@heripo/document-processor 0.1.6 → 0.1.7

package/dist/index.d.cts

@@ -250,6 +250,14 @@ declare class DocumentProcessor {
      * 5. TocExtractor - LLM-based structured extraction
      */
     private extractTableOfContents;
+    /**
+     * Extract the maximum page number from TOC markdown
+     *
+     * Parses page numbers from dot-leader patterns (e.g., "..... 175")
+     * to detect compiled volume scenarios where TOC page numbers exceed
+     * the sub-document's page count.
+     */
+    private extractMaxPageNumber;
     /**
      * Process resource captions (for images and tables)
      *

package/dist/index.d.ts

@@ -250,6 +250,14 @@ declare class DocumentProcessor {
      * 5. TocExtractor - LLM-based structured extraction
      */
     private extractTableOfContents;
+    /**
+     * Extract the maximum page number from TOC markdown
+     *
+     * Parses page numbers from dot-leader patterns (e.g., "..... 175")
+     * to detect compiled volume scenarios where TOC page numbers exceed
+     * the sub-document's page count.
+     */
+    private extractMaxPageNumber;
     /**
      * Process resource captions (for images and tables)
      *

package/dist/index.js

@@ -4464,8 +4464,10 @@ var DocumentProcessor = class {
       );
     }
     const totalPages = Object.keys(doclingDoc.pages).length;
+    const maxTocPageNo = this.extractMaxPageNumber(markdown);
+    const effectiveTotalPages = maxTocPageNo > totalPages ? void 0 : totalPages;
     const tocResult = await this.tocExtractor.extract(markdown, {
-      totalPages
+      totalPages: effectiveTotalPages
     });
     for (const usage of tocResult.usages) {
       this.usageAggregator.track(usage);
@@ -4480,6 +4482,20 @@ var DocumentProcessor = class {
     );
     return tocResult.entries;
   }
+  /**
+   * Extract the maximum page number from TOC markdown
+   *
+   * Parses page numbers from dot-leader patterns (e.g., "..... 175")
+   * to detect compiled volume scenarios where TOC page numbers exceed
+   * the sub-document's page count.
+   */
+  extractMaxPageNumber(markdown) {
+    const matches = [...markdown.matchAll(/\.{2,}\s*(\d+)/g)];
+    if (matches.length === 0) {
+      return 0;
+    }
+    return Math.max(...matches.map((m) => parseInt(m[1], 10)));
+  }
   /**
    * Process resource captions (for images and tables)
    *