@heripo/pdf-parser 0.1.12 → 0.1.14

package/dist/index.cjs

@@ -1780,7 +1780,8 @@ var VlmTextCorrector = class {
               },
               {
                 type: "image",
-                image: `data:image/png;base64,${imageBase64}`
+                image: imageBase64,
+                mediaType: "image/png"
               }
             ]
           }
@@ -1992,7 +1993,7 @@ var VlmTextCorrector = class {
    */
   readPageImage(outputDir, pageNo) {
     const imagePath = (0, import_node_path4.join)(outputDir, "pages", `page_${pageNo - 1}.png`);
-    return (0, import_node_fs4.readFileSync)(imagePath).toString("base64");
+    return new Uint8Array((0, import_node_fs4.readFileSync)(imagePath));
   }
   /**
    * Apply VLM corrections to the DoclingDocument.
@@ -2069,8 +2070,9 @@ Note: Hanja are Chinese characters used in Korean documents, different from mode
 
 Answer whether any Hanja characters are present on this page.
 
-Also identify all languages present on this page. Return an array of BCP 47 language tags ordered by prevalence (primary language first).
-Examples: ["ko-KR", "en-US"], ["ja-JP"], ["zh-TW", "en-US"]`;
+Also identify all languages present on this page. Return an array of ocrmac-compatible language tags ordered by prevalence (primary language first).
+Supported tags: ar-SA, ars-SA, cs-CZ, da-DK, de-DE, en-US, es-ES, fr-FR, id-ID, it-IT, ja-JP, ko-KR, ms-MY, nb-NO, nl-NL, nn-NO, no-NO, pl-PL, pt-BR, ro-RO, ru-RU, sv-SE, th-TH, tr-TR, uk-UA, vi-VT, yue-Hans, yue-Hant, zh-Hans, zh-Hant.
+Examples: ["ko-KR", "en-US"], ["ja-JP"], ["zh-Hant", "en-US"]`;
 var OcrStrategySampler = class {
   logger;
   pageRenderer;
@@ -2261,7 +2263,7 @@ var OcrStrategySampler = class {
     this.logger.debug(
       `[OcrStrategySampler] Analyzing page ${pageNo} for Korean-Hanja mix and language...`
     );
-    const base64Image = (0, import_node_fs5.readFileSync)(pageFile).toString("base64");
+    const imageData = new Uint8Array((0, import_node_fs5.readFileSync)(pageFile));
     const messages = [
       {
         role: "user",
@@ -2269,7 +2271,8 @@ var OcrStrategySampler = class {
           { type: "text", text: KOREAN_HANJA_MIX_PROMPT },
           {
             type: "image",
-            image: `data:image/png;base64,${base64Image}`
+            image: imageData,
+            mediaType: "image/png"
           }
         ]
       }
@@ -2367,6 +2370,36 @@ var LocalFileServer = class {
   }
 };
 
+// src/utils/task-failure-details.ts
+var MAX_RESULT_RETRIES = 3;
+var RESULT_RETRY_DELAY_MS = 2e3;
+async function getTaskFailureDetails(task, logger, logPrefix) {
+  for (let attempt = 0; attempt < MAX_RESULT_RETRIES; attempt++) {
+    try {
+      if (attempt > 0) {
+        await new Promise((r) => setTimeout(r, RESULT_RETRY_DELAY_MS));
+      }
+      const result = await task.getResult();
+      if (result.errors?.length) {
+        return result.errors.map((e) => e.message).join("; ");
+      }
+      return `status: ${result.status ?? "unknown"}`;
+    } catch (err) {
+      if (attempt === MAX_RESULT_RETRIES - 1) {
+        logger.error(
+          `${logPrefix} Failed to retrieve task result after ${MAX_RESULT_RETRIES} attempts:`,
+          err
+        );
+        return "unable to retrieve error details";
+      }
+      logger.warn(
+        `${logPrefix} Result not available yet, retrying (${attempt + 1}/${MAX_RESULT_RETRIES})...`
+      );
+    }
+  }
+  return "unable to retrieve error details";
+}
+
 // src/core/chunked-pdf-converter.ts
 var import_node_fs7 = require("fs");
 var import_promises4 = require("fs/promises");
@@ -2712,14 +2745,15 @@ var ChunkedPDFConverter = class {
       const status = await task.poll();
       if (status.task_status === "success") return;
       if (status.task_status === "failure") {
-        let details = "unknown";
-        try {
-          const result = await task.getResult();
-          if (result.errors?.length) {
-            details = result.errors.map((e) => e.message).join("; ");
-          }
-        } catch {
-        }
+        const elapsed = ((Date.now() - startTime) / 1e3).toFixed(1);
+        this.logger.error(
+          `[ChunkedPDFConverter] Task ${task.taskId} failed after ${elapsed}s`
+        );
+        const details = await getTaskFailureDetails(
+          task,
+          this.logger,
+          "[ChunkedPDFConverter]"
+        );
         throw new Error(`[ChunkedPDFConverter] Chunk task failed: ${details}`);
       }
       await new Promise(
@@ -3356,6 +3390,7 @@ var PDFConverter = class {
     return {
       ...(0, import_es_toolkit.omit)(options, [
         "num_threads",
+        "document_timeout",
         "forceImagePdf",
         "strategySamplerModel",
         "vlmProcessorModel",
@@ -3377,6 +3412,8 @@ var PDFConverter = class {
         framework: "livetext"
       },
       generate_picture_images: true,
+      do_picture_classification: true,
+      do_picture_description: true,
       generate_page_images: false,
       // Page images are rendered by PageRenderer (ImageMagick) after conversion
       images_scale: 2,
@@ -3391,6 +3428,9 @@ var PDFConverter = class {
       accelerator_options: {
         device: "mps",
         num_threads: options.num_threads
+      },
+      ...options.document_timeout !== void 0 && {
+        document_timeout: options.document_timeout
       }
     };
   }
@@ -3477,16 +3517,7 @@ var PDFConverter = class {
    * Fetch detailed error information from a failed task result.
    */
   async getTaskFailureDetails(task) {
-    try {
-      const result = await task.getResult();
-      if (result.errors?.length) {
-        return result.errors.map((e) => e.message).join("; ");
-      }
-      return `status: ${result.status ?? "unknown"}`;
-    } catch (err) {
-      this.logger.error("[PDFConverter] Failed to retrieve task result:", err);
-      return "unable to retrieve error details";
-    }
+    return getTaskFailureDetails(task, this.logger, "[PDFConverter]");
   }
   async downloadResult(taskId) {
     this.logger.info(