processImage().
-     *
-     * @throws {Error} If @gutenye/ocr-node is not installed
-     * @throws {Error} If OCR initialization fails
-     *
-     * @example
-     * ```typescript
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     * ```
-     */
-    initialize(): Promise<void>;
-    /**
-     * Shutdown the backend and release resources.
-     *
-     * This method cleans up all resources associated with the backend,
-     * including the GutenOCR instance and module references.
-     *
-     * @example
-     * ```typescript
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     * // ... use backend ...
-     * await backend.shutdown();
-     * ```
-     */
-    shutdown(): Promise<void>;
-    /**
-     * Process image bytes and extract text using Guten OCR.
-     *
-     * This method:
-     * 1. Decodes the image using sharp (if pixel data is needed) or passes bytes directly
-     * 2. Runs OCR detection to find text regions
-     * 3. Runs OCR recognition on each text region
-     * 4. Returns extracted text with metadata
-     *
-     * @param imageBytes - Raw image data (PNG, JPEG, TIFF, etc.)
-     * @param language - Language code (must be in supportedLanguages())
-     * @returns Promise resolving to OCR result with content and metadata
-     *
-     * @throws {Error} If backend is not initialized
-     * @throws {Error} If OCR processing fails
-     *
-     * @example
-     * ```typescript
-     * import { readFile } from 'fs/promises';
-     *
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     *
-     * const imageBytes = await readFile('scanned.png');
-     * const result = await backend.processImage(imageBytes, 'en');
-     * console.log(result.content);
-     * console.log(result.metadata.confidence);
-     * ```
-     */
-    processImage(imageBytes: Uint8Array | string, language: string): Promise<{
-        content: string;
-        mime_type: string;
-        metadata: {
-            width: number;
-            height: number;
-            confidence: number;
-            text_regions: number;
-            language: string;
-        };
-        tables: never[];
-    }>;
-}
-
-export { GutenOcrBackend };

package/dist/ocr/guten-ocr.d.ts

@@ -1,193 +0,0 @@
-import { OcrBackendProtocol } from '../types.js';
-
-/**
- * Guten OCR backend for document OCR processing.
- *
- * This module provides integration with @gutenye/ocr-node for optical character recognition.
- * Guten OCR uses PaddleOCR models via ONNX Runtime for high-performance text extraction.
- *
- * @module ocr/guten-ocr
- */
-
-/**
- * Guten OCR backend for OCR processing.
- *
- * This backend uses @gutenye/ocr-node for text extraction from images.
- * It uses PaddleOCR models via ONNX Runtime for efficient processing.
- *
- * ## Installation
- *
- * Install the optional dependency:
- * ```bash
- * npm install @gutenye/ocr-node
- * # or
- * pnpm add @gutenye/ocr-node
- * # or
- * bun add @gutenye/ocr-node
- * ```
- *
- * ## Usage
- *
- * ```typescript
- * import { GutenOcrBackend } from '@kreuzberg/node/ocr/guten-ocr';
- * import { registerOcrBackend, extractFile } from '@kreuzberg/node';
- *
- * // Create and register the backend
- * const backend = new GutenOcrBackend();
- * await backend.initialize();
- * registerOcrBackend(backend);
- *
- * // Extract with OCR enabled
- * const result = await extractFile('scanned.pdf', null, {
- *   ocr: { backend: 'guten-ocr', language: 'en' },
- * });
- * console.log(result.content);
- * ```
- *
- * ## Supported Languages
- *
- * Guten OCR supports multiple languages via different model configurations.
- * The default models support English ("en") and Chinese ("ch_sim", "ch_tra").
- *
- * @example
- * ```typescript
- * // Basic usage with default settings
- * const backend = new GutenOcrBackend();
- * await backend.initialize();
- *
- * // Custom model configuration
- * const customBackend = new GutenOcrBackend({
- *   models: {
- *     detectionPath: './models/detection.onnx',
- *     recognitionPath: './models/recognition.onnx',
- *     dictionaryPath: './models/dict.txt'
- *   }
- * });
- * await customBackend.initialize();
- * ```
- */
-declare class GutenOcrBackend implements OcrBackendProtocol {
-    private ocr;
-    private ocrModule;
-    private options?;
-    /**
-     * Create a new Guten OCR backend.
