@heripo/document-processor 0.1.3 → 0.1.4

package/dist/index.cjs

@@ -1908,11 +1908,10 @@ var TocExtractor = class extends TextLLMComponent {
   async extract(markdown) {
     this.log("info", `Starting TOC extraction (${markdown.length} chars)`);
     if (!markdown.trim()) {
-      this.log("info", "Empty markdown, returning empty array");
-      return {
-        entries: [],
-        usage: this.createEmptyUsage("extraction")
-      };
+      this.log("error", "Cannot extract TOC from empty markdown content");
+      throw new TocParseError(
+        "TOC extraction failed: provided markdown content is empty"
+      );
     }
     try {
       const result = await this.callTextLLM(
@@ -4051,10 +4050,13 @@ var DocumentProcessor = class {
       const totalPages = Object.keys(doclingDoc.pages).length;
       markdown = await this.visionTocExtractor.extract(totalPages);
       if (!markdown) {
-        this.logger.warn(
-          "[DocumentProcessor] TOC not found in any method, returning empty"
+        const reason = "Both rule-based search and vision fallback failed to locate TOC";
+        this.logger.error(
+          `[DocumentProcessor] TOC extraction failed: ${reason}`
+        );
+        throw new TocNotFoundError(
+          `Table of contents not found in the document. ${reason}.`
         );
-        return [];
       }
       this.logger.info(
         `[DocumentProcessor] Vision extracted TOC markdown (${markdown.length} chars)`
@@ -4062,6 +4064,11 @@ var DocumentProcessor = class {
     }
     const tocResult = await this.tocExtractor.extract(markdown);
     this.usageAggregator.track(tocResult.usage);
+    if (tocResult.entries.length === 0) {
+      const reason = "TOC area was detected but LLM could not extract any structured entries";
+      this.logger.error(`[DocumentProcessor] TOC extraction failed: ${reason}`);
+      throw new TocNotFoundError(`${reason}.`);
+    }
     this.logger.info(
       `[DocumentProcessor] Extracted ${tocResult.entries.length} top-level TOC entries`
     );
@@ -4301,21 +4308,14 @@ var DocumentProcessor = class {
    * Convert chapters and link resources
    *
    * Generates chapters based on TOC and links images/tables/footnotes using ChapterConverter.
-   * Falls back to single "Document" chapter when TOC is empty.
+   * Throws TocNotFoundError if TOC entries are empty (defensive assertion).
    */
   async convertChapters(doclingDoc, tocEntries, pageRangeMap, images, tables, footnotes) {
     this.logger.info("[DocumentProcessor] Converting chapters...");
     if (tocEntries.length === 0) {
-      this.logger.info(
-        "[DocumentProcessor] No TOC entries, creating fallback chapter"
-      );
-      return this.createFallbackChapter(
-        doclingDoc,
-        pageRangeMap,
-        images,
-        tables,
-        footnotes
-      );
+      const reason = "Cannot convert chapters without TOC entries";
+      this.logger.error(`[DocumentProcessor] ${reason}`);
+      throw new TocNotFoundError(reason);
     }
     const chapters = this.chapterConverter.convert(
       tocEntries,
@@ -4330,48 +4330,6 @@ var DocumentProcessor = class {
     );
     return chapters;
   }
-  /**
-   * Create a fallback chapter when TOC is not available
-   *
-   * Creates a single "Document" chapter containing all text blocks,
-   * images, tables, and footnotes from the document.
-   */
-  createFallbackChapter(doclingDoc, pageRangeMap, images, tables, footnotes) {
-    const textBlocks = doclingDoc.texts.filter(
-      (item) => item.label !== "footnote" && this.textCleaner.isValidText(item.text)
-    ).map((item) => ({
-      text: this.textCleaner.normalize(item.text),
-      pdfPageNo: item.prov?.[0]?.page_no ?? 1
-    }));
-    if (textBlocks.length === 0 && images.length === 0 && tables.length === 0 && footnotes.length === 0) {
-      this.logger.info(
-        "[DocumentProcessor] No content found for fallback chapter"
-      );
-      return [];
-    }
-    const firstPdfPage = Math.min(
-      ...Object.keys(pageRangeMap).map(Number).filter((n) => !isNaN(n)),
-      1
-    );
-    const firstPageRange = pageRangeMap[firstPdfPage];
-    const pageNo = firstPageRange?.startPageNo ?? 1;
-    const fallbackChapter = {
-      id: this.idGenerator.generateChapterId(),
-      originTitle: "Document",
-      title: "Document",
-      pageNo,
-      level: 1,
-      textBlocks,
-      imageIds: images.map((img) => img.id),
-      tableIds: tables.map((tbl) => tbl.id),
-      footnoteIds: footnotes.map((ftn) => ftn.id),
-      children: []
-    };
-    this.logger.info(
-      `[DocumentProcessor] Created fallback chapter with ${textBlocks.length} text blocks, ${images.length} images, ${tables.length} tables, ${footnotes.length} footnotes`
-    );
-    return [fallbackChapter];
-  }
 };
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {