@hawon/nexus 0.4.0 → 0.4.1

package/dist/docparser/index.d.ts

@@ -3,4 +3,5 @@ export { chunkText } from "./chunker.js";
 export { extractPdfText, isPdfSupported } from "./pdf.js";
 export { extractDocxText } from "./docx.js";
 export { extractPlainText } from "./text.js";
+export { convertWithMarkItDown, isMarkItDownSupported, isMarkItDownFormat } from "./markitdown.js";
 export type { ParsedDocument, DocumentChunk, ParseOptions, DocumentFormat } from "./types.js";

package/dist/docparser/index.js

@@ -3,3 +3,4 @@ export { chunkText } from "./chunker.js";
 export { extractPdfText, isPdfSupported } from "./pdf.js";
 export { extractDocxText } from "./docx.js";
 export { extractPlainText } from "./text.js";
+export { convertWithMarkItDown, isMarkItDownSupported, isMarkItDownFormat } from "./markitdown.js";

package/dist/docparser/markitdown.d.ts

@@ -0,0 +1,7 @@
+export declare function isMarkItDownSupported(): boolean;
+/**
+ * Convert any file to markdown using Microsoft MarkItDown.
+ * Supports: PDF, DOCX, PPTX, XLSX, HTML, images, audio, ZIP, etc.
+ */
+export declare function convertWithMarkItDown(filePath: string): string;
+export declare function isMarkItDownFormat(ext: string): boolean;

package/dist/docparser/markitdown.js

@@ -0,0 +1,43 @@
+import { execSync } from "node:child_process";
+let _supported = null;
+export function isMarkItDownSupported() {
+    if (_supported !== null)
+        return _supported;
+    try {
+        execSync("markitdown --help", { stdio: "ignore" });
+        _supported = true;
+    }
+    catch {
+        _supported = false;
+    }
+    return _supported;
+}
+/**
+ * Convert any file to markdown using Microsoft MarkItDown.
+ * Supports: PDF, DOCX, PPTX, XLSX, HTML, images, audio, ZIP, etc.
+ */
+export function convertWithMarkItDown(filePath) {
+    if (!isMarkItDownSupported()) {
+        throw new Error("markitdown not installed. Run: pip install markitdown");
+    }
+    const result = execSync(`markitdown "${filePath}"`, {
+        maxBuffer: 50 * 1024 * 1024,
+        encoding: "utf-8",
+        timeout: 60_000,
+    });
+    return result.trim();
+}
+/** Formats that MarkItDown supports beyond our built-in parsers. */
+const MARKITDOWN_EXTENSIONS = new Set([
+    ".pptx", ".ppt", ".xlsx", ".xls",
+    ".html", ".htm", ".xml",
+    ".jpg", ".jpeg", ".png", ".gif", ".bmp", ".webp",
+    ".mp3", ".wav", ".m4a",
+    ".zip", ".tar", ".gz",
+    ".json", ".yaml", ".yml",
+    ".rst", ".org", ".rtf",
+    ".epub",
+]);
+export function isMarkItDownFormat(ext) {
+    return MARKITDOWN_EXTENSIONS.has(ext.toLowerCase());
+}

package/dist/docparser/parse-document.js

@@ -4,6 +4,7 @@ import { chunkText } from "./chunker.js";
 import { extractPdfText, isPdfSupported } from "./pdf.js";
 import { extractDocxText } from "./docx.js";
 import { extractPlainText } from "./text.js";
+import { convertWithMarkItDown, isMarkItDownSupported, isMarkItDownFormat } from "./markitdown.js";
 export function detectFormat(filePath) {
     const ext = extname(filePath).toLowerCase();
     switch (ext) {
@@ -16,33 +17,78 @@ export function detectFormat(filePath) {
         case ".text":
         case ".log":
         case ".csv": return "txt";
-        default: return null;
+        default:
+            // MarkItDown handles PPTX, XLSX, HTML, images, audio, etc.
+            if (isMarkItDownFormat(ext))
+                return "txt";
+            return null;
     }
 }
+/** Check if a file should use MarkItDown instead of built-in parsers. */
+function shouldUseMarkItDown(filePath) {
+    const ext = extname(filePath).toLowerCase();
+    return isMarkItDownFormat(ext);
+}
 export function parseDocument(filePath, memory, options) {
     if (!existsSync(filePath))
         throw new Error(`File not found: ${filePath}`);
+    const ext = extname(filePath).toLowerCase();
+    const useMarkItDown = shouldUseMarkItDown(filePath);
     const format = options?.format ?? detectFormat(filePath);
-    if (!format)
-        throw new Error(`Unsupported format: ${extname(filePath)}`);
+    if (!format && !useMarkItDown)
+        throw new Error(`Unsupported format: ${ext}`);
     let text;
     let pageCount;
-    switch (format) {
-        case "pdf": {
-            if (!isPdfSupported())
-                throw new Error("PDF requires python3 + pymupdf");
-            const result = extractPdfText(filePath);
-            text = result.text;
-            pageCount = result.pageCount;
-            break;
+    if (useMarkItDown) {
+        // PPTX, XLSX, HTML, images, audio, etc. → MarkItDown
+        if (!isMarkItDownSupported())
+            throw new Error("Install markitdown for this format: pip install markitdown");
+        text = convertWithMarkItDown(filePath);
+    }
+    else {
+        switch (format) {
+            case "pdf": {
+                // Try pymupdf first, fall back to markitdown
+                if (isPdfSupported()) {
+                    const result = extractPdfText(filePath);
+                    text = result.text;
+                    pageCount = result.pageCount;
+                }
+                else if (isMarkItDownSupported()) {
+                    text = convertWithMarkItDown(filePath);
+                }
+                else {
+                    throw new Error("PDF requires python3 + pymupdf or markitdown");
+                }
+                break;
+            }
+            case "docx":
+                try {
+                    text = extractDocxText(filePath);
+                }
+                catch {
+                    // Fall back to markitdown if unzip fails
+                    if (isMarkItDownSupported()) {
+                        text = convertWithMarkItDown(filePath);
+                    }
+                    else {
+                        throw new Error(`Failed to parse DOCX: ${filePath}`);
+                    }
+                }
+                break;
+            case "markdown":
+            case "txt":
+                text = extractPlainText(filePath, format === "markdown");
+                break;
+            default:
+                if (isMarkItDownSupported()) {
+                    text = convertWithMarkItDown(filePath);
+                }
+                else {
+                    throw new Error(`Unsupported format: ${ext}`);
+                }
+                break;
         }
-        case "docx":
-            text = extractDocxText(filePath);
-            break;
-        case "markdown":
-        case "txt":
-            text = extractPlainText(filePath, format === "markdown");
-            break;
     }
     // Truncate if needed
     const maxChars = options?.maxChars ?? 500_000;
@@ -59,7 +105,7 @@ export function parseDocument(filePath, memory, options) {
         memory.save();
     return {
         filePath,
-        format,
+        format: format ?? "txt",
         title,
         text,
         chunks,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hawon/nexus",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "description": "The all-in-one AI developer framework — session intelligence, code review, prompt injection defense, infinite memory, self-evolving skills",
   "type": "module",
   "main": "dist/index.js",