kordoc 1.7.2 → 1.9.0

package/dist/{chunk-NJ3R7LNR.js → chunk-MDRW3HYC.js}

@@ -1,10 +1,22 @@
 #!/usr/bin/env node
+import {
+  KordocError,
+  classifyError,
+  isPathTraversal,
+  precheckZipSize,
+  sanitizeHref,
+  toArrayBuffer
+} from "./chunk-AHW56LNX.js";
+import {
+  parsePageRange
+} from "./chunk-MOL7MDBG.js";
 
 // src/detect.ts
+import JSZip from "jszip";
 function magicBytes(buffer) {
   return new Uint8Array(buffer, 0, Math.min(4, buffer.byteLength));
 }
-function isHwpxFile(buffer) {
+function isZipFile(buffer) {
   const b = magicBytes(buffer);
   return b[0] === 80 && b[1] === 75 && b[2] === 3 && b[3] === 4;
 }
@@ -18,15 +30,28 @@ function isPdfFile(buffer) {
 }
 function detectFormat(buffer) {
   if (buffer.byteLength < 4) return "unknown";
-  if (isHwpxFile(buffer)) return "hwpx";
+  if (isZipFile(buffer)) return "hwpx";
   if (isOldHwpFile(buffer)) return "hwp";
   if (isPdfFile(buffer)) return "pdf";
   return "unknown";
 }
+async function detectZipFormat(buffer) {
+  try {
+    const zip = await JSZip.loadAsync(buffer);
+    if (zip.file("xl/workbook.xml")) return "xlsx";
+    if (zip.file("word/document.xml")) return "docx";
+    if (zip.file("Contents/content.hpf") || zip.file("mimetype")) return "hwpx";
+    const hasSection = Object.keys(zip.files).some((f) => f.startsWith("Contents/"));
+    if (hasSection) return "hwpx";
+    return "unknown";
+  } catch {
+    return "unknown";
+  }
+}
 
 // src/table/builder.ts
 var SAFE_HREF_RE = /^(?:https?:|mailto:|tel:|#)/i;
-function sanitizeHref(href) {
+function sanitizeHref2(href) {
   const trimmed = href.trim();
   if (!trimmed || !SAFE_HREF_RE.test(trimmed)) return null;
   return trimmed;
@@ -36,23 +61,24 @@ var MAX_ROWS = 1e4;
 function buildTable(rows) {
   if (rows.length > MAX_ROWS) rows = rows.slice(0, MAX_ROWS);
   const numRows = rows.length;
-  const tempOccupied = /* @__PURE__ */ new Set();
+  const hasAddr = rows.some((row) => row.some((c) => c.colAddr !== void 0 && c.rowAddr !== void 0));
+  if (hasAddr) return buildTableDirect(rows, numRows);
   let maxCols = 0;
+  const tempOccupied = Array.from({ length: numRows }, () => []);
   for (let rowIdx = 0; rowIdx < numRows; rowIdx++) {
     let colIdx = 0;
     for (const cell of rows[rowIdx]) {
-      while (colIdx < MAX_COLS && tempOccupied.has(rowIdx * MAX_COLS + colIdx)) colIdx++;
+      while (colIdx < MAX_COLS && tempOccupied[rowIdx][colIdx]) colIdx++;
       if (colIdx >= MAX_COLS) break;
       for (let r = rowIdx; r < Math.min(rowIdx + cell.rowSpan, numRows); r++) {
         for (let c = colIdx; c < Math.min(colIdx + cell.colSpan, MAX_COLS); c++) {
-          tempOccupied.add(r * MAX_COLS + c);
+          tempOccupied[r][c] = true;
         }
       }
       colIdx += cell.colSpan;
       if (colIdx > maxCols) maxCols = colIdx;
     }
   }
-  tempOccupied.clear();
   if (maxCols === 0) return { rows: 0, cols: 0, cells: [], hasHeader: false };
   const grid = Array.from(
     { length: numRows },
@@ -80,6 +106,50 @@ function buildTable(rows) {
       cellIdx++;
     }
   }
+  return trimAndReturn(grid, numRows, maxCols);
+}
+function buildTableDirect(rows, numRows) {
+  let maxCols = 0;
+  for (const row of rows) {
+    for (const cell of row) {
+      const end = (cell.colAddr ?? 0) + cell.colSpan;
+      if (end > maxCols) maxCols = end;
+    }
+  }
+  if (maxCols === 0) return { rows: 0, cols: 0, cells: [], hasHeader: false };
+  const grid = Array.from(
+    { length: numRows },
+    () => Array.from({ length: maxCols }, () => ({ text: "", colSpan: 1, rowSpan: 1 }))
+  );
+  for (const row of rows) {
+    for (const cell of row) {
+      const r = cell.rowAddr ?? 0;
+      const c = cell.colAddr ?? 0;
+      if (r >= numRows || c >= maxCols) continue;
+      grid[r][c] = { text: cell.text.trim(), colSpan: cell.colSpan, rowSpan: cell.rowSpan };
+      for (let dr = 0; dr < cell.rowSpan; dr++) {
+        for (let dc = 0; dc < cell.colSpan; dc++) {
+          if (dr === 0 && dc === 0) continue;
+          if (r + dr < numRows && c + dc < maxCols) {
+            grid[r + dr][c + dc] = { text: "", colSpan: 1, rowSpan: 1 };
+          }
+        }
+      }
+    }
+  }
+  return trimAndReturn(grid, numRows, maxCols);
+}
+function trimAndReturn(grid, numRows, maxCols) {
+  let effectiveCols = maxCols;
+  while (effectiveCols > 0) {
+    const colEmpty = grid.every((row) => !row[effectiveCols - 1]?.text?.trim());
+    if (!colEmpty) break;
+    effectiveCols--;
+  }
+  if (effectiveCols < maxCols && effectiveCols > 0) {
+    const trimmed = grid.map((row) => row.slice(0, effectiveCols));
+    return { rows: numRows, cols: effectiveCols, cells: trimmed, hasHeader: numRows > 1 };
+  }
   return { rows: numRows, cols: maxCols, cells: grid, hasHeader: numRows > 1 };
 }
 function convertTableToText(rows) {
@@ -87,13 +157,26 @@ function convertTableToText(rows) {
     (row) => row.map((c) => c.text.trim().replace(/\n/g, " ")).filter(Boolean).join(" | ")
   ).filter(Boolean).join("\n");
 }
