@dragon708/docmind-node 1.12.3 → 1.13.1

package/dist/index.d.ts

@@ -6,7 +6,7 @@ import { AnalyzeDocxIncludeFlags, DocxToHtmlOptions, ExtractStructuredDataFromDo
 export { AnalyzeDocxIncludeFlags, DocxEmbeddedImage, DocxEmbeddedImageConversionOptions, DocxEmbeddedImageConversionResult, DocxEmbeddedImageWebSlice, DocxImageExtractionMode, ExtractImagesFromDocxOptions, ExtractImagesFromDocxResult, ExtractStructuredDataFromDocxOptions, convertDocxEmbeddedImageToWeb, convertDocxImagesForWeb, docxImageIsBrowserRenderable, docxImageRequiresConversion, docxImageSuggestedTargetFormat, docxImageToDataUri, extractImagesFromDocx, extractStructuredDataFromDocx, isNodeJsRuntime } from '@dragon708/docmind-docx';
 import { PdfAnalyzeOptions, ExtractStructuredDataFromPdfOptions } from '@dragon708/docmind-pdf';
 export { ExtractStructuredDataFromPdfIncludeFlags, ExtractStructuredDataFromPdfOptions, extractStructuredDataFromPdf } from '@dragon708/docmind-pdf';
-import { RenderLlmTextOptions, RenderMarkdownOptions, ConvertDocxToMarkdownOptions, ConvertPdfToMarkdownOptions, RenderMarkdownSectionsOptions, MarkdownSection } from '@dragon708/docmind-markdown';
+import { RenderLlmTextOptions, RenderMarkdownOptions, ConvertDocxToMarkdownOptions, ConvertPdfToMarkdownOptions, ExtractMarkdownStrategy, RenderMarkdownSectionsOptions, MarkdownSection } from '@dragon708/docmind-markdown';
 export { MarkdownSection } from '@dragon708/docmind-markdown';
 
 /**
@@ -55,14 +55,22 @@ interface NodeExtractStructuredDataOptions extends DocMindAnalyzeOptions {
 /**
  * {@link extractMarkdown}: inherits {@link NodeExtractStructuredDataOptions}; `markdown` maps to structured-serializer
  * options passed through `extractMarkdown` in `@dragon708/docmind-markdown`. `markdownDocx` / `markdownPdf`
- * configure Mammoth→Turndown and `@opendataloader/pdf` respectively — separate from `docx` / `pdf` used only by {@link extractStructuredData}.
+ * configure Mammoth→Turndown and `@cognipeer/to-markdown` (`convertPdfToMarkdown`) respectively — separate from `docx` / `pdf` used only by {@link extractStructuredData}.
  */
 interface NodeExtractMarkdownOptions extends NodeExtractStructuredDataOptions {
     readonly markdown?: RenderMarkdownOptions;
     /** Forwarded to `extractMarkdown` → `convertDocxToMarkdown` when bytes are DOCX (Node). */
     readonly markdownDocx?: ConvertDocxToMarkdownOptions;
-    /** Forwarded to `extractMarkdown` → `convertPdfToMarkdown` when bytes are PDF. */
+    /** Forwarded to `extractMarkdown` → `convertPdfToMarkdown` when bytes are PDF (Node; `@cognipeer/to-markdown`). */
     readonly markdownPdf?: ConvertPdfToMarkdownOptions;
+    /**
+     * Optional: receive `strategy` and merged `warnings` from `@dragon708/docmind-markdown` `extractMarkdown`
+     * (e.g. `pdf-cognipeer-specialized` vs `pdf-structured-fallback`, `docx-mammoth` vs `docx-structured-fallback`) without changing the `Promise<string>` return type.
+     */
+    readonly onMarkdownExtract?: (info: {
+        strategy: ExtractMarkdownStrategy;
+        warnings: readonly string[];
+    }) => void;
 }
 /**
  * {@link extractLlmContent}: same structured fields; `llm` forwards to `renderLlmText`.
@@ -133,8 +141,14 @@ declare function extractStructuredData(input: NodeAnalyzeInput, options?: NodeEx
 /**
  * End-to-end: {@link extractStructuredData} (for fallback + option parity) plus
  * `extractMarkdown` from `@dragon708/docmind-markdown` on `{ data, filename, mimeType }`.
- * On Node, PDF/DOCX bytes use specialized routes (`@opendataloader/pdf`, Mammoth+Turndown) when detection matches;
- * the structured envelope is always passed as `structuredFallback`.
+ *
+ * On Node, **DOCX** bytes use **Mammoth → Turndown**; **PDF** bytes use **`convertPdfToMarkdown`**
+ * (`@cognipeer/to-markdown`, Node-first, no JVM). The structured envelope is always passed as `structuredFallback`
+ * so specialized routes can fall back to structured → Markdown serialization with clear warnings when needed.
+ *
+ * {@link NodeExtractMarkdownOptions.onMarkdownExtract} receives `strategy` and merged `warnings` from the markdown
+ * package (for example `pdf-cognipeer-specialized`, `pdf-structured-fallback`, `docx-structured-fallback`, and
+ * `[docmind-markdown:extractMarkdown]` trace lines).
  *
  * @param input - Path, buffer, or {@link NodeAnalyzeInput} accepted by the Node facade.
  * @param options - Structured routing (`pdf` / `docx` / `ocr` / `normalize`), optional `markdown` serializer knobs,
@@ -143,7 +157,8 @@ declare function extractStructuredData(input: NodeAnalyzeInput, options?: NodeEx
 declare function extractMarkdown(input: NodeAnalyzeInput, options?: NodeExtractMarkdownOptions): Promise<string>;
 /**
  * {@link extractStructuredData} then `renderLlmText` in `@dragon708/docmind-markdown` (tagged plain text for LLMs).
- * That package's `extractLlmContent` is the same transform on an in-memory structured result only.
+ * This path **does not** run the binary PDF/DOCX Markdown pipelines (no `@cognipeer/to-markdown` / Mammoth→Turndown here);
+ * it linearizes the structured envelope only. Use {@link extractMarkdown} when you need specialized PDF or DOCX Markdown layout.
  *
  * @param options - Structured routing plus optional `llm` passed through to `renderLlmText`.
  */

