@dragon708/docmind-node 1.8.0 → 1.9.1

package/dist/index.d.ts

@@ -1,6 +1,6 @@
 import { DocMindAnalyzeOptions, DetectFileKindInput, NamedInput, AnalysisResult, FileKind, RuntimeDescriptor, DocMindPublicIntent, AnalysisAnalyzer, ProcessingPlanDescriptor, ExplainAnalysisPlanOptions, GetCapabilitiesOptions } from '@dragon708/docmind-shared';
 export { AnalysisAnalyzer, AnalysisResult, CapabilityDescriptor, DetectFileKindInput, DocMindPublicIntent, DocxAnalysisCoreResult, ExplainAnalysisPlanOptions, ExplainAnalysisPlanResult, FileKind, FileKindMetadata, GenericAnalysisResult, GetCapabilitiesOptions, GetCapabilitiesResult, ImageAnalysisCoreResult, PdfAnalysisCoreResult, TextAnalysisResult, detectFileKind } from '@dragon708/docmind-shared';
-import { OcrOptions } from '@dragon708/docmind-ocr';
+import { OcrOptions, OcrTiffOptions, PreprocessImageOptions } from '@dragon708/docmind-ocr';
 import { PdfAnalyzeOptions } from '@dragon708/docmind-pdf';
 import { AnalyzeDocxIncludeFlags, DocxToHtmlOptions } from '@dragon708/docmind-docx';
 export { AnalyzeDocxIncludeFlags, DocxEmbeddedImage, DocxEmbeddedImageConversionOptions, DocxEmbeddedImageConversionResult, DocxEmbeddedImageWebSlice, DocxImageExtractionMode, ExtractImagesFromDocxOptions, ExtractImagesFromDocxResult, convertDocxEmbeddedImageToWeb, convertDocxImagesForWeb, docxImageIsBrowserRenderable, docxImageRequiresConversion, docxImageSuggestedTargetFormat, docxImageToDataUri, extractImagesFromDocx, isNodeJsRuntime } from '@dragon708/docmind-docx';
@@ -10,7 +10,7 @@ export { AnalyzeDocxIncludeFlags, DocxEmbeddedImage, DocxEmbeddedImageConversion
  *
  * - **`pdf`**: forwarded to `@dragon708/docmind-pdf`. `analyzeFile` defaults `pdf.ocr` to `"auto"` when omitted.
  *   {@link extractText} / {@link convertToHtml} merge a default of `ocr: "off"` unless you set `pdf.ocr` explicitly.
- * - **`ocr`**: forwarded to `@dragon708/docmind-ocr` for raster images; language string also feeds PDF OCR when `pdf.ocrLangs` is unset.
+ * - **`ocr`**: forwarded to `@dragon708/docmind-ocr` for raster images (`ocrImageDetailed` + `normalizeImageForOcr`, or `ocrTiff` for TIFF); optional `preprocess` maps to {@link PreprocessImageOptions}. `maxPages` / `pageSeparator` apply to multipage TIFF. Language also feeds PDF OCR when `pdf.ocrLangs` is unset.
  * - **`pdfNativeTextSource`**: when `pdf.ocr` resolves to `"off"`, chooses how native text is obtained (see {@link extractText} default).
  */
 /**
@@ -26,7 +26,9 @@ interface NodeAnalyzeDocxOptionsSlice {
 }
 interface NodeAnalyzeOptions extends DocMindAnalyzeOptions {
     readonly pdf?: PdfAnalyzeOptions;
-    readonly ocr?: OcrOptions;
+    readonly ocr?: OcrOptions & Pick<OcrTiffOptions, "maxPages" | "pageSeparator"> & {
+        readonly preprocess?: PreprocessImageOptions;
+    };
     /** Solo DOCX: ver {@link NodeAnalyzeDocxOptionsSlice}. */
     readonly docx?: NodeAnalyzeDocxOptionsSlice;
     /**
@@ -77,12 +79,14 @@ declare function extractMetadata(input: NodeAnalyzeInput, options?: NodeAnalyzeO
 declare function convertToHtml(input: NodeAnalyzeInput, options?: NodeAnalyzeOptions): Promise<AnalysisResult>;
 /**
  * OCR intent: PDF always runs {@link analyzePdf} with `ocr: "force"` (merged with `options.pdf`).