+var HWP_SHAPE_ALT_TEXT_RE = /(?:모서리가 둥근 |둥근 )?(?:사각형|직사각형|정사각형|원|타원|삼각형|이등변 삼각형|직각 삼각형|선|직선|곡선|화살표|굵은 화살표|이중 화살표|오각형|육각형|팔각형|별|[4-8]점별|십자|십자형|구름|구름형|마름모|도넛|평행사변형|사다리꼴|부채꼴|호|반원|물결|번개|하트|빗금|블록 화살표|수식|표|그림|개체|그리기\s?개체|묶음\s?개체|글상자|수식\s?개체|OLE\s?개체)\s?입니다\.?/g;
+function sanitizeText(text) {
+  let result = text.replace(/[\u{F0000}-\u{FFFFD}]/gu, "").replace(HWP_SHAPE_ALT_TEXT_RE, "").replace(/  +/g, " ").trim();
+  if (result.length <= 30 && result.includes(" ")) {
+    const tokens = result.split(" ");
+    const koreanSingleCharCount = tokens.filter((t) => t.length === 1 && /[\uAC00-\uD7AF\u3131-\u318E]/.test(t)).length;
+    if (tokens.length >= 3 && koreanSingleCharCount / tokens.length >= 0.7) {
+      result = tokens.join("");
+    }
+  }
+  return result;
+}
 function blocksToMarkdown(blocks) {
   const lines = [];
   for (let i = 0; i < blocks.length; i++) {
     const block = blocks[i];
     if (block.type === "heading" && block.text) {
       const prefix = "#".repeat(Math.min(block.level || 2, 6));
-      lines.push("", `${prefix} ${block.text}`, "");
+      const headingText = sanitizeText(block.text);
+      if (headingText) lines.push("", `${prefix} ${headingText}`, "");
       continue;
     }
     if (block.type === "image" && block.text) {
@@ -105,9 +188,11 @@ function blocksToMarkdown(blocks) {
       continue;
     }
     if (block.type === "list" && block.text) {
-      const alreadyNumbered = block.listType === "ordered" && /^\d+\.\s/.test(block.text);
+      const listText = sanitizeText(block.text);
+      if (!listText) continue;
+      const alreadyNumbered = block.listType === "ordered" && /^\d+\.\s/.test(listText);
       const prefix = alreadyNumbered ? "" : block.listType === "ordered" ? "1. " : "- ";
-      lines.push(`${prefix}${block.text}`);
+      lines.push(`${prefix}${listText}`);
       if (block.children) {
         for (const child of block.children) {
           const childPrefix = child.listType === "ordered" ? "1." : "-";
@@ -117,7 +202,8 @@ function blocksToMarkdown(blocks) {
       continue;
     }
     if (block.type === "paragraph" && block.text) {
-      let text = block.text;
+      let text = sanitizeText(block.text);
+      if (!text) continue;
       if (/^\[별표\s*\d+/.test(text)) {
         const nextBlock = blocks[i + 1];
         if (nextBlock?.type === "paragraph" && nextBlock.text && /관련\)?$/.test(nextBlock.text)) {
@@ -133,7 +219,7 @@ function blocksToMarkdown(blocks) {
         continue;
       }
       if (block.href) {
-        const href = sanitizeHref(block.href);
+        const href = sanitizeHref2(block.href);
         if (href) text = `[${text}](${href})`;
       }
       if (block.footnoteText) {
@@ -154,7 +240,7 @@ function tableToMarkdown(table) {
   if (table.rows === 0 || table.cols === 0) return "";
   const { cells, rows: numRows, cols: numCols } = table;
   if (numRows === 1 && numCols === 1) {
-    const content = cells[0][0].text;
+    const content = sanitizeText(cells[0][0].text);
     return content.split(/\n/).map((line) => {
       const trimmed = line.trim();
       if (!trimmed) return "";
@@ -163,13 +249,17 @@ function tableToMarkdown(table) {
       return trimmed;
     }).filter(Boolean).join("\n");
   }
+  if (numCols === 1 && numRows >= 2) {
+    return cells.map((row) => sanitizeText(row[0].text).replace(/\n/g, " ")).filter(Boolean).join("\n");
+  }
   const display = Array.from({ length: numRows }, () => Array(numCols).fill(""));
   const skip = /* @__PURE__ */ new Set();
   for (let r = 0; r < numRows; r++) {
     for (let c = 0; c < numCols; c++) {
       if (skip.has(`${r},${c}`)) continue;
-      const cell = cells[r][c];
-      display[r][c] = cell.text.replace(/\n/g, "<br>");
+      const cell = cells[r]?.[c];
+      if (!cell) continue;
+      display[r][c] = sanitizeText(cell.text).replace(/\n/g, "<br>");
       for (let dr = 0; dr < cell.rowSpan; dr++) {
         for (let dc = 0; dc < cell.colSpan; dc++) {
           if (dr === 0 && dc === 0) continue;
@@ -178,12 +268,28 @@ function tableToMarkdown(table) {
           }
         }
       }
+      c += cell.colSpan - 1;
     }
   }
   const uniqueRows = [];
-  for (const row of display) {
+  let pendingFirstCol = "";
+  for (let r = 0; r < display.length; r++) {
+    const row = display[r];
     const isEmptyPlaceholder = row.every((cell) => cell === "");
-    if (!isEmptyPlaceholder) uniqueRows.push(row);
+    if (isEmptyPlaceholder) continue;
+    const hasSkippedCols = row.some((cell, c) => cell === "" && skip.has(`${r},${c}`));
+    const nonEmptyCols = row.filter((cell) => cell !== "");
+    if (!hasSkippedCols && nonEmptyCols.length === 1 && row[0] !== "" && row.slice(1).every((c) => c === "")) {
+      pendingFirstCol = row[0];
+      continue;
+    }
+    if (pendingFirstCol && row[0] === "") {
+      row[0] = pendingFirstCol;
+      pendingFirstCol = "";
+    } else {
+      pendingFirstCol = "";
+    }
+    uniqueRows.push(row);
   }
   if (uniqueRows.length === 0) return "";
   const md = [];
@@ -195,75 +301,15 @@ function tableToMarkdown(table) {
   return md.join("\n");