-     *
-     * @param options - Optional configuration for Guten OCR
-     * @param options.models - Custom model paths (default: uses bundled models)
-     * @param options.isDebug - Enable debug mode (default: false)
-     * @param options.debugOutputDir - Directory for debug output (default: undefined)
-     * @param options.onnxOptions - Custom ONNX Runtime options (default: undefined)
-     *
-     * @example
-     * ```typescript
-     * // Default configuration
-     * const backend = new GutenOcrBackend();
-     *
-     * // With debug enabled
-     * const debugBackend = new GutenOcrBackend({
-     *   isDebug: true,
-     *   debugOutputDir: './ocr_debug'
-     * });
-     * ```
-     */
-    constructor(options?: {
-        models?: {
-            detectionPath: string;
-            recognitionPath: string;
-            dictionaryPath: string;
-        };
-        isDebug?: boolean;
-        debugOutputDir?: string;
-        onnxOptions?: unknown;
-    });
-    /**
-     * Get the backend name.
-     *
-     * @returns Backend name ("guten-ocr")
-     */
-    name(): string;
-    /**
-     * Get list of supported language codes.
-     *
-     * Guten OCR supports multiple languages depending on the model configuration.
-     * The default models support English and Chinese.
-     *
-     * @returns Array of ISO 639-1/2 language codes
-     */
-    supportedLanguages(): string[];
-    /**
-     * Initialize the OCR backend.
-     *
-     * This method loads the Guten OCR module and creates an OCR instance.
-     * Call this before using processImage().
-     *
-     * @throws {Error} If @gutenye/ocr-node is not installed
-     * @throws {Error} If OCR initialization fails
-     *
-     * @example
-     * ```typescript
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     * ```
-     */
-    initialize(): Promise<void>;
-    /**
-     * Shutdown the backend and release resources.
-     *
-     * This method cleans up all resources associated with the backend,
-     * including the GutenOCR instance and module references.
-     *
-     * @example
-     * ```typescript
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     * // ... use backend ...
-     * await backend.shutdown();
-     * ```
-     */
-    shutdown(): Promise<void>;
-    /**
-     * Process image bytes and extract text using Guten OCR.
-     *
-     * This method:
-     * 1. Decodes the image using sharp (if pixel data is needed) or passes bytes directly
-     * 2. Runs OCR detection to find text regions
-     * 3. Runs OCR recognition on each text region
-     * 4. Returns extracted text with metadata
-     *
-     * @param imageBytes - Raw image data (PNG, JPEG, TIFF, etc.)
-     * @param language - Language code (must be in supportedLanguages())
-     * @returns Promise resolving to OCR result with content and metadata
-     *
-     * @throws {Error} If backend is not initialized
-     * @throws {Error} If OCR processing fails
-     *
-     * @example
-     * ```typescript
-     * import { readFile } from 'fs/promises';
-     *
-     * const backend = new GutenOcrBackend();
-     * await backend.initialize();
-     *
-     * const imageBytes = await readFile('scanned.png');
-     * const result = await backend.processImage(imageBytes, 'en');
-     * console.log(result.content);
-     * console.log(result.metadata.confidence);
-     * ```
-     */
-    processImage(imageBytes: Uint8Array | string, language: string): Promise<{
-        content: string;
-        mime_type: string;
-        metadata: {
-            width: number;
-            height: number;
-            confidence: number;
-            text_regions: number;
-            language: string;
-        };
-        tables: never[];
-    }>;
-}
-
-export { GutenOcrBackend };

package/dist/ocr/guten-ocr.js

@@ -1,234 +0,0 @@
-"use strict";
-var __create = Object.create;
-var __defProp = Object.defineProperty;
-var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
-var __getOwnPropNames = Object.getOwnPropertyNames;
-var __getProtoOf = Object.getPrototypeOf;
-var __hasOwnProp = Object.prototype.hasOwnProperty;
-var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
-var __copyProps = (to, from, except, desc) => {
-  if (from && typeof from === "object" || typeof from === "function") {
-    for (let key of __getOwnPropNames(from))
-      if (!__hasOwnProp.call(to, key) && key !== except)
-        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
-  }
-  return to;
-};
-var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
-  // If the importer is in node compatibility mode or this is not an ESM
-  // file that has been converted to a CommonJS file using a Babel-
-  // compatible transform (i.e. "__esModule" has not been set), then set
-  // "default" to the CommonJS "module.exports" for node compatibility.