- * Raster images run Tesseract via `options.ocr`. DOCX returns structured extract with a notice.
+ * Raster images: `ocrImageDetailed` (normalize → optional preprocess → Tesseract) for supported single-frame inputs;
+ * TIFF via multipage `ocrTiff` (`options.ocr.maxPages`, `pageSeparator`).
+ * DOCX returns structured extract with a notice.
  */
 declare function runOcr(input: NodeAnalyzeInput, options?: NodeAnalyzeOptions): Promise<AnalysisResult>;
 
 /** High-level features the user can ask DocMind for (per input kind and runtime). */
-type PublicCapabilityId = "text" | "metadata" | "html" | "ocr" | "pages";
+type PublicCapabilityId = "text" | "metadata" | "html" | "ocr" | "pages" | "ocr-multipage" | "image-normalization" | "tiff" | "heic-node-only" | "bmp" | "gif-first-frame";
 /**
  * DOCX-only: what the stack can do with OOXML embedded bitmaps/vector payloads under `word/media`.
  * Present on {@link GetCapabilitiesReport} when `kind === "docx"`.
@@ -174,7 +178,7 @@ type NodeExplainAnalysisPlanOptions = ExplainAnalysisPlanOptions & Pick<NodeAnal
 
 /**
  * Epic 1 — **Capabilities:** after resolving {@link NodeAnalyzeInput}, lists which of
- * `text` | `metadata` | `html` | `ocr` | `pages` apply for that kind in Node (for PDF, `text` / `metadata` /
+ * `text` | `metadata` | `html` | `ocr` | `pages` (and image-specific ids such as `ocr-multipage`, `image-normalization`, `tiff`, `heic-node-only`, `bmp`, `gif-first-frame`) apply for that kind in Node (for PDF, `text` / `metadata` /
  * `pages` / `ocr` describe the v2 pdf-parse + PDF.js + OCR stack; see {@link buildNodeCapabilityReport}).
  * For **DOCX**, `docxEmbeddedImages` and `docxStructure` describe ZIP media and optional OOXML v2 extractors (`options.docx.include`).
  * Does not run Mammoth/Tesseract/PDF bodies beyond path resolution.

package/dist/index.js

@@ -1,8 +1,8 @@
-import { assertValidAnalyzeFileInput, detectFileKind, notImplementedResult, UNKNOWN_FORMAT_WARNING, analyzeText, toUint8Array, isNamedInput, isBinaryInput, isBlob, isFile } from '@dragon708/docmind-shared';
+import { assertValidAnalyzeFileInput, detectFileKind, notImplementedResult, UNKNOWN_FORMAT_WARNING, analyzeText, toUint8Array, isNamedInput, isBinaryInput, isBlob, isFile, getMimeType } from '@dragon708/docmind-shared';
 export { detectFileKind } from '@dragon708/docmind-shared';
 import { analyzeDocx } from '@dragon708/docmind-docx';
 export { convertDocxEmbeddedImageToWeb, convertDocxImagesForWeb, docxImageIsBrowserRenderable, docxImageRequiresConversion, docxImageSuggestedTargetFormat, docxImageToDataUri, extractImagesFromDocx, isNodeJsRuntime } from '@dragon708/docmind-docx';
-import { ocr } from '@dragon708/docmind-ocr';
+import { preprocessHasEffect, resolveImageFormat, ocrTiff, ocrImageDetailed } from '@dragon708/docmind-ocr';
 import { extractPdfMetadata, extractTextFromPdf, analyzePdf, extractPdfTextByPage } from '@dragon708/docmind-pdf';
 import { readFile } from 'fs/promises';
 import { basename } from 'path';
@@ -74,6 +74,53 @@ async function analyzeDocxForNode(input, options) {
   const r = docxOpts !== void 0 ? await analyzeDocx(data, docxOpts) : await analyzeDocx(data);
   return docxPackageResultToAnalysisResult(r);
 }
+function meanPageConfidence(pages) {
+  if (pages.length === 0) return 0;
+  return pages.reduce((s, p) => s + p.confidence, 0) / pages.length;
+}
+function mimeHintFromDetectInput(input) {
+  if (!isByteBackedInput(input)) return void 0;
+  return getMimeType(input);
+}
+async function runRasterOcrForNode(data, input, options) {
+  const signal = options?.ocr?.signal ?? options?.signal;
+  const langs = options?.ocr?.langs;
+  const mimeHint = mimeHintFromDetectInput(input);
+  const format = resolveImageFormat(data, mimeHint);
+  if (format === "tiff") {
+    const tiff = await ocrTiff(data, {
+      langs,
+      signal,
+      maxPages: options?.ocr?.maxPages,
+      pageSeparator: options?.ocr?.pageSeparator
+    });
+    const warnings = [...tiff.warnings];
+    return {
+      text: tiff.text.trim(),
+      confidence: meanPageConfidence(tiff.textByPage),
+      ocrUsed: true,
+      warnings,
+      pages: tiff.pagesProcessed,
+      textByPage: tiff.textByPage
+    };
+  }
+  const detailed = await ocrImageDetailed(data, {
+    langs,
+    signal,
+    preprocess: options?.ocr?.preprocess
+  });
+  return {
+    text: detailed.text.trim(),
+    confidence: detailed.confidence,
+    ocrUsed: true,
+    warnings: [...detailed.warnings],
+    pages: detailed.pages,
+    inputFormat: detailed.inputFormat,
+    normalizedFormat: detailed.normalizedFormat
+  };
+}
+
+// src/analyzers/image.ts
 async function analyzeImageForNode(input, options) {
   if (options?.signal?.aborted) {
     const err = new Error("The operation was aborted");
@@ -93,21 +140,26 @@ async function analyzeImageForNode(input, options) {
       warnings: ["No image bytes were provided for analysis."]
     };
   }
-  const ocrOpts = {
-    ...options?.ocr ?? {},
-    signal: options?.ocr?.signal ?? options?.signal
-  };
-  const r = await ocr(data, ocrOpts);
-  return {
+  const ocrPart = await runRasterOcrForNode(data, input, options);
+  const base = {
     fileKind: "image",
     analyzer: "image",
     status: "ok",
     kind: "image",
-    text: r.text,
-    confidence: r.confidence,
-    ocrUsed: r.ocrUsed,
-    warnings: []
+    text: ocrPart.text,
+    confidence: ocrPart.confidence,
+    ocrUsed: true,
+    warnings: ocrPart.warnings
   };
+  const extra = {};
+  if (ocrPart.pages !== void 0) extra.pages = ocrPart.pages;
+  if (ocrPart.textByPage !== void 0) extra.textByPage = ocrPart.textByPage;
+  if (ocrPart.inputFormat !== void 0) extra.inputFormat = ocrPart.inputFormat;
+  if (ocrPart.normalizedFormat !== void 0) extra.normalizedFormat = ocrPart.normalizedFormat;
+  if (Object.keys(extra).length > 0) {
+    return { ...base, ...extra };
+  }
+  return base;
 }
 async function analyzePdfForNode(input, options) {
   if (options?.signal?.aborted) {
@@ -449,34 +501,10 @@ async function runOcr(input, options) {
       };
     }
     case "image": {
-      const data = await bytesFromDetectInput(resolved);
-      if (data.byteLength === 0) {
-        return {
-          fileKind: "image",
-          analyzer: "image",
-          status: "ok",
-          kind: "image",
-          text: "",
-          confidence: 0,
-          ocrUsed: true,
-          warnings: ["No image bytes were provided for analysis."]
-        };
-      }
-      const ocrOpts = {
-        ...options?.ocr ?? {},
-        signal: options?.ocr?.signal ?? signal
-      };
-      const r = await ocr(data, ocrOpts);
-      return {
-        fileKind: "image",
-        analyzer: "image",
-        status: "ok",
-        kind: "image",
-        text: r.text,
-        confidence: r.confidence,
-        ocrUsed: r.ocrUsed,
-        warnings: []
-      };
+      return analyzeImageForNode(resolved, {
+        ...options,
+        ocr: { ...options?.ocr ?? {}, signal: options?.ocr?.signal ?? signal }
+      });
     }
     case "docx": {
       const data = await bytesFromDetectInput(resolved);
@@ -589,11 +617,33 @@ function buildNodeCapabilityReport(kind) {
       break;
     case "image":
       capabilities = [
-        slot("text", true, ["Text is obtained via OCR."]),
+        slot("text", true, [
+          "Text via `@dragon708/docmind-ocr` after format sniff (PNG, JPEG, WebP, TIFF, BMP, GIF first frame, HEIC/HEIF in Node via conversion)."