 }
 
-// src/utils.ts
-var VERSION = true ? "1.7.2" : "0.0.0-dev";
-function toArrayBuffer(buf) {
-  if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
-    return buf.buffer;
-  }
-  return buf.buffer.slice(buf.byteOffset, buf.byteOffset + buf.byteLength);
-}
-var KordocError = class extends Error {
-  constructor(message) {
-    super(message);
-    this.name = "KordocError";
-  }
-};
-function sanitizeError(err) {
-  if (err instanceof KordocError) return err.message;
-  return "\uBB38\uC11C \uCC98\uB9AC \uC911 \uC624\uB958\uAC00 \uBC1C\uC0DD\uD588\uC2B5\uB2C8\uB2E4";
-}
-function isPathTraversal(name) {
-  if (name.includes("\0")) return true;
-  const normalized = name.replace(/\\/g, "/");
-  return normalized.includes("..") || normalized.startsWith("/") || /^[A-Za-z]:/.test(normalized);
-}
-function classifyError(err) {
-  if (!(err instanceof Error)) return "PARSE_ERROR";
-  const msg = err.message;
-  if (msg.includes("\uC554\uD638\uD654")) return "ENCRYPTED";
-  if (msg.includes("DRM")) return "DRM_PROTECTED";
-  if (msg.includes("ZIP bomb") || msg.includes("ZIP \uBE44\uC555\uCD95 \uD06C\uAE30 \uCD08\uACFC") || msg.includes("ZIP \uC5D4\uD2B8\uB9AC \uC218 \uCD08\uACFC")) return "ZIP_BOMB";
-  if (msg.includes("bomb") || msg.includes("\uD06C\uAE30 \uCD08\uACFC") || msg.includes("\uC555\uCD95 \uD574\uC81C")) return "DECOMPRESSION_BOMB";
-  if (msg.includes("\uC774\uBBF8\uC9C0 \uAE30\uBC18")) return "IMAGE_BASED_PDF";
-  if (msg.includes("\uC139\uC158") && (msg.includes("\uCC3E\uC744 \uC218 \uC5C6") || msg.includes("\uC5C6\uC74C"))) return "NO_SECTIONS";
-  if (msg.includes("\uC2DC\uADF8\uB2C8\uCC98") || msg.includes("\uBCF5\uAD6C\uD560 \uC218 \uC5C6")) return "CORRUPTED";
-  return "PARSE_ERROR";
-}
-
 // src/hwpx/parser.ts
-import JSZip from "jszip";
+import JSZip2 from "jszip";
 import { inflateRawSync } from "zlib";
 import { DOMParser } from "@xmldom/xmldom";
 
-// src/page-range.ts
-function parsePageRange(spec, maxPages) {
-  const result = /* @__PURE__ */ new Set();
-  if (maxPages <= 0) return result;
-  if (Array.isArray(spec)) {
-    for (const n of spec) {
-      const page = Math.round(n);
-      if (page >= 1 && page <= maxPages) result.add(page);
-    }
-    return result;
-  }
-  if (typeof spec !== "string" || spec.trim() === "") return result;
-  const parts = spec.split(",");
-  for (const part of parts) {
-    const trimmed = part.trim();
-    if (!trimmed) continue;
-    const rangeMatch = trimmed.match(/^(\d+)\s*-\s*(\d+)$/);
-    if (rangeMatch) {
-      const start = Math.max(1, parseInt(rangeMatch[1], 10));
-      const end = Math.min(maxPages, parseInt(rangeMatch[2], 10));
-      for (let i = start; i <= end; i++) result.add(i);
-    } else {
-      const page = parseInt(trimmed, 10);
-      if (!isNaN(page) && page >= 1 && page <= maxPages) result.add(page);
-    }
-  }
-  return result;
-}
+// src/types.ts
+var HEADING_RATIO_H1 = 1.5;
+var HEADING_RATIO_H2 = 1.3;
+var HEADING_RATIO_H3 = 1.15;
 
 // src/hwpx/parser.ts
 var MAX_DECOMPRESS_SIZE = 100 * 1024 * 1024;
@@ -357,16 +403,10 @@ function stripDtd(xml) {
   return xml.replace(/<!DOCTYPE\s[^[>]*(\[[\s\S]*?\])?\s*>/gi, "");
 }
 async function parseHwpxDocument(buffer, options) {
-  const precheck = precheckZipSize(buffer);
-  if (precheck.totalUncompressed > MAX_DECOMPRESS_SIZE) {
-    throw new KordocError("ZIP \uBE44\uC555\uCD95 \uD06C\uAE30 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
-  }
-  if (precheck.entryCount > MAX_ZIP_ENTRIES) {
-    throw new KordocError("ZIP \uC5D4\uD2B8\uB9AC \uC218 \uCD08\uACFC (ZIP bomb \uC758\uC2EC)");
-  }
+  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE, MAX_ZIP_ENTRIES);
   let zip;
   try {
-    zip = await JSZip.loadAsync(buffer);
+    zip = await JSZip2.loadAsync(buffer);
   } catch {
     return extractFromBrokenZip(buffer);
   }
@@ -529,7 +569,7 @@ function parseDublinCoreMetadata(xml, metadata) {
 async function extractHwpxMetadataOnly(buffer) {
   let zip;
   try {
-    zip = await JSZip.loadAsync(buffer);
+    zip = await JSZip2.loadAsync(buffer);
   } catch {
     throw new KordocError("HWPX ZIP\uC744 \uC5F4 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
   }
@@ -539,46 +579,17 @@ async function extractHwpxMetadataOnly(buffer) {
   metadata.pageCount = sectionPaths.length;
   return metadata;
 }
-function precheckZipSize(buffer) {
-  try {
-    const data = new DataView(buffer);
-    const len = buffer.byteLength;
-    if (len < 22) return { totalUncompressed: 0, entryCount: 0 };
-    const searchStart = Math.max(0, len - 22 - 65535);
-    let eocdOffset = -1;
-    for (let i = len - 22; i >= searchStart; i--) {
-      if (data.getUint32(i, true) === 101010256) {
-        eocdOffset = i;
-        break;
-      }
-    }
-    if (eocdOffset < 0) return { totalUncompressed: 0, entryCount: 0 };
-    const entryCount = data.getUint16(eocdOffset + 10, true);
-    const cdSize = data.getUint32(eocdOffset + 12, true);
-    const cdOffset = data.getUint32(eocdOffset + 16, true);
-    if (cdOffset + cdSize > len) return { totalUncompressed: 0, entryCount };
