@dragon708/docmind-browser 1.8.3 → 1.8.5

package/dist/index.d.ts

@@ -1,7 +1,7 @@
 import { DocMindAnalyzeOptions, AnalyzeFileOutputOptions, NormalizeStructuredOptions, AnalysisResult, StructuredDocumentResult, FileKind, RuntimeDescriptor, DocMindPublicIntent, AnalysisAnalyzer, ProcessingPlanDescriptor, ExplainAnalysisPlanOptions, GetCapabilitiesOptions } from '@dragon708/docmind-shared';
 export { AnalysisAnalyzer, AnalysisResult, CapabilityDescriptor, DetectFileKindInput, DocMindPublicIntent, DocumentBlock, DocumentImageRef, DocumentPage, DocumentTable, DocxAnalysisCoreResult, ExplainAnalysisPlanOptions, ExplainAnalysisPlanResult, FileKind, FileKindMetadata, GenericAnalysisResult, GetCapabilitiesOptions, GetCapabilitiesResult, ImageAnalysisCoreResult, PdfAnalysisCoreResult, StructuredDocumentResult, TextAnalysisResult, analyzeFileRequestsStructured, detectFileKind, isStructuredDocumentResult } from '@dragon708/docmind-shared';
-import { RenderLlmTextOptions, RenderMarkdownOptions, RenderMarkdownSectionsOptions, MarkdownSection } from '@dragon708/docmind-markdown';
-export { MarkdownSection } from '@dragon708/docmind-markdown';
+import { RenderLlmTextOptions, RenderMarkdownOptions, ConvertHtmlToMarkdownOptions, ConvertCsvToMarkdownOptions, ConvertSpreadsheetToMarkdownOptions, ExtractMarkdownStrategy, ExtractMarkdownRoutingInfo, RenderMarkdownSectionsOptions, MarkdownSection } from '@dragon708/docmind-markdown';
+export { ConvertCsvToMarkdownOptions, ConvertCsvToMarkdownResult, ConvertHtmlToMarkdownOptions, ConvertHtmlToMarkdownResult, ConvertSpreadsheetToMarkdownOptions, ConvertSpreadsheetToMarkdownResult, CsvStringInputMode, DetectedBinaryFormat, ExtractMarkdownMediaHint, ExtractMarkdownRoutingInfo, ExtractMarkdownStrategy, HtmlStringInputMode, MarkdownSection, convertCsvToMarkdown, convertHtmlToMarkdown, convertSpreadsheetToMarkdown, detectBinaryFormat } from '@dragon708/docmind-markdown';
 import { AnalyzeDocxIncludeFlags, DocxToHtmlOptions } from '@dragon708/docmind-docx';
 export { AnalyzeDocxIncludeFlags, ExtractStructuredDataFromDocxOptions, extractStructuredDataFromDocx } from '@dragon708/docmind-docx';
 import { OcrOptions, OcrTiffOptions, PreprocessImageOptions } from '@dragon708/docmind-ocr';
@@ -45,12 +45,26 @@ 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`). **Cognipeer** (`convertPdfToMarkdown`, `convertHtmlToMarkdown`, `convertCsvToMarkdown`,
+ * `convertSpreadsheetToMarkdown`) is **Node-only**; in-browser those routes yield `*-unsupported-runtime` and use
+ * structured fallback. DOCX bytes→Mammoth→Turndown is also Node-only; the browser still gets DOCX Markdown via structured
+ * serialization. Optional `markdownHtml` / `markdownCsv` / `markdownSpreadsheet` forward to the same package for API parity
+ * with `@dragon708/docmind-node` `NodeExtractMarkdownOptions`.
  */
 interface BrowserExtractMarkdownOptions extends BrowserExtractStructuredDataOptions {
     readonly markdown?: RenderMarkdownOptions;
+    readonly markdownHtml?: ConvertHtmlToMarkdownOptions;
+    readonly markdownCsv?: ConvertCsvToMarkdownOptions;
+    readonly markdownSpreadsheet?: ConvertSpreadsheetToMarkdownOptions;
+    /**
+     * Optional: `strategy`, merged `warnings`, and `routing` from `@dragon708/docmind-markdown` `extractMarkdown`
+     * (e.g. `html-unsupported-runtime` + `routingSummary` in the browser).
+     */
+    readonly onMarkdownExtract?: (info: {
+        strategy: ExtractMarkdownStrategy;
+        warnings: readonly string[];
+        routing?: ExtractMarkdownRoutingInfo;
+    }) => void;
 }
 /** {@link extractLlmContent}: optional `llm` passed to `renderLlmText`. */
 interface BrowserExtractLlmContentOptions extends BrowserExtractStructuredDataOptions {
@@ -126,18 +140,22 @@ 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 / HTML / CSV / Excel:** Cognipeer-backed specialized routes are **Node-only**. The markdown package detects
+ *   format, returns `pdf-unsupported-runtime` / `html-unsupported-runtime` / etc. with warnings, and uses structured fallback
+ *   — for PDF in this facade the envelope is an **empty stub** (see {@link getCapabilities}); for HTML/CSV/XLSX classified as
+ *   `text`, structured fallback still yields Markdown from UTF-8 rollup when applicable.
  * - **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.
  * - **Text / image:** unidentified or non-binary bytes use the same structured serializer.
  *
- * @param options - `markdown` options plus the same routing as {@link extractStructuredData} (`ocr`, `docx`, `normalize`).
+ * @param options - `markdown` and optional `markdownHtml` / `markdownCsv` / `markdownSpreadsheet` (forwarded to the package),
+ *   plus the same routing as {@link extractStructuredData} (`ocr`, `docx`, `normalize`).
  */
 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 +166,7 @@ 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 +248,9 @@ 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** Cognipeer PDF/HTML/CSV/Excel — specialized Markdown unsupported here; use `@dragon708/docmind-node`) | `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