+        ]),
         slot("metadata", false, [IMAGE_META]),
         slot("html", false, [IMAGE_HTML]),
-        slot("ocr", true),
-        slot("pages", false)
+        slot("ocr", true, [
+          "Single-frame pipeline: `normalizeImageForOcr` \u2192 optional `preprocessImageForOcr` (`options.ocr.preprocess`) \u2192 Tesseract via `ocrImageDetailed`."
+        ]),
+        slot("ocr-multipage", true, [
+          "Multipage TIFF: `ocrTiff` with per-page text; `text` joins pages (see `options.ocr.pageSeparator`)."
+        ]),
+        slot("image-normalization", true, [
+          "Bytes are normalized to a Tesseract-friendly raster (PNG-oriented) inside the OCR package before recognition."
+        ]),
+        slot("tiff", true, [
+          "TIFF detected by magic/MIME routes to `ocrTiff` (not the single-frame `ocrImageDetailed` path)."
+        ]),
+        slot("heic-node-only", true, [
+          "HEIC/HEIF uses Node `sharp` conversion in `@dragon708/docmind-ocr`; not available in the browser facade."
+        ]),
+        slot("bmp", true, ["BMP inputs are supported through the universal normalization path."]),
+        slot("gif-first-frame", true, [
+          "Animated GIF: only the first frame is normalized and OCR'd."
+        ]),
+        slot("pages", true, [
+          "TIFF: `pages` and `textByPage` mirror frames processed. Other formats may expose `pages` when the normalizer reports it (e.g. GIF)."
+        ])
       ];
       break;
     case "text":
@@ -643,11 +693,12 @@ function finalizeDocxExplainReport(report) {
   };
 }
 var DOCX_MAMMOTH_PLUS_OPTIONAL = "Mammoth extracts text and HTML from OOXML; optional parallel OOXML/ZIP extractors run when options.docx.include is set (structure, headings, tables, blocks, pagesApprox, embeddedImages).";
-function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
+var NODE_IMAGE_OCR_PIPELINE = "Node raster OCR: detect format (magic/MIME) \u2192 `normalizeImageForOcr` \u2192 optional `preprocessImageForOcr` when `options.ocr.preprocess` is set \u2192 Tesseract. TIFF is handled with `ocrTiff` (multipage; per-page `textByPage` and joined `text`). HEIC/HEIF is decoded via `sharp` on Node (not in the browser package). BMP and static images share the normalization path; GIF uses the first frame only.";
+function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude, ocrSlice) {
   const runtime = { id: "node" };
   const primaryAnalyzer = kind === "pdf" ? "pdf" : kind === "docx" ? "docx" : kind === "image" ? "image" : kind === "text" ? "text" : "none";
   let nativeExtraction;
-  let ocr3;
+  let ocr;
   let limitations = [];
   if (kind === "unknown") {
     limitations = lim(
@@ -672,7 +723,7 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "pdf-parse supplies embedded text, metadata, and page count; PDF.js drives raster OCR when enabled."
         };
-        ocr3 = {
+        ocr = {
           mayUse: pdfOcr !== "off",
           description: pdfOcr === "off" ? "Raster OCR pipeline is off (pdf.ocr: off)." : pdfOcr === "force" ? "Raster OCR may run on all pages when pdf.ocr is force." : "Raster OCR may run when native text looks insufficient (pdf.ocr: auto + heuristics)."
         };
@@ -681,19 +732,22 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: docxIncludeRequested(docxInclude) ? "Mammoth plus parallel OOXML extractors (per options.docx.include)." : DOCX_MAMMOTH_PLUS_OPTIONAL
         };
-        ocr3 = { mayUse: false, description: "DOCX does not use OCR in DocMind." };
+        ocr = { mayUse: false, description: "DOCX does not use OCR in DocMind." };
       } else if (kind === "image") {
         nativeExtraction = {
           willAttempt: false,
-          description: "Images have no native text layer; text comes from OCR only."
+          description: NODE_IMAGE_OCR_PIPELINE
+        };
+        ocr = {
+          mayUse: true,
+          description: "Same stack as `runOcr` / `analyzeFile` for images: universal normalization, optional preprocess, then `ocrImageDetailed` or multipage `ocrTiff` for TIFF."
