@kreuzberg/wasm 4.4.0 → 4.4.2

package/README.md

@@ -22,7 +22,7 @@
     <img src="https://img.shields.io/maven-central/v/dev.kreuzberg/kreuzberg?label=Java&color=007ec6" alt="Java">
   </a>
   <a href="https://github.com/kreuzberg-dev/kreuzberg/releases">
-    <img src="https://img.shields.io/github/v/tag/kreuzberg-dev/kreuzberg?label=Go&color=007ec6&filter=v4.4.0" alt="Go">
+    <img src="https://img.shields.io/github/v/tag/kreuzberg-dev/kreuzberg?label=Go&color=007ec6&filter=v4.4.2" alt="Go">
   </a>
   <a href="https://www.nuget.org/packages/Kreuzberg/">
     <img src="https://img.shields.io/nuget/v/Kreuzberg?label=C%23&color=007ec6" alt="C#">

package/dist/extraction/batch.d.ts

@@ -33,8 +33,11 @@ export declare function batchExtractBytes(files: Array<{
 /**
  * Batch extract content from multiple byte arrays synchronously
  *
- * Synchronous version of batchExtractBytes. Extracts content from multiple documents
- * in a single batch operation without async operations.
+ * Synchronous version of {@link batchExtractBytes}. Extracts content from multiple
+ * documents in a single batch operation without async/await.
+ *
+ * **Note:** This function blocks the current thread until all extractions complete.
+ * For large batches, prefer the async {@link batchExtractBytes} function.
  *
  * @param files - Array of objects containing data (Uint8Array) and mimeType (string)
  * @param config - Optional extraction configuration applied to all files
@@ -44,8 +47,8 @@ export declare function batchExtractBytes(files: Array<{
  * @example
  * ```typescript
  * const files = [
- *   { data: pdfBytes, mimeType: 'application/pdf' },
- *   { data: docxBytes, mimeType: 'application/vnd.openxmlformats-officedocument.wordprocessingml.document' }
+ *   { data: txtBytes, mimeType: 'text/plain' },
+ *   { data: htmlBytes, mimeType: 'text/html' }
  * ];
  * const results = batchExtractBytesSync(files);
  * results.forEach((result) => console.log(result.content));

package/dist/extraction/batch.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"batch.d.ts","sourceRoot":"","sources":["../../typescript/extraction/batch.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAGH,OAAO,KAAK,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAG9F;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAsB,iBAAiB,CACtC,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,CAAC,EACpD,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,EAAE,CAAC,CA6D7B;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAgB,qBAAqB,CACpC,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,CAAC,EACpD,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,gBAAgB,EAAE,CA6DpB;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAsB,iBAAiB,CACtC,KAAK,EAAE,IAAI,EAAE,EACb,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,EAAE,CAAC,CAiC7B"}
+{"version":3,"file":"batch.d.ts","sourceRoot":"","sources":["../../typescript/extraction/batch.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAGH,OAAO,KAAK,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAG9F;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAsB,iBAAiB,CACtC,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,CAAC,EACpD,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,EAAE,CAAC,CA6D7B;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,wBAAgB,qBAAqB,CACpC,KAAK,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,UAAU,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAA;CAAE,CAAC,EACpD,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,gBAAgB,EAAE,CA6DpB;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAsB,iBAAiB,CACtC,KAAK,EAAE,IAAI,EAAE,EACb,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,EAAE,CAAC,CAiC7B"}

package/dist/extraction/bytes.d.ts

@@ -49,11 +49,14 @@ export declare function extractBytes(data: Uint8Array, mimeType: string, config?
 /**
  * Extract content from bytes synchronously
  *
- * Synchronous version of extractBytes. Performs extraction without async operations.
- * Note: Some extraction features may still be async internally, but the wrapper is synchronous.
+ * Synchronous version of {@link extractBytes}. Extracts text, metadata, tables,
+ * and other content from document bytes without async/await.
+ *
+ * **Note:** This function blocks the current thread until extraction completes.
+ * For large documents, prefer the async {@link extractBytes} function.
  *
  * @param data - The document bytes to extract from
- * @param mimeType - MIME type of the document
+ * @param mimeType - MIME type of the document (e.g., 'application/pdf', 'image/jpeg')
  * @param config - Optional extraction configuration
  * @returns The extraction result
  * @throws {Error} If WASM module is not initialized or extraction fails
@@ -61,7 +64,7 @@ export declare function extractBytes(data: Uint8Array, mimeType: string, config?
