@dragon708/docmind-browser 1.6.0 → 1.8.0

package/dist/index.d.ts

@@ -1,9 +1,11 @@
-import { DocMindAnalyzeOptions, AnalyzeFileOutputOptions, AnalysisResult, NormalizeStructuredOptions, StructuredDocumentResult, FileKind, RuntimeDescriptor, DocMindPublicIntent, AnalysisAnalyzer, ProcessingPlanDescriptor, ExplainAnalysisPlanOptions, GetCapabilitiesOptions } from '@dragon708/docmind-shared';
+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 { OcrOptions, OcrTiffOptions, PreprocessImageOptions } from '@dragon708/docmind-ocr';
-export { ExtractStructuredDataFromImageOptions, extractStructuredDataFromImage } from '@dragon708/docmind-ocr';
+import { RenderLlmTextOptions, RenderMarkdownOptions, RenderMarkdownSectionsOptions, MarkdownSection } from '@dragon708/docmind-markdown';
+export { MarkdownSection } 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';
+export { ExtractStructuredDataFromImageOptions, extractStructuredDataFromImage } from '@dragon708/docmind-ocr';
 
 /**
  * Opciones DOCX para el facade browser (Mammoth + inclusiones v2 de `@dragon708/docmind-docx`; sin APIs Node-only).
@@ -37,6 +39,27 @@ interface BrowserAnalyzeOptions extends DocMindAnalyzeOptions, AnalyzeFileOutput
     /** Solo DOCX: ver {@link BrowserAnalyzeDocxOptionsSlice}. */
     readonly docx?: BrowserAnalyzeDocxOptionsSlice;
 }
+/** Options for {@link extractStructuredData}: same as {@link BrowserAnalyzeOptions} plus shared normalize knobs. */
+type BrowserExtractStructuredDataOptions = BrowserAnalyzeOptions & {
+    readonly normalize?: NormalizeStructuredOptions;
+};
+/**
+ * {@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}.
+ */
+interface BrowserExtractMarkdownOptions extends BrowserExtractStructuredDataOptions {
+    readonly markdown?: RenderMarkdownOptions;
+}
+/** {@link extractLlmContent}: optional `llm` passed to `renderLlmText`. */
+interface BrowserExtractLlmContentOptions extends BrowserExtractStructuredDataOptions {
+    readonly llm?: RenderLlmTextOptions;
+}
+/** {@link extractStructuredChunks}: optional `chunks` (split / section sizing). */
+interface BrowserExtractStructuredChunksOptions extends BrowserExtractStructuredDataOptions {
+    readonly chunks?: RenderMarkdownSectionsOptions;
+}
 
 /**
  * Inputs supported by the browser entry (DOM types only — no `fs`, no Node `Buffer` in the public surface).
@@ -90,10 +113,6 @@ declare function runOcr(input: BrowserAnalyzeInput, options?: BrowserAnalyzeOpti
  * PDF is not supported in this runtime (clear warnings, no PDF package import).
  */
 
-/** Options for {@link extractStructuredData}: same as {@link BrowserAnalyzeOptions} plus shared normalize knobs. */
-type BrowserExtractStructuredDataOptions = BrowserAnalyzeOptions & {
-    readonly normalize?: NormalizeStructuredOptions;
-};
 /**
  * Returns a {@link StructuredDocumentResult} for inputs the browser runtime actually supports:
  * **DOCX** (`extractStructuredDataFromDocx`), **images** (`extractStructuredDataFromImage` when OCR is not off),
@@ -103,8 +122,34 @@ type BrowserExtractStructuredDataOptions = BrowserAnalyzeOptions & {
  */
 declare function extractStructuredData(input: BrowserAnalyzeInput, options?: BrowserExtractStructuredDataOptions): Promise<StructuredDocumentResult>;
 
+/**
+ * {@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}).
+ * - **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`).
+ */
+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).
+ */
+declare function extractLlmContent(input: BrowserAnalyzeInput, options?: BrowserExtractLlmContentOptions): Promise<string>;
+/**
+ * Structured extract → `renderMarkdownSections` (`splitStructuredIntoChunks` with Markdown; same as
+ * `extractStructuredChunks` alias in `@dragon708/docmind-markdown`).
+ */
+declare function extractStructuredChunks(input: BrowserAnalyzeInput, options?: BrowserExtractStructuredChunksOptions): Promise<readonly MarkdownSection[]>;
+
 /** High-level features the user can ask DocMind for (per input kind and runtime). */
-type PublicCapabilityId = "text" | "metadata" | "html" | "ocr" | "pages" | "structured-output" | "image-normalization" | "gif-first-frame" | "bmp" | "heic" | "tiff";
+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). */
+ | "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). */
 interface DocxEmbeddedImageCapabilities {
@@ -176,12 +221,17 @@ interface ExplainAnalysisPlanReport {
     readonly warnings?: readonly string[];
 }
 
+/** Browser facade intents that run `@dragon708/docmind-markdown` after structured extraction. */
+type BrowserMarkdownFacadeIntent = "extractMarkdown" | "extractLlmContent" | "extractStructuredChunks";
 /** Options for {@link explainAnalysisPlan}: shared fields plus optional `ocr` / `docx` for accurate step preview. */
-type BrowserExplainAnalysisPlanOptions = ExplainAnalysisPlanOptions & Pick<BrowserAnalyzeOptions, "ocr" | "docx" | "structuredOutput" | "output">;
+type BrowserExplainAnalysisPlanOptions = Omit<ExplainAnalysisPlanOptions, "intent"> & Pick<BrowserAnalyzeOptions, "ocr" | "docx" | "structuredOutput" | "output"> & {
+    readonly intent?: DocMindPublicIntent | BrowserMarkdownFacadeIntent;
+};
 
 /**
  * Epic 1 — **Capabilities:** detects kind from the same hints as `analyzeFile`, then lists which of
- * `text` | `metadata` | `html` | `ocr` | `pages` | `structured-output` and image-specific ids (`image-normalization`, `bmp`, `gif-first-frame`, `heic`, `tiff`) apply in the browser (PDF always unsupported).
+ * `text` | `metadata` | `html` | `ocr` | `pages` | `structured-output` | `markdown` (package `extractMarkdown` + structured fallback; PDF empty here) | `llm-text` | `structured-chunks` (split + Markdown sections)
+ * 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.
  */
 declare function getCapabilities(input: BrowserAnalyzeInput, options?: GetCapabilitiesOptions): Promise<GetCapabilitiesReport>;
@@ -191,4 +241,4 @@ declare function getCapabilities(input: BrowserAnalyzeInput, options?: GetCapabi
  */
 declare function explainAnalysisPlan(input: BrowserAnalyzeInput, options?: BrowserExplainAnalysisPlanOptions): Promise<ExplainAnalysisPlanReport>;
 
-export { BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING, BROWSER_PDF_UNSUPPORTED_WARNING, type BrowserAnalyzeDocxOptionsSlice, type BrowserAnalyzeInput, type BrowserAnalyzeOptions, type BrowserExplainAnalysisPlanOptions, type BrowserExtractStructuredDataOptions, type BrowserOcrMode, type BrowserOcrOptions, DOCX_EMBEDDED_IMAGE_CAPABILITIES_BROWSER, DOCX_STRUCTURE_CAPABILITIES_BROWSER, type DocxEmbeddedImageCapabilities, type DocxStructuralCapabilities, type ExplainAnalysisPlanReport, type GetCapabilitiesReport, type NativeExtractionPlan, type OcrPlan, type PublicCapabilityId, type PublicCapabilitySupport, analyzeFile, convertToHtml, docxIncludeRequested, explainAnalysisPlan, extractMetadata, extractStructuredData, extractText, getCapabilities, runOcr };
+export { BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING, BROWSER_PDF_UNSUPPORTED_WARNING, type BrowserAnalyzeDocxOptionsSlice, type BrowserAnalyzeInput, type BrowserAnalyzeOptions, type BrowserExplainAnalysisPlanOptions, type BrowserExtractLlmContentOptions, type BrowserExtractMarkdownOptions, type BrowserExtractStructuredChunksOptions, type BrowserExtractStructuredDataOptions, type BrowserMarkdownFacadeIntent, type BrowserOcrMode, type BrowserOcrOptions, DOCX_EMBEDDED_IMAGE_CAPABILITIES_BROWSER, DOCX_STRUCTURE_CAPABILITIES_BROWSER, type DocxEmbeddedImageCapabilities, type DocxStructuralCapabilities, type ExplainAnalysisPlanReport, type GetCapabilitiesReport, type NativeExtractionPlan, type OcrPlan, type PublicCapabilityId, type PublicCapabilitySupport, analyzeFile, convertToHtml, docxIncludeRequested, explainAnalysisPlan, extractLlmContent, extractMarkdown, extractMetadata, extractStructuredChunks, extractStructuredData, extractText, getCapabilities, runOcr };

package/dist/index.js

@@ -4,6 +4,7 @@ import { extractStructuredDataFromDocx, analyzeDocx } from '@dragon708/docmind-d
 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';
 