         };
-        ocr3 = { mayUse: true, description: "Tesseract runs on supported raster formats." };
       } else {
         nativeExtraction = {
           willAttempt: true,
           description: "UTF-8 decode with BOM handling for plain text."
         };
-        ocr3 = { mayUse: false, description: "OCR does not apply to text files." };
+        ocr = { mayUse: false, description: "OCR does not apply to text files." };
       }
       break;
     case "extractText":
@@ -702,7 +756,7 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "Node: pdf-parse for metadata/page baseline, then PDF.js per-page text merged into `text` (pdfNativeTextSource pdfjs-per-page default)."
         };
-        ocr3 = {
+        ocr = {
           mayUse: false,
           description: "extractText defaults pdf.ocr off; set pdf.ocr explicitly to allow auto/force raster OCR."
         };
@@ -711,16 +765,22 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: docxIncludeRequested(docxInclude) ? "Same DOCX router as analyzeFile: Mammoth text + optional OOXML fields; HTML cleared in extractText." : "Mammoth plain text; HTML cleared. Optional OOXML v2 fields when options.docx.include is set."
         };
-        ocr3 = { mayUse: false, description: "DOCX does not use OCR." };
+        ocr = { mayUse: false, description: "DOCX does not use OCR." };
       } else if (kind === "image") {
-        nativeExtraction = { willAttempt: false, description: "No embedded text layer." };
-        ocr3 = { mayUse: true, description: "OCR produces text for images." };
+        nativeExtraction = {
+          willAttempt: false,
+          description: NODE_IMAGE_OCR_PIPELINE
+        };
+        ocr = {
+          mayUse: true,
+          description: "Same Node image pipeline as analyzeFile (normalize \u2192 optional preprocess \u2192 `ocrImageDetailed` or `ocrTiff`)."
+        };
       } else {
         nativeExtraction = {
           willAttempt: true,
           description: "UTF-8 decode only."
         };
-        ocr3 = { mayUse: false, description: "OCR does not apply." };
+        ocr = { mayUse: false, description: "OCR does not apply." };
       }
       break;
     case "extractMetadata":
@@ -729,13 +789,13 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "Lightweight PDF info/XMP normalization without full OCR."
         };
-        ocr3 = { mayUse: false, description: "extractMetadata does not run the OCR pipeline." };
+        ocr = { mayUse: false, description: "extractMetadata does not run the OCR pipeline." };
       } else if (kind === "docx" || kind === "image") {
         nativeExtraction = {
           willAttempt: false,
           description: "Stub response; no heavy extractor."
         };
-        ocr3 = { mayUse: false, description: "OCR not used for this metadata path." };
+        ocr = { mayUse: false, description: "OCR not used for this metadata path." };
         limitations = lim(
           kind === "docx" ? "Structured DOCX metadata is not exposed separately; use analyzeFile / extractText / convertToHtml with options.docx.include for OOXML structure." : "Raster images have no document metadata bundle."
         );
@@ -744,7 +804,7 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "Decoded text only; no structured document metadata."
         };
-        ocr3 = { mayUse: false, description: "OCR does not apply." };
+        ocr = { mayUse: false, description: "OCR does not apply." };
         limitations = lim("Plain text has no structured document metadata.");
       }
       break;
@@ -754,26 +814,26 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "Text layer extracted then wrapped in <pre> (not visual layout)."
         };
-        ocr3 = { mayUse: false, description: "convertToHtml does not run PDF OCR." };
+        ocr = { mayUse: false, description: "convertToHtml does not run PDF OCR." };
         limitations = lim("PDF HTML is a plain-text preview, not page layout.");
       } else if (kind === "docx") {
         nativeExtraction = {
           willAttempt: true,
           description: docxIncludeRequested(docxInclude) ? "Mammoth HTML plus optional OOXML extractors (same router as analyzeFile)." : "Mammoth HTML via analyzeFile routing; optional OOXML v2 when options.docx.include is set."
         };
-        ocr3 = { mayUse: false, description: "DOCX path does not use OCR." };
+        ocr = { mayUse: false, description: "DOCX path does not use OCR." };
       } else if (kind === "text") {
         nativeExtraction = {
           willAttempt: true,
           description: "UTF-8 decode then <pre> wrapper."