  * @example
  * ```typescript
  * const bytes = new Uint8Array(buffer);
- * const result = extractBytesSync(bytes, 'application/pdf');
+ * const result = extractBytesSync(bytes, 'text/plain');
  * console.log(result.content);
  * ```
  */

package/dist/extraction/bytes.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"bytes.d.ts","sourceRoot":"","sources":["../../typescript/extraction/bytes.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAGH,OAAO,KAAK,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAG9F;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAwCG;AACH,wBAAsB,YAAY,CACjC,IAAI,EAAE,UAAU,EAChB,QAAQ,EAAE,MAAM,EAChB,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,CAAC,CA4B3B;AAED;;;;;;;;;;;;;;;;;;GAkBG;AACH,wBAAgB,gBAAgB,CAC/B,IAAI,EAAE,UAAU,EAChB,QAAQ,EAAE,MAAM,EAChB,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,gBAAgB,CA4BlB"}
+{"version":3,"file":"bytes.d.ts","sourceRoot":"","sources":["../../typescript/extraction/bytes.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAGH,OAAO,KAAK,EAAE,gBAAgB,IAAI,oBAAoB,EAAE,gBAAgB,EAAE,MAAM,aAAa,CAAC;AAG9F;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAwCG;AACH,wBAAsB,YAAY,CACjC,IAAI,EAAE,UAAU,EAChB,QAAQ,EAAE,MAAM,EAChB,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,OAAO,CAAC,gBAAgB,CAAC,CA4B3B;AAED;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,wBAAgB,gBAAgB,CAC/B,IAAI,EAAE,UAAU,EAChB,QAAQ,EAAE,MAAM,EAChB,MAAM,CAAC,EAAE,oBAAoB,GAAG,IAAI,GAClC,gBAAgB,CA4BlB"}

package/dist/index.js

@@ -606,9 +606,11 @@ async function initWasm(options) {
         }
       }
       if (loadedModule && typeof loadedModule.initialize_pdfium_render === "function") {
-        initializePdfiumAsync(loadedModule).catch((error) => {
+        try {
+          await initializePdfiumAsync(loadedModule);
+        } catch (error) {
           console.warn("PDFium auto-initialization failed (PDF extraction disabled):", error);
-        });
+        }
       }
       setInitialized(true);
       setInitializationError(null);
@@ -1298,6 +1300,161 @@ var TesseractWasmBackend = class {
   }
 };
 
+// typescript/ocr/worker-bridge.ts
+var workerHandle = null;
+var pendingRequests = /* @__PURE__ */ new Map();
+var nextRequestId = 0;
+var workerReady = false;
+var readyResolve = null;
+var readyReject = null;
+var useFallback = false;
+var fallbackFn = null;
+function handleWorkerMessage(msg) {
+  switch (msg["type"]) {
+    case "ready":
+      workerReady = true;
+      readyResolve?.();
+      readyResolve = null;
+      readyReject = null;
+      break;
+    case "init-error":
+      readyReject?.(new Error(msg["error"]));
+      readyResolve = null;
+      readyReject = null;
+      break;
+    case "result": {
+      const id = msg["id"];
+      const pending = pendingRequests.get(id);
+      if (pending) {
+        pendingRequests.delete(id);
+        pending.resolve(msg["text"]);
+      }
+      break;
+    }
+    case "error": {
+      const id = msg["id"];
+      const pending = pendingRequests.get(id);
+      if (pending) {
+        pendingRequests.delete(id);
+        pending.reject(new Error(msg["error"]));
+      }
+      break;
+    }
+  }
+}
+async function createOcrWorker(wasmGluePath, wasmBinary, directFallback) {
+  fallbackFn = directFallback;
+  if (workerHandle) return;
+  const readyPromise = new Promise((resolve, reject) => {
+    readyResolve = resolve;
+    readyReject = reject;
+  });
+  try {
+    if (isNode()) {
+      await createNodeWorker(wasmGluePath, wasmBinary);
+    } else if (typeof Worker !== "undefined") {
+      await createBrowserWorker(wasmGluePath, wasmBinary);
+    } else {
+      useFallback = true;
+      return;
+    }
+    await readyPromise;
+  } catch {
+    workerHandle = null;
+    useFallback = true;
+  }
+}
+async function createNodeWorker(wasmGluePath, wasmBinary) {
+  const { Worker: Worker2 } = await import(
+    /* @vite-ignore */
+    "worker_threads"
+  );
+  const nodePath = await import(
+    /* @vite-ignore */
+    "path"
+  );
+  const nodeUrl = await import(
+    /* @vite-ignore */
+    "url"
+  );
+  const __dirname = nodePath.dirname(nodeUrl.fileURLToPath(import.meta.url));
+  const workerPath = nodePath.join(__dirname, "ocr-worker.js");
+  const worker = new Worker2(workerPath, {
+    workerData: { wasmGluePath, wasmBinary }
+  });
+  worker.on("message", (msg) => handleWorkerMessage(msg));