-  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
-  mod
-));
-var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
-
-// typescript/ocr/guten-ocr.ts
-var guten_ocr_exports = {};
-__export(guten_ocr_exports, {
-  GutenOcrBackend: () => GutenOcrBackend
-});
-module.exports = __toCommonJS(guten_ocr_exports);
-var GutenOcrBackend = class {
-  ocr = null;
-  ocrModule = null;
-  options;
-  /**
-   * Create a new Guten OCR backend.
-   *
-   * @param options - Optional configuration for Guten OCR
-   * @param options.models - Custom model paths (default: uses bundled models)
-   * @param options.isDebug - Enable debug mode (default: false)
-   * @param options.debugOutputDir - Directory for debug output (default: undefined)
-   * @param options.onnxOptions - Custom ONNX Runtime options (default: undefined)
-   *
-   * @example
-   * ```typescript
-   * // Default configuration
-   * const backend = new GutenOcrBackend();
-   *
-   * // With debug enabled
-   * const debugBackend = new GutenOcrBackend({
-   *   isDebug: true,
-   *   debugOutputDir: './ocr_debug'
-   * });
-   * ```
-   */
-  constructor(options) {
-    if (options !== void 0) {
-      this.options = options;
-    }
-  }
-  /**
-   * Get the backend name.
-   *
-   * @returns Backend name ("guten-ocr")
-   */
-  name() {
-    return "guten-ocr";
-  }
-  /**
-   * Get list of supported language codes.
-   *
-   * Guten OCR supports multiple languages depending on the model configuration.
-   * The default models support English and Chinese.
-   *
-   * @returns Array of ISO 639-1/2 language codes
-   */
-  supportedLanguages() {
-    return ["en", "eng", "ch_sim", "ch_tra", "chinese"];
-  }
-  /**
-   * Initialize the OCR backend.
-   *
-   * This method loads the Guten OCR module and creates an OCR instance.
-   * Call this before using processImage().
-   *
-   * @throws {Error} If @gutenye/ocr-node is not installed
-   * @throws {Error} If OCR initialization fails
-   *
-   * @example
-   * ```typescript
-   * const backend = new GutenOcrBackend();
-   * await backend.initialize();
-   * ```
-   */
-  async initialize() {
-    if (this.ocr !== null) {
-      return;
-    }
-    try {
-      this.ocrModule = await import("@gutenye/ocr-node").then((m) => m.default || m);
-    } catch (e) {
-      const error = e;
-      throw new Error(
-        `Guten OCR support requires the '@gutenye/ocr-node' package. Install with: npm install @gutenye/ocr-node. Error: ${error.message}`
-      );
-    }
-    try {
-      this.ocr = await this.ocrModule?.create(this.options) ?? null;
-    } catch (e) {
-      const error = e;
-      throw new Error(`Failed to initialize Guten OCR: ${error.message}`);
-    }
-  }
-  /**
-   * Shutdown the backend and release resources.
-   *
-   * This method cleans up all resources associated with the backend,
-   * including the GutenOCR instance and module references.
-   *
-   * @example
-   * ```typescript
-   * const backend = new GutenOcrBackend();
-   * await backend.initialize();
-   * // ... use backend ...
-   * await backend.shutdown();
-   * ```
-   */
-  async shutdown() {
-    if (this.ocr !== null) {
-      this.ocr = null;
-    }
-    if (this.ocrModule !== null) {
-      this.ocrModule = null;
-    }
-  }
-  /**
-   * Process image bytes and extract text using Guten OCR.
-   *
-   * This method:
-   * 1. Decodes the image using sharp (if pixel data is needed) or passes bytes directly
-   * 2. Runs OCR detection to find text regions
-   * 3. Runs OCR recognition on each text region
-   * 4. Returns extracted text with metadata
-   *
-   * @param imageBytes - Raw image data (PNG, JPEG, TIFF, etc.)