         };
-        ocr3 = { mayUse: false, description: "OCR does not apply." };
+        ocr = { mayUse: false, description: "OCR does not apply." };
       } else {
         nativeExtraction = {
           willAttempt: false,
           description: "No HTML path for raster images."
         };
-        ocr3 = { mayUse: false, description: "OCR does not emit layout HTML here." };
+        ocr = { mayUse: false, description: "OCR does not emit layout HTML here." };
         limitations = lim("Use extractText or runOcr for image text.");
       }
       break;
@@ -783,32 +843,44 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
           willAttempt: true,
           description: "pdf-parse runs first; text may be replaced by raster OCR output."
         };
-        ocr3 = {
+        ocr = {
           mayUse: true,
           description: 'runOcr always sets pdf.ocr to "force" for PDFs.'
         };
         limitations = lim("Forced OCR may run even when a text layer exists.");
       } else if (kind === "image") {
-        nativeExtraction = { willAttempt: false, description: "No native text layer." };
-        ocr3 = { mayUse: true, description: "Tesseract OCR on the image bytes." };
+        nativeExtraction = {
+          willAttempt: false,
+          description: NODE_IMAGE_OCR_PIPELINE
+        };
+        ocr = {
+          mayUse: true,
+          description: "Forced OCR path for rasters: TIFF \u2192 `ocrTiff` with `textByPage` when multipage; other formats \u2192 `ocrImageDetailed` after normalization (HEIC converted with `sharp` on Node)."
+        };
       } else if (kind === "docx") {
         nativeExtraction = {
           willAttempt: true,
           description: docxIncludeRequested(docxInclude) ? "Mammoth text/HTML plus optional OOXML extractors; still not OCR." : "Full Mammoth extract (text + HTML); optional OOXML v2 via options.docx.include; not OCR."
         };
-        ocr3 = { mayUse: false, description: "DOCX is not OCR'd." };
+        ocr = { mayUse: false, description: "DOCX is not OCR'd." };
         limitations = lim("Result is structured extract, not OCR output.");
       } else {
         nativeExtraction = {
           willAttempt: true,
           description: "UTF-8 decode only."
         };
-        ocr3 = { mayUse: false, description: "OCR does not apply to text files." };
+        ocr = { mayUse: false, description: "OCR does not apply to text files." };
       }
       break;
     default:
       nativeExtraction = { willAttempt: false, description: "Generic intent; see plan." };
-      ocr3 = { mayUse: false, description: "See plan steps." };
+      ocr = { mayUse: false, description: "See plan steps." };
+  }
+  if (kind === "image" && preprocessHasEffect(ocrSlice?.preprocess)) {
+    limitations = [
+      ...limitations,
+      "options.ocr.preprocess applies to the single-frame `ocrImageDetailed` path; multipage TIFF (`ocrTiff`) does not run this preprocess on each frame."
+    ];
   }
   return finalizeDocxExplainReport({
     kind,
@@ -817,17 +889,27 @@ function buildNodeExplainReport(kind, intent, pdfOcr, plan, docxInclude) {
     intent,
     primaryAnalyzer,
     nativeExtraction,
-    ocr: ocr3,
+    ocr,
     limitations,
     plan
   });
 }
-
-// src/introspection.ts
 function resolvePdfOcrMode(pdf) {
   return pdf?.ocr ?? "auto";
 }
-function planAnalyzeFile(kind, pdfOcr, docxInclude) {
+function imageOcrPlanSteps(ocr) {
+  return [
+    { id: "detect_kind", status: "done" },
+    { id: "image_format_detect", status: "planned" },
+    { id: "normalize_image_for_ocr", status: "planned" },
+    {
+      id: "preprocess_image_for_ocr",
+      status: preprocessHasEffect(ocr?.preprocess) ? "planned" : "skipped"