 // src/analyzeFile.ts
 function assertBrowserInput(input) {
@@ -519,6 +520,39 @@ async function runOcr(input, options) {
       return notImplementedResult(kind, "none", [UNKNOWN_FORMAT_WARNING]);
   }
 }
+function browserFileHints(input) {
+  if (input instanceof File) {
+    return {
+      filename: input.name,
+      mimeType: input.type ? input.type : void 0
+    };
+  }
+  return {};
+}
+async function extractMarkdown(input, options) {
+  throwIfAborted(options?.signal);
+  const { markdown: markdownOpts, ...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 }
+  );
+  return r.markdown;
+}
+async function extractLlmContent(input, options) {
+  throwIfAborted(options?.signal);
+  const { llm: llmOpts, ...structuredOpts } = options ?? {};
+  const structured = await extractStructuredData(input, structuredOpts);
+  return renderLlmText(structured, llmOpts);
+}
+async function extractStructuredChunks(input, options) {
+  throwIfAborted(options?.signal);
+  const { chunks: chunkOpts, ...structuredOpts } = options ?? {};
+  const structured = await extractStructuredData(input, structuredOpts);
+  return renderMarkdownSections(structured, chunkOpts);
+}
 
 // src/capabilityReport.ts
 function docxIncludeRequested(flags) {
@@ -557,6 +591,8 @@ 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_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 };
 }
@@ -573,7 +609,10 @@ function buildBrowserCapabilityReport(kind) {
         slot("html", false, [pdf]),
         slot("ocr", false, [pdf]),
         slot("pages", false, [pdf]),
-        slot("structured-output", false, [BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING])
+        slot("structured-output", false, [BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING]),
+        slot("markdown", false, [MARKDOWN_PDF_BROWSER]),
+        slot("llm-text", false, [MARKDOWN_PDF_BROWSER]),
+        slot("structured-chunks", false, [MARKDOWN_PDF_BROWSER])
       ];
       break;
     case "docx":
@@ -593,6 +632,15 @@ function buildBrowserCapabilityReport(kind) {
         ]),
         slot("structured-output", true, [
           "`extractStructuredData` uses `@dragon708/docmind-docx` (Mammoth + OOXML) and returns `StructuredDocumentResult`; optional `options.docx` slices are forwarded."
+        ]),
+        slot("markdown", true, [
+          "extractMarkdown: `@dragon708/docmind-markdown` `extractMarkdown` on bytes + structured fallback. The package\u2019s DOCX-bytes Mammoth\u2192Turndown path is Node-only; in-browser, Markdown is produced from `extractStructuredData` (Mammoth/OOXML in `@dragon708/docmind-docx`) via structured serialization, with a clear package warning that the binary shortcut is skipped."
+        ]),
+        slot("llm-text", true, [
+          "extractLlmContent: structured envelope \u2192 `renderLlmText` (LLM-ready plain text; no binary PDF/DOCX Markdown routes)."
+        ]),
+        slot("structured-chunks", true, [
+          "extractStructuredChunks: structured \u2192 `renderMarkdownSections` / `splitStructuredIntoChunks` (heading-aware chunking + optional parallel `text`)."
         ])
       ];
       break;
@@ -629,6 +677,18 @@ function buildBrowserCapabilityReport(kind) {
           "`extractStructuredData` uses `extractStructuredDataFromImage` (same OCR path as analyzeFile when `ocr.mode` is not off).",
           STRUCTURED_OCR_OFF,
           "HEIC/HEIF and TIFF limitations match `getCapabilities` (`heic`, `tiff`) and OCR warnings."
+        ]),
+        slot("markdown", true, [
+          "extractMarkdown: same bytes + structured fallback through package `extractMarkdown` when applicable; OCR structured layout \u2192 Markdown when OCR runs. HEIC unsupported; TIFF best-effort.",
+          MARKDOWN_IMAGE_OCR_OFF
+        ]),
+        slot("llm-text", true, [
+          "extractLlmContent: OCR structured \u2192 LLM plain text under the same OCR and format limits.",
+          MARKDOWN_IMAGE_OCR_OFF
+        ]),
+        slot("structured-chunks", true, [
+          "extractStructuredChunks: OCR structured \u2192 sectioned Markdown; empty when OCR is off or HEIC.",
+          MARKDOWN_IMAGE_OCR_OFF
         ])
       ];
       break;
@@ -641,6 +701,15 @@ function buildBrowserCapabilityReport(kind) {
         slot("pages", false),
         slot("structured-output", true, [
           "`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)."
+        ]),
+        slot("llm-text", true, [
+          "extractLlmContent: UTF-8 structured rollup \u2192 `renderLlmText` in `@dragon708/docmind-markdown`."
+        ]),
+        slot("structured-chunks", true, [
+          "extractStructuredChunks: typically one Markdown section when only paragraph rollup exists."
         ])
       ];
       break;
@@ -652,7 +721,10 @@ function buildBrowserCapabilityReport(kind) {
         slot("html", false),
         slot("ocr", false),
         slot("pages", false),
-        slot("structured-output", false, [UNKNOWN_KIND])
+        slot("structured-output", false, [UNKNOWN_KIND]),
+        slot("markdown", false, [UNKNOWN_KIND]),
+        slot("llm-text", false, [UNKNOWN_KIND]),
+        slot("structured-chunks", false, [UNKNOWN_KIND])
       ];
   }
   return {
@@ -695,9 +767,10 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
   let limitations = [];
   const ocrOffNote = ocrMode === "off" ? 'Image OCR is skipped when ocr.mode is "off".' : "";
   if (kind === "pdf") {
+    const structuredLikeIntent = intent === "extractStructuredData" || intent === "extractMarkdown" || intent === "extractLlmContent" || intent === "extractStructuredChunks";
     limitations = lim(
       BROWSER_PDF_UNSUPPORTED_WARNING,
-      intent === "extractStructuredData" ? "`extractStructuredData` only returns an empty `StructuredDocumentResult` with warnings for PDF in-browser; use @dragon708/docmind-node for real PDF structured extraction." : ""