-   * @param language - Language code (must be in supportedLanguages())
-   * @returns Promise resolving to OCR result with content and metadata
-   *
-   * @throws {Error} If backend is not initialized
-   * @throws {Error} If OCR processing fails
-   *
-   * @example
-   * ```typescript
-   * import { readFile } from 'fs/promises';
-   *
-   * const backend = new GutenOcrBackend();
-   * await backend.initialize();
-   *
-   * const imageBytes = await readFile('scanned.png');
-   * const result = await backend.processImage(imageBytes, 'en');
-   * console.log(result.content);
-   * console.log(result.metadata.confidence);
-   * ```
-   */
-  async processImage(imageBytes, language) {
-    if (this.ocr === null) {
-      await this.initialize();
-    }
-    if (this.ocr === null) {
-      throw new Error("Guten OCR backend failed to initialize");
-    }
-    try {
-      const buffer = typeof imageBytes === "string" ? Buffer.from(imageBytes, "base64") : Buffer.from(imageBytes);
-      const debugEnv = process.env["KREUZBERG_DEBUG_GUTEN"];
-      if (debugEnv === "1") {
-        const header = Array.from(buffer.subarray(0, 8));
-        console.log("[Guten OCR] Debug input header:", header);
-        console.log(
-          "[Guten OCR] Buffer?",
-          Buffer.isBuffer(buffer),
-          "constructor",
-          imageBytes?.constructor?.name,
-          "length",
-          buffer.length,
-          "type",
-          typeof imageBytes
-        );
-      }
-      let width = 0;
-      let height = 0;
-      try {
-        const sharpModule = await import("sharp");
-        const sharp = sharpModule.default || sharpModule;
-        const image = sharp(buffer);
-        const metadata = await image.metadata();
-        const metadataRecord = metadata;
-        width = metadataRecord["width"] ?? 0;
-        height = metadataRecord["height"] ?? 0;
-      } catch (metadataError) {
-        const error = metadataError;
-        console.warn(`[Guten OCR] Unable to read image metadata via sharp: ${error.message}`);
-      }
-      const result = await this.ocr.detect(buffer);
-      const textLines = result.map((line) => line.text);
-      const content = textLines.join("\n");
-      const avgConfidence = result.length > 0 ? result.reduce((sum, line) => sum + line.mean, 0) / result.length : 0;
-      return {
-        content,
-        mime_type: "text/plain",
-        metadata: {
-          width,
-          height,
-          confidence: avgConfidence,
-          text_regions: result.length,
-          language
-        },
-        tables: []
-      };
-    } catch (e) {
-      const error = e;
-      throw new Error(`Guten OCR processing failed: ${error.message}`);
-    }
-  }
-};
-// Annotate the CommonJS export names for ESM import in node:
-0 && (module.exports = {
-  GutenOcrBackend
-});
-//# sourceMappingURL=guten-ocr.js.map