@@ -5,6 +5,7 @@ export { extractStructuredDataFromDocx } from '@dragon708/docmind-docx';
 import { extractStructuredDataFromImage, preprocessHasEffect, resolveImageFormat, normalizeImageForOcr, ocrTiff, ocrImageDetailed } from '@dragon708/docmind-ocr';
 export { extractStructuredDataFromImage } from '@dragon708/docmind-ocr';
 import { extractMarkdown as extractMarkdown$1, renderLlmText, renderMarkdownSections } from '@dragon708/docmind-markdown';
+export { convertCsvToMarkdown, convertHtmlToMarkdown, convertSpreadsheetToMarkdown, detectBinaryFormat } from '@dragon708/docmind-markdown';
 
 // src/analyzeFile.ts
 function assertBrowserInput(input) {
@@ -531,14 +532,28 @@ function browserFileHints(input) {
 }
 async function extractMarkdown(input, options) {
   throwIfAborted(options?.signal);
-  const { markdown: markdownOpts, ...structuredOpts } = options ?? {};
+  const {
+    markdown: markdownOpts,
+    markdownHtml,
+    markdownCsv,
+    markdownSpreadsheet,
+    onMarkdownExtract,
+    ...structuredOpts
+  } = options ?? {};
   const structured = await extractStructuredData(input, structuredOpts);
   const data = await toUint8Array(input);
   const hints = browserFileHints(input);
   const r = await extractMarkdown$1(
     { data, filename: hints.filename, mimeType: hints.mimeType },
-    { ...markdownOpts ?? {}, structuredFallback: structured }
+    {
+      ...markdownOpts ?? {},
+      ...markdownHtml !== void 0 ? { html: markdownHtml } : {},
+      ...markdownCsv !== void 0 ? { csv: markdownCsv } : {},
+      ...markdownSpreadsheet !== void 0 ? { spreadsheet: markdownSpreadsheet } : {},
+      structuredFallback: structured
+    }
   );
+  onMarkdownExtract?.({ strategy: r.strategy, warnings: r.warnings, routing: r.routing });
   return r.markdown;
 }
 async function extractLlmContent(input, options) {
@@ -591,7 +606,8 @@ 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_COGNIPEER_HTML_CSV_XLSX_BROWSER = "HTML / CSV / Excel: `convertHtmlToMarkdown`, `convertCsvToMarkdown`, and `convertSpreadsheetToMarkdown` (same Cognipeer stack as PDF) are Node-only. In-browser `extractMarkdown` detects those formats, warns (`html-unsupported-runtime`, `csv-unsupported-runtime`, `spreadsheet-unsupported-runtime`), and uses structured fallback (UTF-8 text rollup for `text` kind, or OCR/layout for images). Use @dragon708/docmind-node for specialized conversion.";
 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 };
@@ -703,7 +719,8 @@ function buildBrowserCapabilityReport(kind) {
           "`extractStructuredData` decodes UTF-8 (via `analyzeText`) and normalizes to `StructuredDocumentResult` (paragraph block rollup)."
         ]),
         slot("markdown", true, [
-          "extractMarkdown: bytes + structured fallback through `@dragon708/docmind-markdown` `extractMarkdown` (typically structured serializer for UTF-8 text)."
+          "extractMarkdown: bytes + structured fallback through `@dragon708/docmind-markdown` `extractMarkdown` (typically structured serializer for UTF-8 text).",
+          MARKDOWN_COGNIPEER_HTML_CSV_XLSX_BROWSER
         ]),
         slot("llm-text", true, [
           "extractLlmContent: UTF-8 structured rollup \u2192 `renderLlmText` in `@dragon708/docmind-markdown`."
@@ -770,7 +787,7 @@ 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 +970,7 @@ 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`. Cognipeer (PDF/HTML/CSV/Excel) is Node-only in-browser \u2014 `*-unsupported-runtime` + structured fallback. 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": "1.8.3",
+  "version": "1.8.5",
   "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": "^1.1.3",
+    "@dragon708/docmind-markdown": "^1.2.8",
     "@dragon708/docmind-ocr": "^1.1.4",
     "@dragon708/docmind-shared": "^1.2.0"
   },