@dragon708/docmind-browser 1.8.2 → 1.8.4

package/dist/index.d.ts

@@ -45,9 +45,9 @@ type BrowserExtractStructuredDataOptions = BrowserAnalyzeOptions & {
 };
 /**
  * {@link extractMarkdown}: structured options plus `markdown` (passed through to `extractMarkdown` in
- * `@dragon708/docmind-markdown`, including structured-serializer knobs). Binary PDF/DOCX converters inside that
- * package are not used for PDF in-browser and the DOCX bytes→Turndown path is Node-only; the browser still
- * gets correct DOCX Markdown via structured fallback from {@link extractStructuredData}.
+ * `@dragon708/docmind-markdown`). PDF: `convertPdfToMarkdown` / `@cognipeer/to-markdown` never run in-browser — only
+ * structured fallback (empty PDF stub here). DOCX bytes→Mammoth→Turndown is Node-only; the browser still gets
+ * DOCX Markdown via structured fallback from {@link extractStructuredData}.
  */
 interface BrowserExtractMarkdownOptions extends BrowserExtractStructuredDataOptions {
     readonly markdown?: RenderMarkdownOptions;
@@ -126,8 +126,9 @@ declare function extractStructuredData(input: BrowserAnalyzeInput, options?: Bro
  * {@link extractStructuredData} for a full structured envelope, then `extractMarkdown` from
  * `@dragon708/docmind-markdown` on `{ data, filename?, mimeType? }` with that result as `structuredFallback`.
  *
- * - **PDF:** the markdown package does not load `@opendataloader/pdf` here; output comes from the structured
- *   fallback (empty in-browser stub — see {@link getCapabilities}).
+ * - **PDF:** specialized PDF→Markdown (`convertPdfToMarkdown` / `@cognipeer/to-markdown`) is **Node-only** and is
+ *   never loaded in the browser. The markdown package detects PDF, returns `pdf-unsupported-runtime` with clear
+ *   warnings, and uses structured fallback — in this facade that envelope is an **empty PDF stub** (see {@link getCapabilities}).
  * - **DOCX:** the package’s direct bytes → Mammoth → Turndown path is **Node-only**; in-browser, Markdown is
  *   produced via `convertStructuredToMarkdown` on the structured envelope (still Mammoth/OOXML-backed via
  *   `@dragon708/docmind-docx`), with an explanatory warning from the markdown package.
@@ -137,7 +138,8 @@ declare function extractStructuredData(input: BrowserAnalyzeInput, options?: Bro
  */
 declare function extractMarkdown(input: BrowserAnalyzeInput, options?: BrowserExtractMarkdownOptions): Promise<string>;
 /**
- * {@link extractStructuredData} then `renderLlmText` (`@dragon708/docmind-markdown`). For a structured value you already have, that package's `extractLlmContent` matches `renderLlmText` (no file I/O).
+ * {@link extractStructuredData} then `renderLlmText` (`@dragon708/docmind-markdown`). No binary PDF/DOCX Markdown
+ * pipelines run here; for PDF in-browser the envelope is empty and warnings describe the limitation.
  */
 declare function extractLlmContent(input: BrowserAnalyzeInput, options?: BrowserExtractLlmContentOptions): Promise<string>;
 /**
@@ -148,7 +150,7 @@ declare function extractStructuredChunks(input: BrowserAnalyzeInput, options?: B
 
 /** High-level features the user can ask DocMind for (per input kind and runtime). */
 type PublicCapabilityId = "text" | "metadata" | "html" | "ocr" | "pages" | "structured-output"
-/** Browser: {@link extractMarkdown} via `@dragon708/docmind-markdown` `extractMarkdown` + structured fallback (PDF empty; DOCX structured path when binary converter is Node-only). */
+/** Browser: {@link extractMarkdown} via `@dragon708/docmind-markdown` + structured fallback (PDF: no specialized Markdown; DOCX: structured path when bytes→Turndown is Node-only). */
  | "markdown" | "llm-text" | "structured-chunks" | "image-normalization" | "gif-first-frame" | "bmp" | "heic" | "tiff";
 declare function docxIncludeRequested(flags?: AnalyzeDocxIncludeFlags): boolean;
 /** DOCX `word/media` en runtime browser (JSZip; sin pipeline Node). */
@@ -230,9 +232,9 @@ type BrowserExplainAnalysisPlanOptions = Omit<ExplainAnalysisPlanOptions, "inten
 