package/dist/ocr/guten-ocr.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../../typescript/ocr/guten-ocr.ts"],"sourcesContent":["/**\n * Guten OCR backend for document OCR processing.\n *\n * This module provides integration with @gutenye/ocr-node for optical character recognition.\n * Guten OCR uses PaddleOCR models via ONNX Runtime for high-performance text extraction.\n *\n * @module ocr/guten-ocr\n */\n\nimport type { OcrBackendProtocol } from \"../types.js\";\n\n/**\n * Text line detected by Guten OCR.\n */\ninterface TextLine {\n\ttext: string;\n\tmean: number;\n\tbox: number[][];\n}\n\n/**\n * Guten OCR instance interface.\n */\ninterface GutenOcr {\n\tdetect(imagePath: string | Buffer, options?: { onnxOptions?: unknown }): Promise<TextLine[]>;\n}\n\n/**\n * Guten OCR module interface.\n */\ninterface GutenOcrModule {\n\tcreate(options?: {\n\t\tmodels?: {\n\t\t\tdetectionPath: string;\n\t\t\trecognitionPath: string;\n\t\t\tdictionaryPath: string;\n\t\t};\n\t\tisDebug?: boolean;\n\t\tdebugOutputDir?: string;\n\t\tonnxOptions?: unknown;\n\t}): Promise<GutenOcr>;\n}\n\n/**\n * Guten OCR backend for OCR processing.\n *\n * This backend uses @gutenye/ocr-node for text extraction from images.\n * It uses PaddleOCR models via ONNX Runtime for efficient processing.\n *\n * ## Installation\n *\n * Install the optional dependency:\n * ```bash\n * npm install @gutenye/ocr-node\n * # or\n * pnpm add @gutenye/ocr-node\n * # or\n * bun add @gutenye/ocr-node\n * ```\n *\n * ## Usage\n *\n * ```typescript\n * import { GutenOcrBackend } from '@kreuzberg/node/ocr/guten-ocr';\n * import { registerOcrBackend, extractFile } from '@kreuzberg/node';\n *\n * // Create and register the backend\n * const backend = new GutenOcrBackend();\n * await backend.initialize();\n * registerOcrBackend(backend);\n *\n * // Extract with OCR enabled\n * const result = await extractFile('scanned.pdf', null, {\n *   ocr: { backend: 'guten-ocr', language: 'en' },\n * });\n * console.log(result.content);\n * ```\n *\n * ## Supported Languages\n *\n * Guten OCR supports multiple languages via different model configurations.\n * The default models support English (\"en\") and Chinese (\"ch_sim\", \"ch_tra\").\n *\n * @example\n * ```typescript\n * // Basic usage with default settings\n * const backend = new GutenOcrBackend();\n * await backend.initialize();\n *\n * // Custom model configuration\n * const customBackend = new GutenOcrBackend({\n *   models: {\n *     detectionPath: './models/detection.onnx',\n *     recognitionPath: './models/recognition.onnx',\n *     dictionaryPath: './models/dict.txt'\n *   }\n * });\n * await customBackend.initialize();\n * ```\n */\nexport class GutenOcrBackend implements OcrBackendProtocol {\n\tprivate ocr: GutenOcr | null = null;\n\tprivate ocrModule: GutenOcrModule | null = null;\n\tprivate options?: {\n\t\tmodels?: {\n\t\t\tdetectionPath: string;\n\t\t\trecognitionPath: string;\n\t\t\tdictionaryPath: string;\n\t\t};\n\t\tisDebug?: boolean;\n\t\tdebugOutputDir?: string;\n\t\tonnxOptions?: unknown;\n\t};\n\n\t/**\n\t * Create a new Guten OCR backend.