+  worker.on("error", (err) => {
+    for (const pending of pendingRequests.values()) {
+      pending.reject(err);
+    }
+    pendingRequests.clear();
+    readyReject?.(err);
+  });
+  workerHandle = {
+    postMessage: (data) => worker.postMessage(data),
+    terminate: () => worker.terminate()
+  };
+}
+async function createBrowserWorker(wasmGluePath, wasmBinary) {
+  const workerUrl = new URL("./ocr-worker.js", import.meta.url);
+  const worker = new Worker(workerUrl, { type: "module" });
+  worker.onmessage = (e) => handleWorkerMessage(e.data);
+  worker.onerror = (e) => {
+    const err = new Error(e.message);
+    for (const pending of pendingRequests.values()) {
+      pending.reject(err);
+    }
+    pendingRequests.clear();
+    readyReject?.(err);
+  };
+  workerHandle = {
+    postMessage: (data) => worker.postMessage(data),
+    terminate: () => worker.terminate()
+  };
+  worker.postMessage({
+    type: "init",
+    wasmGluePath,
+    wasmBinary
+  });
+}
+function runOcrInWorker(imageData, tessdata, language) {
+  if (useFallback || !workerHandle || !workerReady) {
+    if (fallbackFn) {
+      try {
+        const text = fallbackFn(imageData, tessdata, language);
+        return Promise.resolve(text);
+      } catch (e) {
+        return Promise.reject(e instanceof Error ? e : new Error(String(e)));
+      }
+    }
+    return Promise.reject(new Error("OCR worker not initialized and no fallback available"));
+  }
+  const id = nextRequestId++;
+  return new Promise((resolve, reject) => {
+    pendingRequests.set(id, { resolve, reject });
+    workerHandle.postMessage({
+      type: "ocr",
+      id,
+      imageData,
+      tessdata,
+      language
+    });
+  });
+}
+async function terminateOcrWorker() {
+  if (workerHandle) {
+    await workerHandle.terminate();
+    workerHandle = null;
+  }
+  workerReady = false;
+  useFallback = false;
+  fallbackFn = null;
+  for (const pending of pendingRequests.values()) {
+    pending.reject(new Error("OCR worker terminated"));
+  }
+  pendingRequests.clear();
+}
+
 // typescript/ocr/enabler.ts
 var TESSDATA_CDN_BASE = "https://raw.githubusercontent.com/tesseract-ocr/tessdata_fast/main";
 var NativeWasmOcrBackend = class {
@@ -1360,19 +1517,47 @@ var NativeWasmOcrBackend = class {
         "Native WASM OCR is not available. Build with the 'ocr-wasm' feature to enable kreuzberg-tesseract."
       );
     }
+    let wasmGluePath;
+    let wasmBinary;
+    if (isNode()) {
+      const nodePath = await import(
+        /* @vite-ignore */
+        "path"
+      );
+      const nodeUrl = await import(
+        /* @vite-ignore */
+        "url"
+      );
+      const nodeFs = await import(
+        /* @vite-ignore */
+        "fs/promises"
+      );
+      const __dirname = nodePath.dirname(nodeUrl.fileURLToPath(import.meta.url));
+      wasmGluePath = nodePath.join(__dirname, "..", "pkg", "kreuzberg_wasm.js");
+      try {
+        const wasmPath = nodePath.join(__dirname, "..", "pkg", "kreuzberg_wasm_bg.wasm");
+        const buf = await nodeFs.readFile(wasmPath);
+        wasmBinary = new Uint8Array(buf);
+      } catch {
+      }
+    } else {
+      wasmGluePath = new URL("./pkg/kreuzberg_wasm.js", import.meta.url).href;
+    }
+    const directFallback = (imageData, tessdata, language) => {
+      if (!wasm2.ocrRecognize) throw new Error("ocrRecognize not available");
+      return wasm2.ocrRecognize(imageData, tessdata, language);
+    };
+    await createOcrWorker(wasmGluePath, wasmBinary, directFallback);
   }
   async shutdown() {
     this.tessdataCache.clear();
     this.progressCallback = null;
+    await terminateOcrWorker();
   }
   setProgressCallback(callback) {
     this.progressCallback = callback;
   }
   async processImage(imageBytes, language) {
-    const wasm2 = getWasmModule();
-    if (!wasm2?.ocrRecognize) {
-      throw new Error("Native WASM OCR function not available");
-    }
     const normalizedLang = language.toLowerCase();
     this.reportProgress(10);
     const tessdata = await this.getTessdata(normalizedLang);
@@ -1388,7 +1573,7 @@ var NativeWasmOcrBackend = class {
       imageData = imageBytes;
     }
     this.reportProgress(50);
-    const text = wasm2.ocrRecognize(imageData, tessdata, normalizedLang);
+    const text = await runOcrInWorker(imageData, tessdata, normalizedLang);
     this.reportProgress(90);
     return {
       content: text,