+    },
+    { id: "ocr_tesseract", status: "planned" }
+  ];
+}
+function planAnalyzeFile(kind, pdfOcr, docxInclude, ocr) {
   switch (kind) {
     case "pdf":
       return {
@@ -855,10 +937,7 @@ function planAnalyzeFile(kind, pdfOcr, docxInclude) {
     case "image":
       return {
         intent: "analyzeFile",
-        steps: [
-          { id: "detect_kind", status: "done" },
-          { id: "image_ocr", status: "planned" }
-        ]
+        steps: imageOcrPlanSteps(ocr)
       };
     case "text":
       return {
@@ -878,9 +957,9 @@ function planAnalyzeFile(kind, pdfOcr, docxInclude) {
       };
   }
 }
-function planForIntent(intentOpt, kind, pdfOcrForAnalyze, docxInclude) {
+function planForIntent(intentOpt, kind, pdfOcrForAnalyze, docxInclude, ocr) {
   const intent = intentOpt ?? "analyzeFile";
-  if (intent === "analyzeFile") return planAnalyzeFile(kind, pdfOcrForAnalyze, docxInclude);
+  if (intent === "analyzeFile") return planAnalyzeFile(kind, pdfOcrForAnalyze, docxInclude, ocr);
   if (intent === "extractText") {
     if (kind === "pdf") {
       return {
@@ -893,7 +972,7 @@ function planForIntent(intentOpt, kind, pdfOcrForAnalyze, docxInclude) {
         ]
       };
     }
-    const p = planAnalyzeFile(kind, "off", docxInclude);
+    const p = planAnalyzeFile(kind, "off", docxInclude, ocr);
     return { ...p, intent: "extractText" };
   }
   if (intent === "extractMetadata") {
@@ -977,10 +1056,7 @@ function planForIntent(intentOpt, kind, pdfOcrForAnalyze, docxInclude) {
     if (kind === "image") {
       return {
         intent: "runOcr",
-        steps: [
-          { id: "detect_kind", status: "done" },
-          { id: "tesseract_ocr", status: "planned" }
-        ]
+        steps: imageOcrPlanSteps(ocr)
       };
     }
     if (kind === "docx") {
@@ -1002,7 +1078,7 @@ function planForIntent(intentOpt, kind, pdfOcrForAnalyze, docxInclude) {
       ]
     };
   }
-  return planAnalyzeFile(kind, pdfOcrForAnalyze, docxInclude);
+  return planAnalyzeFile(kind, pdfOcrForAnalyze, docxInclude, ocr);
 }
 async function getCapabilities(input, options) {
   throwIfAborted(options?.signal);
@@ -1019,8 +1095,9 @@ async function explainAnalysisPlan(input, options) {
   const intent = options?.intent ?? "analyzeFile";
   const pdfOcrAnalyze = resolvePdfOcrMode(options?.pdf);
   const docxInc = options?.docx?.include;
-  const plan = planForIntent(intent, kind, pdfOcrAnalyze, docxInc);
-  return buildNodeExplainReport(kind, intent, pdfOcrAnalyze, plan, docxInc);
+  const ocrSlice = options?.ocr;
+  const plan = planForIntent(intent, kind, pdfOcrAnalyze, docxInc, ocrSlice);
+  return buildNodeExplainReport(kind, intent, pdfOcrAnalyze, plan, docxInc, ocrSlice);
 }
 
 export { DOCX_EMBEDDED_IMAGE_CAPABILITIES, DOCX_STRUCTURE_CAPABILITIES, analyzeFile, bufferToInput, convertToHtml, docxIncludeRequested, explainAnalysisPlan, extractMetadata, extractText, getCapabilities, readFileToInput, resolveNodeAnalyzeInput, runOcr };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@dragon708/docmind-node",
-  "version": "1.8.0",
+  "version": "1.9.1",
   "description": "Official DocMind Node facade: analyzeFile, intent APIs, PDF/DOCX/OCR, and fs helpers.",
   "type": "module",
   "main": "./dist/index.js",
@@ -32,10 +32,10 @@
   ],
   "license": "MIT",
   "dependencies": {
-    "@dragon708/docmind-docx": "^1.7.0",
-    "@dragon708/docmind-ocr": "^1.0.0",
-    "@dragon708/docmind-pdf": "^2.0.1",
-    "@dragon708/docmind-shared": "^1.1.0"
+    "@dragon708/docmind-docx": "^1.7.1",
+    "@dragon708/docmind-ocr": "^1.1.3",
+    "@dragon708/docmind-pdf": "^2.1.1",
+    "@dragon708/docmind-shared": "^1.1.1"
   },
   "devDependencies": {
     "@types/node": "^20.19.37",