\n\t *\n\t * @param options - Optional configuration for Guten OCR\n\t * @param options.models - Custom model paths (default: uses bundled models)\n\t * @param options.isDebug - Enable debug mode (default: false)\n\t * @param options.debugOutputDir - Directory for debug output (default: undefined)\n\t * @param options.onnxOptions - Custom ONNX Runtime options (default: undefined)\n\t *\n\t * @example\n\t * ```typescript\n\t * // Default configuration\n\t * const backend = new GutenOcrBackend();\n\t *\n\t * // With debug enabled\n\t * const debugBackend = new GutenOcrBackend({\n\t *   isDebug: true,\n\t *   debugOutputDir: './ocr_debug'\n\t * });\n\t * ```\n\t */\n\tconstructor(options?: {\n\t\tmodels?: {\n\t\t\tdetectionPath: string;\n\t\t\trecognitionPath: string;\n\t\t\tdictionaryPath: string;\n\t\t};\n\t\tisDebug?: boolean;\n\t\tdebugOutputDir?: string;\n\t\tonnxOptions?: unknown;\n\t}) {\n\t\tif (options !== undefined) {\n\t\t\tthis.options = options;\n\t\t}\n\t}\n\n\t/**\n\t * Get the backend name.\n\t *\n\t * @returns Backend name (\"guten-ocr\")\n\t */\n\tname(): string {\n\t\treturn \"guten-ocr\";\n\t}\n\n\t/**\n\t * Get list of supported language codes.\n\t *\n\t * Guten OCR supports multiple languages depending on the model configuration.\n\t * The default models support English and Chinese.\n\t *\n\t * @returns Array of ISO 639-1/2 language codes\n\t */\n\tsupportedLanguages(): string[] {\n\t\treturn [\"en\", \"eng\", \"ch_sim\", \"ch_tra\", \"chinese\"];\n\t}\n\n\t/**\n\t * Initialize the OCR backend.\n\t *\n\t * This method loads the Guten OCR module and creates an OCR instance.\n\t * Call this before using processImage().\n\t *\n\t * @throws {Error} If @gutenye/ocr-node is not installed\n\t * @throws {Error} If OCR initialization fails\n\t *\n\t * @example\n\t * ```typescript\n\t * const backend = new GutenOcrBackend();\n\t * await backend.initialize();\n\t * ```\n\t */\n\tasync initialize(): Promise<void> {\n\t\tif (this.ocr !== null) {\n\t\t\treturn;\n\t\t}\n\n\t\ttry {\n\t\t\tthis.ocrModule = await import(\"@gutenye/ocr-node\").then((m) => (m.default || m) as GutenOcrModule);\n\t\t} catch (e) {\n\t\t\tconst error = e as Error;\n\t\t\tthrow new Error(\n\t\t\t\t`Guten OCR support requires the '@gutenye/ocr-node' package. ` +\n\t\t\t\t\t`Install with: npm install @gutenye/ocr-node. ` +\n\t\t\t\t\t`Error: ${error.message}`,\n\t\t\t);\n\t\t}\n\n\t\ttry {\n\t\t\tthis.ocr = (await this.ocrModule?.create(this.options)) ?? null;\n\t\t} catch (e) {\n\t\t\tconst error = e as Error;\n\t\t\tthrow new Error(`Failed to initialize Guten OCR: ${error.message}`);\n\t\t}\n\t}\n\n\t/**\n\t * Shutdown the backend and release resources.\n\t *\n\t * This method cleans up all resources associated with the backend,\n\t * including the GutenOCR instance and module references.\n\t *\n\t * @example\n\t * ```typescript\n\t * const backend = new GutenOcrBackend();\n\t * await backend.initialize();\n\t * // ... use backend ...\n\t * await backend.shutdown();\n\t * ```\n\t */\n\tasync shutdown(): Promise<void> {\n\t\tif (this.ocr !== null) {\n\t\t\tthis.ocr = null;\n\t\t}\n\n\t\tif (this.ocrModule !== null) {\n\t\t\tthis.ocrModule = null;\n\t\t}\n\t}\n\n\t/**\n\t * Process image bytes and extract text using Guten OCR.