+      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." : ""
     );
     nativeExtraction = {
       willAttempt: false,
@@ -721,9 +794,10 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
     });
   }
   if (kind === "unknown") {
+    const structuredLikeIntent = intent === "extractStructuredData" || intent === "extractMarkdown" || intent === "extractLlmContent" || intent === "extractStructuredChunks";
     limitations = lim(
       "Could not classify the file from name, MIME, or bytes; analysis will return not_implemented until hints improve.",
-      intent === "extractStructuredData" ? "`extractStructuredData` needs a known kind (text, DOCX, or image) to produce structured output." : ""
+      structuredLikeIntent ? "Structured and Markdown/LLM/chunk exports need a known kind (text, DOCX, or image) in the browser runtime." : ""
     );
     nativeExtraction = { willAttempt: false, description: "No analyzer selected without a known file kind." };
     ocr = { mayUse: false, description: "OCR is not used for unknown kinds." };
@@ -836,11 +910,26 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
       }
       break;
     case "extractStructuredData":
+    case "extractMarkdown":
+    case "extractLlmContent":
+    case "extractStructuredChunks":
       if (kind === "docx") {
-        nativeExtraction = {
-          willAttempt: true,
-          description: "`extractStructuredDataFromDocx`: Mammoth plus required OOXML extractors (structure, headings, tables, blocks, pagesApprox, embeddedImages unless disabled), then `normalizeToStructuredResult`. Optional `options.docx` is forwarded."
-        };
+        if (intent === "extractStructuredData") {
+          nativeExtraction = {
+            willAttempt: true,
+            description: "`extractStructuredDataFromDocx`: Mammoth plus required OOXML extractors (structure, headings, tables, blocks, pagesApprox, embeddedImages unless disabled), then `normalizeToStructuredResult`. Optional `options.docx` is forwarded."
+          };
+        } else if (intent === "extractMarkdown") {
+          nativeExtraction = {
+            willAttempt: true,
+            description: "extractMarkdown: `extractStructuredData` (Mammoth/OOXML) for a full structured envelope, then `extractMarkdown` in `@dragon708/docmind-markdown`. The package\u2019s DOCX-bytes Mammoth\u2192Turndown shortcut is Node-only; in-browser Markdown uses structured serialization on that envelope (with a package warning)."
+          };
+        } else {
+          nativeExtraction = {
+            willAttempt: true,
+            description: `${String(intent)}: same structured DOCX pipeline as extractStructuredData, then \`@dragon708/docmind-markdown\` (\`renderLlmText\` or \`renderMarkdownSections\`).`
+          };
+        }
         ocr = { mayUse: false, description: "DOCX structured path does not use OCR." };
         limitations = lim(DOCX_ZIP_NOTE_BROWSER);
       } else if (kind === "image") {
@@ -856,10 +945,30 @@ function buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInclude, ocr
       } else {
         nativeExtraction = {
           willAttempt: true,
-          description: "UTF-8 decode via `analyzeText`, then `normalizeToStructuredResult` with a paragraph block rollup."
+          description: intent === "extractStructuredData" ? "UTF-8 decode via `analyzeText`, then `normalizeToStructuredResult` with a paragraph block rollup." : `${String(intent)}: UTF-8 structured envelope, then \`@dragon708/docmind-markdown\`.`
         };
         ocr = { mayUse: false, description: "OCR does not apply to text files." };
       }
+      if (intent === "extractMarkdown") {
+        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."
+          )
+        ];
+      } else if (intent === "extractLlmContent") {
+        limitations = [
+          ...limitations,
+          ...lim("Output: compact plain text via `renderLlmText` (structured input only in this runtime).")
+        ];
+      } else if (intent === "extractStructuredChunks") {
+        limitations = [
+          ...limitations,
+          ...lim(
+            "Output: MarkdownSection[] via `renderMarkdownSections` (`splitStructuredIntoChunks` / `extractStructuredChunks` alias)."
+          )
+        ];
+      }
       break;
     default:
       nativeExtraction = { willAttempt: false, description: "Intent not specialized in this runtime." };
@@ -1012,6 +1121,40 @@ function planForIntent(intentOpt, kind, ocrMode, docxInclude, ocr, analyzeFileOu
         };
     }
   }
+  if (intent === "extractMarkdown") {
+    const sub = planForIntent(
+      "extractStructuredData",
+      kind,
+      ocrMode,
+      docxInclude,
+      ocr,
+      analyzeFileOutput
+    );
+    return {
+      intent,
+      steps: [...sub.steps ?? [], { id: "markdown_hybrid_package", status: "planned" }]
+    };
+  }
+  if (intent === "extractLlmContent" || intent === "extractStructuredChunks") {
+    const sub = planForIntent(
+      "extractStructuredData",
+      kind,
+      ocrMode,
+      docxInclude,
+      ocr,
+      analyzeFileOutput
+    );
+    return {
+      intent,
+      steps: [
+        ...sub.steps ?? [],
+        {
+          id: intent === "extractLlmContent" ? "docmind_markdown_llm_text" : "docmind_markdown_chunk_sections",
+          status: "planned"
+        }
+      ]
+    };
+  }
   if (intent === "analyzeFile") {
     const base = planForAnalyzeFile(kind, ocrMode, docxInclude, ocr);
     if (!analyzeFileRequestsStructured(analyzeFileOutput)) return base;
@@ -1121,6 +1264,6 @@ async function explainAnalysisPlan(input, options) {
   return buildBrowserExplainReport(kind, intent, ocrMode, plan, docxInc, ocrSlice);
 }
 