-    let totalUncompressed = 0;
-    let pos = cdOffset;
-    for (let i = 0; i < entryCount && pos + 46 <= cdOffset + cdSize; i++) {
-      if (data.getUint32(pos, true) !== 33639248) break;
-      totalUncompressed += data.getUint32(pos + 24, true);
-      const nameLen = data.getUint16(pos + 28, true);
-      const extraLen = data.getUint16(pos + 30, true);
-      const commentLen = data.getUint16(pos + 32, true);
-      pos += 46 + nameLen + extraLen + commentLen;
-    }
-    return { totalUncompressed, entryCount };
-  } catch {
-    return { totalUncompressed: 0, entryCount: 0 };
-  }
-}
 function extractFromBrokenZip(buffer) {
   const data = new Uint8Array(buffer);
   const view = new DataView(buffer);
   let pos = 0;
   const blocks = [];
+  const warnings = [
+    { code: "BROKEN_ZIP_RECOVERY", message: "\uC190\uC0C1\uB41C ZIP \uAD6C\uC870 \u2014 Local File Header \uAE30\uBC18 \uBCF5\uAD6C \uBAA8\uB4DC" }
+  ];
   let totalDecompressed = 0;
   let entryCount = 0;
+  let sectionNum = 0;
   while (pos < data.length - 30) {
     if (data[pos] !== 80 || data[pos + 1] !== 75 || data[pos + 2] !== 3 || data[pos + 3] !== 4) {
       pos++;
@@ -624,14 +635,15 @@ function extractFromBrokenZip(buffer) {
       }
       totalDecompressed += content.length * 2;
       if (totalDecompressed > MAX_DECOMPRESS_SIZE) throw new KordocError("\uC555\uCD95 \uD574\uC81C \uD06C\uAE30 \uCD08\uACFC");
-      blocks.push(...parseSectionXml(content));
+      sectionNum++;
+      blocks.push(...parseSectionXml(content, void 0, warnings, sectionNum));
     } catch {
       continue;
     }
   }
   if (blocks.length === 0) throw new KordocError("\uC190\uC0C1\uB41C HWPX\uC5D0\uC11C \uC139\uC158 \uB370\uC774\uD130\uB97C \uBCF5\uAD6C\uD560 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
   const markdown = blocksToMarkdown(blocks);
-  return { markdown, blocks };
+  return { markdown, blocks, warnings: warnings.length > 0 ? warnings : void 0 };
 }
 async function resolveSectionPaths(zip) {
   const manifestPaths = ["Contents/content.hpf", "content.hpf"];
@@ -695,9 +707,9 @@ function detectHwpxHeadings(blocks, styleMap) {
     let level = 0;
     if (baseFontSize > 0 && block.style?.fontSize) {
       const ratio = block.style.fontSize / baseFontSize;
-      if (ratio >= 1.5) level = 1;
-      else if (ratio >= 1.3) level = 2;
-      else if (ratio >= 1.15) level = 3;
+      if (ratio >= HEADING_RATIO_H1) level = 1;
+      else if (ratio >= HEADING_RATIO_H2) level = 2;
+      else if (ratio >= HEADING_RATIO_H3) level = 3;
     }
     if (/^제\d+[조장절편]/.test(text) && text.length <= 50) {
       if (level === 0) level = 3;
@@ -783,6 +795,14 @@ function walkSection(node, blocks, tableCtx, tableStack, styleMap, warnings, sec
           }
         }
         break;
+      case "cellAddr":
+        if (tableCtx?.cell) {
+          const ca = parseInt(el.getAttribute("colAddr") || "", 10);
+          const ra = parseInt(el.getAttribute("rowAddr") || "", 10);
+          if (!isNaN(ca)) tableCtx.cell.colAddr = ca;
+          if (!isNaN(ra)) tableCtx.cell.rowAddr = ra;
+        }
+        break;
       case "cellSpan":
         if (tableCtx?.cell) {
           const cs = parseInt(el.getAttribute("colSpan") || "1", 10);
@@ -829,39 +849,47 @@ function walkParagraphChildren(node, blocks, tableCtx, tableStack, styleMap, war
   if (depth > MAX_XML_DEPTH) return tableCtx;
   const children = node.childNodes;
   if (!children) return tableCtx;
-  for (let i = 0; i < children.length; i++) {
-    const el = children[i];
-    if (el.nodeType !== 1) continue;
-    const tag = el.tagName || el.localName || "";
-    const localTag = tag.replace(/^[^:]+:/, "");
-    if (localTag === "tbl") {
-      if (tableCtx) tableStack.push(tableCtx);
-      const newTable = { rows: [], currentRow: [], cell: null };
-      walkSection(el, blocks, newTable, tableStack, styleMap, warnings, sectionNum, depth + 1);
-      if (newTable.rows.length > 0) {
-        if (tableStack.length > 0) {
-          const parentTable = tableStack.pop();
-          const nestedText = convertTableToText(newTable.rows);
-          if (parentTable.cell) {
-            parentTable.cell.text += (parentTable.cell.text ? "\n" : "") + nestedText;
+  const walkChildren = (parent, d) => {
+    if (d > MAX_XML_DEPTH) return;
+    const kids = parent.childNodes;
+    if (!kids) return;
+    for (let i = 0; i < kids.length; i++) {
+      const el = kids[i];
+      if (el.nodeType !== 1) continue;
+      const tag = el.tagName || el.localName || "";
+      const localTag = tag.replace(/^[^:]+:/, "");
+      if (localTag === "tbl") {
+        if (tableCtx) tableStack.push(tableCtx);
+        const newTable = { rows: [], currentRow: [], cell: null };
+        walkSection(el, blocks, newTable, tableStack, styleMap, warnings, sectionNum, d + 1);
+        if (newTable.rows.length > 0) {
+          if (tableStack.length > 0) {
+            const parentTable = tableStack.pop();
+            const nestedText = convertTableToText(newTable.rows);
+            if (parentTable.cell) {
+              parentTable.cell.text += (parentTable.cell.text ? "\n" : "") + nestedText;
+            }
+            tableCtx = parentTable;
+          } else {
+            blocks.push({ type: "table", table: buildTable(newTable.rows), pageNumber: sectionNum });
+            tableCtx = null;
           }
-          tableCtx = parentTable;
         } else {
-          blocks.push({ type: "table", table: buildTable(newTable.rows), pageNumber: sectionNum });
-          tableCtx = null;
+          tableCtx = tableStack.length > 0 ? tableStack.pop() : null;
         }
-      } else {
-        tableCtx = tableStack.length > 0 ? tableStack.pop() : null;
-      }
-    } else if (localTag === "pic" || localTag === "shape" || localTag === "drawingObject") {
-      const imgRef = extractImageRef(el);
-      if (imgRef) {
-        blocks.push({ type: "image", text: imgRef, pageNumber: sectionNum });
-      } else if (warnings && sectionNum) {
-        warnings.push({ page: sectionNum, message: `\uC2A4\uD0B5\uB41C \uC694\uC18C: ${localTag}`, code: "SKIPPED_IMAGE" });
+      } else if (localTag === "pic" || localTag === "shape" || localTag === "drawingObject") {
+        const imgRef = extractImageRef(el);
+        if (imgRef) {
+          blocks.push({ type: "image", text: imgRef, pageNumber: sectionNum });
+        } else if (warnings && sectionNum) {
+          warnings.push({ page: sectionNum, message: `\uC2A4\uD0B5\uB41C \uC694\uC18C: ${localTag}`, code: "SKIPPED_IMAGE" });
+        }
+      } else if (localTag === "r" || localTag === "run" || localTag === "ctrl") {
+        walkChildren(el, d + 1);
       }
     }
-  }
+  };
+  walkChildren(node, depth);
   return tableCtx;
 }
 function extractParagraphInfo(para, styleMap) {
@@ -900,7 +928,10 @@ function extractParagraphInfo(para, styleMap) {
         // 하이퍼링크
         case "hyperlink": {
           const url = child.getAttribute("url") || child.getAttribute("href") || "";
-          if (url) href = url;
+          if (url) {
+            const safe = sanitizeHref(url);
+            if (safe) href = safe;
+          }
           walk(child);
           break;
         }
@@ -913,6 +944,29 @@ function extractParagraphInfo(para, styleMap) {
           if (noteText) footnote = (footnote ? footnote + "; " : "") + noteText;
           break;
         }
+        // 제어 요소 — 필드, 컨트롤, 매개변수 등 스킵
+        case "ctrl":
+        case "fieldBegin":
+        case "fieldEnd":
+        case "parameters":
+        case "stringParam":
+        case "integerParam":
+        case "boolParam":
+        case "floatParam":
+        case "secPr":
+        // 섹션 속성 (페이지 설정 등)
+        case "colPr":
+        // 다단 속성
+        case "linesegarray":
+        case "lineseg":
+        // 레이아웃 정보
+        // 도형/이미지 요소 — 대체텍스트("사각형입니다." 등) 누출 방지
+        case "pic":
+        case "shape":
+        case "drawingObject":
+        case "shapeComment":
+        case "drawText":
+          break;
         // run 요소에서 charPrIDRef 추출
         case "r": {
           const runCharPr = child.getAttribute("charPrIDRef");
@@ -927,7 +981,10 @@ function extractParagraphInfo(para, styleMap) {
     }
   };
   walk(para);
-  const cleanText = text.replace(/[ \t]+/g, " ").trim();
+  let cleanText = text.replace(/[ \t]+/g, " ").trim();
+  if (/^그림입니다\.?\s*원본\s*그림의\s*(이름|크기)/.test(cleanText)) cleanText = "";
+  cleanText = cleanText.replace(/그림입니다\.?\s*원본\s*그림의\s*(이름|크기)[^\n]*(\n[^\n]*원본\s*그림의\s*(이름|크기)[^\n]*)*/g, "").trim();
+  cleanText = cleanText.replace(/(?:모서리가 둥근 |둥근 )?(?:사각형|직사각형|정사각형|원|타원|삼각형|선|직선|곡선|화살표|오각형|육각형|팔각형|별|십자|구름|마름모|도넛|평행사변형|사다리꼴|개체|그리기\s?개체|묶음\s?개체|글상자|수식|표|그림|OLE\s?개체)\s?입니다\.?/g, "").trim();
   let style;
   if (styleMap && charPrId) {
     const charProp = styleMap.charProperties.get(charPrId);
@@ -1205,9 +1262,9 @@ function detectHwp5Headings(blocks, docInfo) {
     if (/^\d+$/.test(text)) continue;
     const ratio = block.style.fontSize / baseFontSize;
     let level = 0;
-    if (ratio >= 1.5) level = 1;
-    else if (ratio >= 1.3) level = 2;
-    else if (ratio >= 1.15) level = 3;
+    if (ratio >= HEADING_RATIO_H1) level = 1;
+    else if (ratio >= HEADING_RATIO_H2) level = 2;
+    else if (ratio >= HEADING_RATIO_H3) level = 3;
     if (/^제\d+[조장절편]/.test(text) && text.length <= 50) {
       if (level === 0) level = 3;
     }
@@ -1308,20 +1365,22 @@ function detectImageMime(data) {
 }
 function extractHwp5Images(cfb, blocks, compressed, warnings) {
   const binDataMap = /* @__PURE__ */ new Map();
-  for (let idx = 0; idx < 1e4; idx++) {
-    const entry = CFB.find(cfb, `/BinData/BIN${String(idx).padStart(4, "0")}`) || CFB.find(cfb, `/BinData/Bin${String(idx).padStart(4, "0")}`);
-    if (!entry?.content) {
-      if (idx > 0) break;
-      continue;
-    }
-    let data = Buffer.from(entry.content);
-    if (compressed) {
-      try {
-        data = decompressStream(data);
-      } catch {
+  const binDataRe = /\/BinData\/[Bb][Ii][Nn](\d{4})$/;
+  if (cfb.FileIndex) {
+    for (const entry of cfb.FileIndex) {
+      if (!entry?.name || !entry.content) continue;
+      const match = entry.name.match(binDataRe);
+      if (!match) continue;
+      const idx = parseInt(match[1], 10);
+      let data = Buffer.from(entry.content);
+      if (compressed) {
+        try {
+          data = decompressStream(data);
+        } catch {
+        }
       }
+      binDataMap.set(idx, { data, name: entry.name });
     }
-    binDataMap.set(idx, { data, name: entry.name || `BIN${idx}` });
   }
   if (binDataMap.size === 0) return [];
   const images = [];
@@ -1468,6 +1527,16 @@ function parseTableBlock(records, startIdx) {
     i++;
   }
   if (rows === 0 || cols === 0 || cells.length === 0) return { table: null, nextIdx: i };
+  const hasAddr = cells.some((c) => c.colAddr !== void 0 && c.rowAddr !== void 0);
+  if (hasAddr) {
+    const cellRows2 = arrangeCells(rows, cols, cells);
+    const irCells = cellRows2.map((row) => row.map((c) => ({
+      text: c.text.trim(),
+      colSpan: c.colSpan,
+      rowSpan: c.rowSpan
+    })));
+    return { table: { rows, cols, cells: irCells, hasHeader: rows > 1 }, nextIdx: i };
+  }
   const cellRows = arrangeCells(rows, cols, cells);
   return { table: buildTable(cellRows), nextIdx: i };
 }
@@ -1731,7 +1800,36 @@ function buildTableGrids(horizontals, verticals) {
     };
     grids.push({ rowYs, colXs, bbox });
   }
-  return grids;
+  return mergeAdjacentGrids(grids);
+}
+function mergeAdjacentGrids(grids) {
+  if (grids.length <= 1) return grids;
+  const sorted = [...grids].sort((a, b) => b.bbox.y2 - a.bbox.y2);
+  const merged = [sorted[0]];
+  for (let i = 1; i < sorted.length; i++) {
+    const prev = merged[merged.length - 1];
+    const curr = sorted[i];
+    if (prev.colXs.length === curr.colXs.length) {
+      const colMatch = prev.colXs.every((x, ci) => Math.abs(x - curr.colXs[ci]) <= COORD_MERGE_TOL * 3);
+      const verticalGap = prev.bbox.y1 - curr.bbox.y2;
+      if (colMatch && verticalGap >= -COORD_MERGE_TOL && verticalGap <= 20) {
+        const allRowYs = [.../* @__PURE__ */ new Set([...prev.rowYs, ...curr.rowYs])].sort((a, b) => b - a);
+        merged[merged.length - 1] = {
+          rowYs: allRowYs,
+          colXs: prev.colXs,
+          bbox: {
+            x1: Math.min(prev.bbox.x1, curr.bbox.x1),
+            y1: Math.min(prev.bbox.y1, curr.bbox.y1),
+            x2: Math.max(prev.bbox.x2, curr.bbox.x2),
+            y2: Math.max(prev.bbox.y2, curr.bbox.y2)
+          }
+        };
+        continue;
+      }
+    }
+    merged.push(curr);
+  }
+  return merged;
 }
 function clusterCoordinates(values) {
   if (values.length === 0) return [];
@@ -1918,7 +2016,11 @@ function cellTextToString(items) {
     for (let j = 1; j < s.length; j++) {
       const gap = s[j].x - (s[j - 1].x + s[j - 1].w);
       const avgFs = (s[j].fontSize + s[j - 1].fontSize) / 2;
-      if (gap < avgFs * 0.3 && /[가-힣]$/.test(result) && /^[가-힣]/.test(s[j].text)) {
+      const prevIsKorean = /[가-힣]$/.test(result);
+      const currIsKorean = /^[가-힣]/.test(s[j].text);
+      if (gap < avgFs * 0.15) {
+        result += s[j].text;
+      } else if (gap < avgFs * 0.35 && (prevIsKorean || currIsKorean)) {
         result += s[j].text;
       } else {
         result += " " + s[j].text;
@@ -1933,6 +2035,12 @@ function cellTextToString(items) {
     const curr = textLines[i];
     if (/[가-힣]$/.test(prev) && /^[가-힣]+$/.test(curr) && curr.length <= 8 && !curr.includes(" ")) {
       merged[merged.length - 1] = prev + curr;
+    } else if (curr.trim().length <= 3 && /^[)\]%}]/.test(curr.trim())) {
+      merged[merged.length - 1] = prev + curr.trim();
+    } else if (/[,(]$/.test(prev.trim()) && curr.trim().length <= 15) {
+      merged[merged.length - 1] = prev + curr.trim();
+    } else if (/[\d,]$/.test(prev) && /^[\d,]+[)\]]?$/.test(curr.trim()) && curr.trim().length <= 10) {
+      merged[merged.length - 1] = prev + curr.trim();
     } else {
       merged.push(curr);
     }
@@ -2145,21 +2253,26 @@ async function loadPdfWithTimeout(buffer) {
     disableFontFace: true,
     isEvalSupported: false
   });
-  return Promise.race([
-    loadingTask.promise,
-    new Promise(
-      (_, reject) => setTimeout(() => {
-        loadingTask.destroy();
-        reject(new KordocError("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
-      }, PDF_LOAD_TIMEOUT_MS)
-    )
-  ]);
+  let timer;
+  try {
+    return await Promise.race([
+      loadingTask.promise,
+      new Promise((_, reject) => {
+        timer = setTimeout(() => {
+          loadingTask.destroy();
+          reject(new KordocError("PDF \uB85C\uB529 \uD0C0\uC784\uC544\uC6C3 (30\uCD08 \uCD08\uACFC)"));
+        }, PDF_LOAD_TIMEOUT_MS);
+      })
+    ]);
+  } finally {
+    if (timer !== void 0) clearTimeout(timer);
+  }
 }
 async function parsePdfDocument(buffer, options) {
   const doc = await loadPdfWithTimeout(buffer);
   try {
     const pageCount = doc.numPages;
-    if (pageCount === 0) return { success: false, fileType: "pdf", pageCount: 0, error: "PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.", code: "PARSE_ERROR" };