\n\t *\n\t * This method:\n\t * 1. Decodes the image using sharp (if pixel data is needed) or passes bytes directly\n\t * 2. Runs OCR detection to find text regions\n\t * 3. Runs OCR recognition on each text region\n\t * 4. Returns extracted text with metadata\n\t *\n\t * @param imageBytes - Raw image data (PNG, JPEG, TIFF, etc.)\n\t * @param language - Language code (must be in supportedLanguages())\n\t * @returns Promise resolving to OCR result with content and metadata\n\t *\n\t * @throws {Error} If backend is not initialized\n\t * @throws {Error} If OCR processing fails\n\t *\n\t * @example\n\t * ```typescript\n\t * import { readFile } from 'fs/promises';\n\t *\n\t * const backend = new GutenOcrBackend();\n\t * await backend.initialize();\n\t *\n\t * const imageBytes = await readFile('scanned.png');\n\t * const result = await backend.processImage(imageBytes, 'en');\n\t * console.log(result.content);\n\t * console.log(result.metadata.confidence);\n\t * ```\n\t */\n\tasync processImage(\n\t\timageBytes: Uint8Array | string,\n\t\tlanguage: string,\n\t): Promise<{\n\t\tcontent: string;\n\t\tmime_type: string;\n\t\tmetadata: {\n\t\t\twidth: number;\n\t\t\theight: number;\n\t\t\tconfidence: number;\n\t\t\ttext_regions: number;\n\t\t\tlanguage: string;\n\t\t};\n\t\ttables: never[];\n\t}> {\n\t\tif (this.ocr === null) {\n\t\t\tawait this.initialize();\n\t\t}\n\n\t\tif (this.ocr === null) {\n\t\t\tthrow new Error(\"Guten OCR backend failed to initialize\");\n\t\t}\n\n\t\ttry {\n\t\t\tconst buffer = typeof imageBytes === \"string\" ? Buffer.from(imageBytes, \"base64\") : Buffer.from(imageBytes);\n\n\t\t\t// biome-ignore lint/complexity/useLiteralKeys: required for strict TypeScript noUncheckedIndexedAccess\n\t\t\tconst debugEnv = process.env[\"KREUZBERG_DEBUG_GUTEN\"];\n\t\t\tif (debugEnv === \"1\") {\n\t\t\t\tconst header = Array.from(buffer.subarray(0, 8));\n\t\t\t\tconsole.log(\"[Guten OCR] Debug input header:\", header);\n\t\t\t\tconsole.log(\n\t\t\t\t\t\"[Guten OCR] Buffer?\",\n\t\t\t\t\tBuffer.isBuffer(buffer),\n\t\t\t\t\t\"constructor\",\n\t\t\t\t\timageBytes?.constructor?.name,\n\t\t\t\t\t\"length\",\n\t\t\t\t\tbuffer.length,\n\t\t\t\t\t\"type\",\n\t\t\t\t\ttypeof imageBytes,\n\t\t\t\t);\n\t\t\t}\n\n\t\t\tlet width = 0;\n\t\t\tlet height = 0;\n\n\t\t\ttry {\n\t\t\t\tconst sharpModule = await import(\"sharp\");\n\t\t\t\tconst sharp = (sharpModule as unknown as { default?: unknown }).default || sharpModule;\n\t\t\t\tconst image = (sharp as (buffer: Buffer) => { metadata: () => Promise<Record<string, unknown>> })(buffer);\n\t\t\t\tconst metadata = await image.metadata();\n\t\t\t\tconst metadataRecord = metadata as Record<string, unknown>;\n\t\t\t\t// biome-ignore lint/complexity/useLiteralKeys: TypeScript TS4111 requires bracket notation for index signature properties\n\t\t\t\twidth = (metadataRecord[\"width\"] as number | undefined) ?? 0;\n\t\t\t\t// biome-ignore lint/complexity/useLiteralKeys: TypeScript TS4111 requires bracket notation for index signature properties\n\t\t\t\theight = (metadataRecord[\"height\"] as number | undefined) ?? 