package/dist/index.js

@@ -620,7 +620,13 @@ async function runOcr(input, options) {
 }
 async function extractMarkdown(input, options) {
   throwIfAborted(options?.signal);
-  const { markdown: markdownOpts, markdownDocx, markdownPdf, ...structuredOpts } = options ?? {};
+  const {
+    markdown: markdownOpts,
+    markdownDocx,
+    markdownPdf,
+    onMarkdownExtract,
+    ...structuredOpts
+  } = options ?? {};
   const resolved = await resolveNodeAnalyzeInput(input);
   const structured = await extractStructuredData(resolved, structuredOpts);
   const data = await bytesFromDetectInput(resolved);
@@ -639,6 +645,7 @@ async function extractMarkdown(input, options) {
       structuredFallback: structured
     }
   );
+  onMarkdownExtract?.({ strategy: r.strategy, warnings: r.warnings });
   return r.markdown;
 }
 async function extractLlmContent(input, options) {
@@ -717,7 +724,7 @@ function buildNodeCapabilityReport(kind) {
           "extractStructuredData: analyzePdf + PDF.js per-page text, outline, links, annotations, merged via normalizeToStructuredResult; respects pdf.ocr like analyzeFile."
         ]),
         slot("markdown", true, [
-          "extractMarkdown: `@dragon708/docmind-markdown` hybrid \u2014 PDF bytes prefer `@opendataloader/pdf` \u2192 Markdown; structured PDF (`extractStructuredData`, respects pdf.ocr) is always built as fallback and for non-binary inputs."
+          "extractMarkdown: `@dragon708/docmind-markdown` hybrid \u2014 PDF bytes use `@cognipeer/to-markdown` (Node-first, no Java); structured PDF (`extractStructuredData`, respects pdf.ocr) is always built as fallback."
         ]),
         slot("llm-text", true, [
           "extractLlmContent: structured envelope \u2192 LLM-oriented plain text (`renderLlmText` in `@dragon708/docmind-markdown`)."
@@ -1040,7 +1047,7 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude, ocrSlic
       if (kind === "pdf") {
         nativeExtraction = {
           willAttempt: true,
-          description: intent === "extractStructuredData" ? "extractStructuredData: analyzePdf plus PDF.js page rows, outline, links, and annotations, normalized to StructuredDocumentResult." : intent === "extractMarkdown" ? "extractMarkdown: structured PDF extract (same as extractStructuredData) for fallback; primary Markdown from `@dragon708/docmind-markdown` tries `@opendataloader/pdf` on PDF bytes when possible." : `${intent}: same structured PDF pipeline as extractStructuredData, then \`@dragon708/docmind-markdown\` (renderLlmText or chunk sections).`
+          description: intent === "extractStructuredData" ? "extractStructuredData: analyzePdf plus PDF.js page rows, outline, links, and annotations, normalized to StructuredDocumentResult." : intent === "extractMarkdown" ? "extractMarkdown: structured PDF extract (same as extractStructuredData) for fallback; primary Markdown on PDF bytes uses `@cognipeer/to-markdown` via `@dragon708/docmind-markdown` (Node-first, no Java)." : `${intent}: same structured PDF pipeline as extractStructuredData, then \`@dragon708/docmind-markdown\` (renderLlmText or chunk sections).`
         };
         ocr = {
           mayUse: pdfOcr !== "off",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dragon708/docmind-node",
-  "version": "1.12.3",
+  "version": "1.13.1",
   "description": "Official DocMind Node facade: analyzeFile, intent APIs, PDF/DOCX/OCR, and fs helpers.",
   "type": "module",
   "main": "./dist/index.js",
@@ -33,7 +33,7 @@
   "license": "MIT",
   "dependencies": {
     "@dragon708/docmind-docx": "^1.8.0",
-    "@dragon708/docmind-markdown": "^1.1.3",
+    "@dragon708/docmind-markdown": "^1.2.1",
     "@dragon708/docmind-ocr": "^1.1.4",
     "@dragon708/docmind-pdf": "^2.2.0",
     "@dragon708/docmind-shared": "^1.2.0"