@kreuzberg/node 4.2.14 → 4.3.0

package/dist/index.mjs

@@ -281,6 +281,8 @@ function normalizeOcrConfig(ocr) {
   if (tesseract) {
     setIfDefined(normalized, "tesseractConfig", tesseract);
   }
+  setIfDefined(normalized, "paddleOcrConfig", ocr.paddleOcrConfig);
+  setIfDefined(normalized, "elementConfig", ocr.elementConfig);
   return normalized;
 }
 function normalizeChunkingConfig(chunking) {
@@ -429,6 +431,7 @@ function normalizeExtractionConfig(config) {
   setIfDefined(normalized, "useCache", config.useCache);
   setIfDefined(normalized, "enableQualityProcessing", config.enableQualityProcessing);
   setIfDefined(normalized, "forceOcr", config.forceOcr);
+  setIfDefined(normalized, "includeDocumentStructure", config.includeDocumentStructure);
   setIfDefined(normalized, "maxConcurrentExtractions", config.maxConcurrentExtractions);
   const ocr = normalizeOcrConfig(config.ocr);
   setIfDefined(normalized, "ocr", ocr);
@@ -610,7 +613,9 @@ function convertPageContent(rawPage) {
     // biome-ignore lint/complexity/useLiteralKeys: required for strict TypeScript noPropertyAccessFromIndexSignature
     tables: Array.isArray(page["tables"]) ? page["tables"] : [],
     // biome-ignore lint/complexity/useLiteralKeys: required for strict TypeScript noPropertyAccessFromIndexSignature
-    images: Array.isArray(page["images"]) ? page["images"].map((image) => convertImage(image)) : []
+    images: Array.isArray(page["images"]) ? page["images"].map((image) => convertImage(image)) : [],
+    // biome-ignore lint/complexity/useLiteralKeys: required for strict TypeScript noPropertyAccessFromIndexSignature
+    isBlank: page["isBlank"] ?? null
   };
 }
 function convertResult(rawResult) {
@@ -624,7 +629,8 @@ function convertResult(rawResult) {
       chunks: null,
       images: null,
       elements: null,
-      pages: null
+      pages: null,
+      document: null
     };
   }
   const result = rawResult;
@@ -643,7 +649,9 @@ function convertResult(rawResult) {
     chunks: null,
     images: null,
     elements: null,
-    pages: null
+    pages: null,
+    // biome-ignore lint/complexity/useLiteralKeys: required for strict TypeScript noPropertyAccessFromIndexSignature
+    document: result["document"] ?? null
   };
   const chunksData = result["chunks"];
   if (Array.isArray(chunksData)) {
@@ -661,6 +669,10 @@ function convertResult(rawResult) {
   if (Array.isArray(pagesData)) {
     returnObj.pages = pagesData.map((page) => convertPageContent(page));
   }
+  const ocrElementsData = result["ocrElements"];
+  if (Array.isArray(ocrElementsData)) {
+    returnObj.ocrElements = ocrElementsData;
+  }
   return returnObj;
 }
 
@@ -750,7 +762,7 @@ async function extractBytes(dataOrPath, mimeType, config = null) {
     data = dataOrPath;
   }
   const validated = assertUint8Array(data, "data");
-  if (process.env["KREUZBERG_DEBUG_GUTEN"] === "1") {
+  if (process.env["KREUZBERG_DEBUG_OCR"] === "1") {
     console.log("[TypeScript] Debug input header:", Array.from(validated.slice(0, 8)));
   }
   const normalizedConfig = normalizeExtractionConfig(config);
@@ -903,202 +915,6 @@ function listValidators() {
   return binding2.listValidators();
 }
 
-// typescript/ocr/guten-ocr.ts
-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}`);
-    }
-  }
-};
-
 // typescript/plugins/ocr-backends.ts
 function isOcrProcessTuple(value) {
   return Array.isArray(value) && value.length === 2 && typeof value[1] === "string" && (typeof value[0] === "string" || Buffer.isBuffer(value[0]) || value[0] instanceof Uint8Array);
@@ -1119,7 +935,7 @@ function registerOcrBackend(backend) {
     supportedLanguages: typeof backend.supportedLanguages === "function" ? backend.supportedLanguages() : backend.supportedLanguages ?? ["en"],
     async processImage(...processArgs) {
       const [imagePayload, maybeLanguage] = processArgs;
-      if (process.env["KREUZBERG_DEBUG_GUTEN"] === "1") {
+      if (process.env["KREUZBERG_DEBUG_OCR"] === "1") {
         console.log("[registerOcrBackend] JS arguments", { length: processArgs.length });
         console.log("[registerOcrBackend] Raw args", {
           imagePayloadType: Array.isArray(imagePayload) ? "tuple" : typeof imagePayload,
@@ -1139,7 +955,7 @@ function registerOcrBackend(backend) {
       if (typeof language !== "string") {
         throw new Error("OCR backend did not receive a language parameter");
       }
-      if (process.env["KREUZBERG_DEBUG_GUTEN"] === "1") {
+      if (process.env["KREUZBERG_DEBUG_OCR"] === "1") {
         const length = typeof rawBytes === "string" ? rawBytes.length : rawBytes.length;
         console.log(
           "[registerOcrBackend] Received payload",
@@ -1288,12 +1104,11 @@ function getEmbeddingPreset(name) {
 }
 
 // typescript/index.ts
-var __version__ = "4.2.14";
+var __version__ = "4.3.0";
 export {
   CacheError,
   ErrorCode,
   ExtractionConfig,
-  GutenOcrBackend,
   ImageProcessingError,
   KreuzbergError,
   MissingDependencyError,