-export { BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING, BROWSER_PDF_UNSUPPORTED_WARNING, DOCX_EMBEDDED_IMAGE_CAPABILITIES_BROWSER, DOCX_STRUCTURE_CAPABILITIES_BROWSER, analyzeFile, convertToHtml, docxIncludeRequested, explainAnalysisPlan, extractMetadata, extractStructuredData, extractText, getCapabilities, runOcr };
+export { BROWSER_PDF_STRUCTURED_UNSUPPORTED_WARNING, BROWSER_PDF_UNSUPPORTED_WARNING, DOCX_EMBEDDED_IMAGE_CAPABILITIES_BROWSER, DOCX_STRUCTURE_CAPABILITIES_BROWSER, analyzeFile, convertToHtml, docxIncludeRequested, explainAnalysisPlan, extractLlmContent, extractMarkdown, extractMetadata, extractStructuredChunks, extractStructuredData, extractText, getCapabilities, runOcr };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dragon708/docmind-browser",
-  "version": "1.6.0",
+  "version": "1.8.0",
   "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,
@@ -34,6 +34,7 @@
   "license": "MIT",
   "dependencies": {
     "@dragon708/docmind-docx": "^1.8.0",
+    "@dragon708/docmind-markdown": "^1.1.0",
     "@dragon708/docmind-ocr": "^1.1.4",
     "@dragon708/docmind-shared": "^1.2.0"
   },