0;\n\t\t\t} catch (metadataError) {\n\t\t\t\tconst error = metadataError as Error;\n\t\t\t\tconsole.warn(`[Guten OCR] Unable to read image metadata via sharp: ${error.message}`);\n\t\t\t}\n\n\t\t\tconst result = await this.ocr.detect(buffer);\n\n\t\t\tconst textLines = result.map((line) => line.text);\n\t\t\tconst content = textLines.join(\"\\n\");\n\n\t\t\tconst avgConfidence = result.length > 0 ? result.reduce((sum, line) => sum + line.mean, 0) / result.length : 0;\n\n\t\t\treturn {\n\t\t\t\tcontent,\n\t\t\t\tmime_type: \"text/plain\",\n\t\t\t\tmetadata: {\n\t\t\t\t\twidth,\n\t\t\t\t\theight,\n\t\t\t\t\tconfidence: avgConfidence,\n\t\t\t\t\ttext_regions: result.length,\n\t\t\t\t\tlanguage,\n\t\t\t\t},\n\t\t\t\ttables: [],\n\t\t\t};\n\t\t} catch (e) {\n\t\t\tconst error = e as Error;\n\t\t\tthrow new Error(`Guten OCR processing failed: ${error.message}`);\n\t\t}\n\t}\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAoGO,IAAM,kBAAN,MAAoD;AAAA,EAClD,MAAuB;AAAA,EACvB,YAAmC;AAAA,EACnC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgCR,YAAY,SAST;AACF,QAAI,YAAY,QAAW;AAC1B,WAAK,UAAU;AAAA,IAChB;AAAA,EACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,OAAe;AACd,WAAO;AAAA,EACR;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,qBAA+B;AAC9B,WAAO,CAAC,MAAM,OAAO,UAAU,UAAU,SAAS;AAAA,EACnD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAiBA,MAAM,aAA4B;AACjC,QAAI,KAAK,QAAQ,MAAM;AACtB;AAAA,IACD;AAEA,QAAI;AACH,WAAK,YAAY,MAAM,OAAO,mBAAmB,EAAE,KAAK,CAAC,MAAO,EAAE,WAAW,CAAoB;AAAA,IAClG,SAAS,GAAG;AACX,YAAM,QAAQ;AACd,YAAM,IAAI;AAAA,QACT,mHAEW,MAAM,OAAO;AAAA,MACzB;AAAA,IACD;AAEA,QAAI;AACH,WAAK,MAAO,MAAM,KAAK,WAAW,OAAO,KAAK,OAAO,KAAM;AAAA,IAC5D,SAAS,GAAG;AACX,YAAM,QAAQ;AACd,YAAM,IAAI,MAAM,mCAAmC,MAAM,OAAO,EAAE;AAAA,IACnE;AAAA,EACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAgBA,MAAM,WAA0B;AAC/B,QAAI,KAAK,QAAQ,MAAM;AACtB,WAAK,MAAM;AAAA,IACZ;AAEA,QAAI,KAAK,cAAc,MAAM;AAC5B,WAAK,YAAY;AAAA,IAClB;AAAA,EACD;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EA+BA,MAAM,aACL,YACA,UAYE;AACF,QAAI,KAAK,QAAQ,MAAM;AACtB,YAAM,KAAK,WAAW;AAAA,IACvB;AAEA,QAAI,KAAK,QAAQ,MAAM;AACtB,YAAM,IAAI,MAAM,wCAAwC;AAAA,IACzD;AAEA,QAAI;AACH,YAAM,SAAS,OAAO,eAAe,WAAW,OAAO,KAAK,YAAY,QAAQ,IAAI,OAAO,KAAK,UAAU;AAG1G,YAAM,WAAW,QAAQ,IAAI,uBAAuB;AACpD,UAAI,aAAa,KAAK;AACrB,cAAM,SAAS,MAAM,KAAK,OAAO,SAAS,GAAG,CAAC,CAAC;AAC/C,gBAAQ,IAAI,mCAAmC,MAAM;AACrD,gBAAQ;AAAA,UACP;AAAA,UACA,OAAO,SAAS,MAAM;AAAA,UACtB;AAAA,UACA,YAAY,aAAa;AAAA,UACzB;AAAA,UACA,OAAO;AAAA,UACP;AAAA,UACA,OAAO;AAAA,QACR;AAAA,MACD;AAEA,UAAI,QAAQ;AACZ,UAAI,SAAS;AAEb,UAAI;AACH,cAAM,cAAc,MAAM,OAAO,OAAO;AACxC,cAAM,QAAS,YAAiD,WAAW;AAC3E,cAAM,QAAS,MAAmF,MAAM;AACxG,cAAM,WAAW,MAAM,MAAM,SAAS;AACtC,cAAM,iBAAiB;AAEvB,gBAAS,eAAe,OAAO,KAA4B;AAE3D,iBAAU,eAAe,QAAQ,KAA4B;AAAA,MAC9D,SAAS,eAAe;AACvB,cAAM,QAAQ;AACd,gBAAQ,KAAK,wDAAwD,MAAM,OAAO,EAAE;AAAA,MACrF;AAEA,YAAM,SAAS,MAAM,KAAK,IAAI,OAAO,MAAM;AAE3C,YAAM,YAAY,OAAO,IAAI,CAAC,SAAS,KAAK,IAAI;AAChD,YAAM,UAAU,UAAU,KAAK,IAAI;AAEnC,YAAM,gBAAgB,OAAO,SAAS,IAAI,OAAO,OAAO,CAAC,KAAK,SAAS,MAAM,KAAK,MAAM,CAAC,IAAI,OAAO,SAAS;AAE7G,aAAO;AAAA,QACN;AAAA,QACA,WAAW;AAAA,QACX,UAAU;AAAA,UACT;AAAA,UACA;AAAA,UACA,YAAY;AAAA,UACZ,cAAc,OAAO;AAAA,UACrB;AAAA,QACD;AAAA,QACA,QAAQ,CAAC;AAAA,MACV;AAAA,IACD,SAAS,GAAG;AACX,YAAM,QAAQ;AACd,YAAM,IAAI,MAAM,gCAAgC,MAAM,OAAO,EAAE;AAAA,IAChE;AAAA,EACD;AACD;","names":[]}