+    if (pageCount === 0) throw new KordocError("PDF\uC5D0 \uD398\uC774\uC9C0\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4.");
     const metadata = { pageCount };
     await extractPdfMetadata(doc, metadata);
     const blocks = [];
@@ -2212,14 +2325,14 @@ async function parsePdfDocument(buffer, options) {
           const ocrBlocks = await ocrPages(doc, options.ocr, pageFilter, effectivePageCount);
           if (ocrBlocks.length > 0) {
             const ocrMarkdown = ocrBlocks.map((b) => b.text || "").filter(Boolean).join("\n\n");
-            return { success: true, fileType: "pdf", markdown: ocrMarkdown, pageCount: parsedPageCount, blocks: ocrBlocks, metadata, isImageBased: true, warnings };
+            return { markdown: ocrMarkdown, blocks: ocrBlocks, metadata, warnings, isImageBased: true };
           }
         } catch {
         }
       }
-      return { success: false, fileType: "pdf", pageCount, isImageBased: true, error: `\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`, code: "IMAGE_BASED_PDF" };
+      throw Object.assign(new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
     }
-    if (options?.removeHeaderFooter && parsedPageCount >= 3) {
+    if (options?.removeHeaderFooter !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);
       for (let ri = removed.length - 1; ri >= 0; ri--) {
         blocks.splice(removed[ri], 1);
@@ -2229,9 +2342,10 @@ async function parsePdfDocument(buffer, options) {
     if (medianFontSize > 0) {
       detectHeadings(blocks, medianFontSize);
     }
+    detectMarkerHeadings(blocks);
     const outline = blocks.filter((b) => b.type === "heading" && b.level && b.text).map((b) => ({ level: b.level, text: b.text, pageNumber: b.pageNumber }));
     let markdown = cleanPdfText(blocksToMarkdown(blocks));
-    return { success: true, fileType: "pdf", markdown, pageCount: parsedPageCount, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0 };
+    return { markdown, blocks, metadata, outline: outline.length > 0 ? outline : void 0, warnings: warnings.length > 0 ? warnings : void 0 };
   } finally {
     await doc.destroy().catch(() => {
     });
@@ -2302,12 +2416,67 @@ function detectHeadings(blocks, medianFontSize) {
     if (/^\d+$/.test(text)) continue;
     const ratio = block.style.fontSize / medianFontSize;
     let level = 0;
-    if (ratio >= 1.5) level = 1;
-    else if (ratio >= 1.3) level = 2;
-    else if (ratio >= 1.15) level = 3;
+    if (ratio >= HEADING_RATIO_H1) level = 1;
+    else if (ratio >= HEADING_RATIO_H2) level = 2;
+    else if (ratio >= HEADING_RATIO_H3) level = 3;
     if (level > 0) {
       block.type = "heading";
       block.level = level;
+      block.text = collapseEvenSpacing(text);
+    }
+  }
+}
+function collapseEvenSpacing(text) {
+  const tokens = text.split(" ");
+  const singleCharCount = tokens.filter((t) => t.length === 1).length;
+  if (tokens.length >= 3 && singleCharCount / tokens.length >= 0.7) {
+    return tokens.join("");
+  }
+  return text;
+}
+function shouldDemoteTable(table) {
+  const allCells = table.cells.flatMap((row) => row.map((c) => c.text.trim())).filter(Boolean);
+  const allText = allCells.join(" ");
+  if (allText.length > 200) return false;
+  if (/[□■◆○●▶]/.test(allText) && table.rows <= 3) return true;
+  const totalCells = table.rows * table.cols;
+  const emptyCells = totalCells - allCells.length;
+  if (table.rows <= 2 && emptyCells > totalCells * 0.5) return true;
+  if (table.rows === 1 && !/\d{2,}/.test(allText)) return true;
+  return false;
+}
+function demoteTableToText(table) {
+  const lines = [];
+  for (let r = 0; r < table.rows; r++) {
+    const cells = table.cells[r].map((c) => c.text.trim()).filter(Boolean);
+    if (cells.length === 0) continue;
+    if (table.cols === 2 && cells.length === 2) {
+      lines.push(`${cells[0]} : ${cells[1]}`);
+    } else {
+      lines.push(cells.join(" "));
+    }
+  }
+  return lines.join("\n");
+}
+function detectMarkerHeadings(blocks) {
+  for (let i = 0; i < blocks.length; i++) {
+    const block = blocks[i];
+    if (block.type !== "paragraph" || !block.text) continue;
+    const text = block.text.trim();
+    if (text.length < 50 && /^[□■◆◇▶]\s*[가-힣]/.test(text)) {
+      block.type = "heading";
+      block.level = 4;
+      continue;
+    }
+    if (/^[가-힣]{2,6}$/.test(text) && block.style?.fontSize) {
+      const prev = blocks[i - 1];
+      const next = blocks[i + 1];
+      const prevIsStructural = !prev || prev.type === "table" || prev.type === "heading" || prev.type === "separator";
+      const nextIsStructural = !next || next.type === "table" || next.type === "heading" || next.type === "paragraph" && next.text && /^[□■◆○●]/.test(next.text.trim());
+      if (prevIsStructural || nextIsStructural) {
+        block.type = "heading";
+        block.level = 3;
+      }
     }
   }
 }
@@ -2344,7 +2513,7 @@ function computeRegion(items) {
   }
   return { items, minX, minY, maxX, maxY };
 }