 /**
  * Epic 1 — **Capabilities:** detects kind from the same hints as `analyzeFile`, then lists which of
- * `text` | `metadata` | `html` | `ocr` | `pages` | `structured-output` | `markdown` (package `extractMarkdown` + structured fallback; PDF empty here) | `llm-text` | `structured-chunks` (split + Markdown sections)
+ * `text` | `metadata` | `html` | `ocr` | `pages` | `structured-output` | `markdown` (hybrid `extractMarkdown`; **no** Node-only PDF `@cognipeer/to-markdown` — PDF Markdown/LLM/chunks unsupported here) | `llm-text` | `structured-chunks`
  * and image-specific ids (`image-normalization`, `bmp`, `gif-first-frame`, `heic`, `tiff`) apply in the browser (PDF always unsupported for meaningful content).
- * No Mammoth/Tesseract/PDF parsing. For DOCX, {@link GetCapabilitiesReport.docxStructure} / `docxEmbeddedImages` describe v2 opt-in features.
+ * No PDF parser; Mammoth/Tesseract apply to DOCX/images only. For DOCX, {@link GetCapabilitiesReport.docxStructure} / `docxEmbeddedImages` describe v2 opt-in features.
  */
 declare function getCapabilities(input: BrowserAnalyzeInput, options?: GetCapabilitiesOptions): Promise<GetCapabilitiesReport>;
 /**

package/dist/index.js

@@ -591,7 +591,7 @@ var TEXT_META_NOTE = "Plain text has no structured document metadata; extractMet
 var OCR_OFF_NOTE = 'Image OCR may be skipped when `ocr.mode` is "off" in analyze options.';
 var STRUCTURED_OCR_OFF = 'Structured image output uses OCR; when `ocr.mode` is "off", `extractStructuredData` returns an empty envelope with a warning.';
 var UNKNOWN_KIND = "Could not determine file kind from name, MIME, or bytes; all features are reported as unsupported until the kind is known.";
-var MARKDOWN_PDF_BROWSER = "PDF: no parser in-browser \u2014 `@opendataloader/pdf` is not loaded here. extractMarkdown still calls `extractMarkdown` in `@dragon708/docmind-markdown`, which falls back to the empty structured stub (same empty Markdown). extractLlmContent / extractStructuredChunks use the structured envelope only. Use @dragon708/docmind-node for real PDF \u2192 Markdown / LLM text / chunks.";
+var MARKDOWN_PDF_BROWSER = "PDF: no specialized PDF\u2192Markdown in-browser (`@cognipeer/to-markdown` / `convertPdfToMarkdown` are Node-only). `extractMarkdown` still calls `@dragon708/docmind-markdown` `extractMarkdown`, which skips the Node pipeline, warns (`pdf-unsupported-runtime`), and uses structured fallback \u2014 here that envelope is empty, so Markdown stays empty. `extractLlmContent` / `extractStructuredChunks` use the same empty structured envelope with browser PDF warnings. Use @dragon708/docmind-node for real PDF extraction and Markdown / LLM text / chunks.";
 var MARKDOWN_IMAGE_OCR_OFF = 'Same as structured-output: when ocr.mode is "off", structured (and thus Markdown/LLM/chunk exports) are empty aside from warnings.';
 function slot(id, supported, warnings) {
   return warnings?.length ? { id, supported, warnings } : { id, supported };
@@ -770,7 +770,7 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
     const structuredLikeIntent = intent === "extractStructuredData" || intent === "extractMarkdown" || intent === "extractLlmContent" || intent === "extractStructuredChunks";
     limitations = lim(
       BROWSER_PDF_UNSUPPORTED_WARNING,
-      structuredLikeIntent ? "`extractStructuredData` / extractMarkdown / extractLlmContent / extractStructuredChunks only see an empty structured envelope in-browser for PDF; use @dragon708/docmind-node for real PDF extraction and Markdown/LLM/chunk exports." : ""
+      structuredLikeIntent ? "`extractStructuredData` / extractMarkdown / extractLlmContent / extractStructuredChunks only see an empty structured envelope in-browser for PDF (no `@cognipeer/to-markdown`; specialized PDF\u2192Markdown runs on Node via @dragon708/docmind-node)." : ""
     );
     nativeExtraction = {
       willAttempt: false,
@@ -953,7 +953,7 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
         limitations = [
           ...limitations,
           ...lim(
-            "Output: Markdown string from `extractMarkdown` in `@dragon708/docmind-markdown`. PDF in-browser: empty (no `@opendataloader/pdf`). DOCX: structured Markdown path when the binary converter is Node-only."
+            "Output: Markdown string from `extractMarkdown` in `@dragon708/docmind-markdown`. PDF in-browser: empty specialized path (no `@cognipeer/to-markdown`; structured fallback only). DOCX: structured Markdown when bytes\u2192Turndown is Node-only."
           )
         ];
       } else if (intent === "extractLlmContent") {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dragon708/docmind-browser",
-  "version": "1.8.2",
+  "version": "1.8.4",
   "description": "Official DocMind browser facade: analyzeFile and intent APIs (DOCX, image OCR, text). PDF and fs paths use @dragon708/docmind-node.",
   "type": "module",
   "sideEffects": false,
@@ -34,7 +34,7 @@
   "license": "MIT",
   "dependencies": {
     "@dragon708/docmind-docx": "^1.8.0",
-    "@dragon708/docmind-markdown": "^1.1.2",
+    "@dragon708/docmind-markdown": "^1.2.1",
     "@dragon708/docmind-ocr": "^1.1.4",
     "@dragon708/docmind-shared": "^1.2.0"
   },