-function findYSplit(items, region, gapThreshold) {
+function findYSplit(items, _region, gapThreshold) {
   const sorted = [...items].sort((a, b) => b.y - a.y);
   let bestGap = gapThreshold;
   let bestSplit = null;
@@ -2359,7 +2528,7 @@ function findYSplit(items, region, gapThreshold) {
   }
   return bestSplit;
 }
-function findXSplit(items, region, gapThreshold) {
+function findXSplit(items, _region, gapThreshold) {
   const sorted = [...items].sort((a, b) => a.x - b.x);
   let bestGap = gapThreshold;
   let bestSplit = null;
@@ -2418,7 +2587,8 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
     );
     for (const cell of cells) {
       const cellItems = cellTextMap.get(cell) || [];
-      const text = cellTextToString(cellItems);
+      let text = cellTextToString(cellItems);
+      text = text.replace(/^[\s]*[-–—]\s*\d+\s*[-–—][\s]*$/gm, "").trim();
       irGrid[cell.row][cell.col] = {
         text,
         colSpan: cell.colSpan,
@@ -2433,18 +2603,21 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
     };
     const hasContent = irGrid.some((row) => row.some((cell) => cell.text.trim() !== ""));
     if (!hasContent) continue;
-    blocks.push({
-      type: "table",
-      table: irTable,
-      pageNumber: pageNum,
-      bbox: {
-        page: pageNum,
-        x: grid.bbox.x1,
-        y: grid.bbox.y1,
-        width: grid.bbox.x2 - grid.bbox.x1,
-        height: grid.bbox.y2 - grid.bbox.y1
+    const tableBbox = {
+      page: pageNum,
+      x: grid.bbox.x1,
+      y: grid.bbox.y1,
+      width: grid.bbox.x2 - grid.bbox.x1,
+      height: grid.bbox.y2 - grid.bbox.y1
+    };
+    if (shouldDemoteTable(irTable)) {
+      const demoted = demoteTableToText(irTable);
+      if (demoted) {
+        blocks.push({ type: "paragraph", text: demoted, pageNumber: pageNum, bbox: tableBbox, style: dominantStyle(tableItems) });
       }
-    });
+      continue;
+    }
+    blocks.push({ type: "table", table: irTable, pageNumber: pageNum, bbox: tableBbox });
   }
   const remaining = items.filter((i) => !usedItems.has(i));
   if (remaining.length > 0) {
@@ -2456,9 +2629,29 @@ function extractBlocksWithGrids(items, pageNum, grids, horizontals, verticals) {
       const by = b.bbox ? b.bbox.y + b.bbox.height : 0;
       return by - ay;
     });
-    return allBlocks;
+    return mergeAdjacentTableBlocks(allBlocks);
   }
-  return blocks;
+  return mergeAdjacentTableBlocks(blocks);
+}
+function mergeAdjacentTableBlocks(blocks) {
+  if (blocks.length <= 1) return blocks;
+  const result = [blocks[0]];
+  for (let i = 1; i < blocks.length; i++) {
+    const prev = result[result.length - 1];
+    const curr = blocks[i];
+    if (prev.type === "table" && curr.type === "table" && prev.table && curr.table && prev.table.cols === curr.table.cols) {
+      const merged = {
+        rows: prev.table.rows + curr.table.rows,
+        cols: prev.table.cols,
+        cells: [...prev.table.cells, ...curr.table.cells],
+        hasHeader: prev.table.hasHeader
+      };
+      result[result.length - 1] = { ...prev, table: merged };
+    } else {
+      result.push(curr);
+    }
+  }
+  return result;
 }
 function extractPageBlocksFallback(items, pageNum) {
   if (items.length === 0) return [];
@@ -2481,11 +2674,13 @@ function extractPageBlocksFallback(items, pageNum) {
     }));
     const clusterResults = detectClusterTables(clusterItems, pageNum);
     if (clusterResults.length > 0) {
+      const ciToIdx = /* @__PURE__ */ new Map();
+      for (let ci = 0; ci < clusterItems.length; ci++) ciToIdx.set(clusterItems[ci], ci);
       const usedIndices = /* @__PURE__ */ new Set();
       for (const cr of clusterResults) {
         for (const ci of cr.usedItems) {
-          const idx = clusterItems.indexOf(ci);
-          if (idx >= 0) usedIndices.add(idx);
+          const idx = ciToIdx.get(ci);
+          if (idx !== void 0) usedIndices.add(idx);
         }
         blocks.push({ type: "table", table: cr.table, pageNumber: pageNum, bbox: cr.bbox });
       }
@@ -2796,7 +2991,8 @@ function mergeLineSimple(items) {
     const gap = sorted[i].x - (sorted[i - 1].x + sorted[i - 1].w);
     const avgFs = (sorted[i].fontSize + sorted[i - 1].fontSize) / 2;
     if (gap > 15) result += "	";
-    else if (gap < avgFs * 0.3 && /[가-힣]$/.test(result) && /^[가-힣]/.test(sorted[i].text)) {
+    else if (gap < avgFs * 0.15) {
+    } else if (gap < avgFs * 0.35 && (/[가-힣]$/.test(result) || /^[가-힣]/.test(sorted[i].text))) {
     } else if (gap > 3) result += " ";
     result += sorted[i].text;
   }
@@ -2804,8 +3000,8 @@ function mergeLineSimple(items) {
 }
 function cleanPdfText(text) {
   return mergeKoreanLines(
-    text.replace(/^[\s]*[-–—]\s*\d+\s*[-–—][\s]*$/gm, "").replace(/^\s*\d+\s*\/\s*\d+\s*$/gm, "")
-  ).replace(/\n{3,}/g, "\n\n").trim();
+    text.replace(/^[\s]*[-–—]\s*[-–—]?\d+[-–—]?[\s]*[-–—]?[\s]*$/gm, "").replace(/^\s*\d+\s*\/\s*\d+\s*$/gm, "").replace(/\n\d{1,4}\n/g, "\n").replace(/\n\d{1,4}$/, "")
+  ).replace(/^(?!\|).{3,30}$/gm, (line) => collapseEvenSpacing(line)).replace(/\n{3,}/g, "\n\n").trim();
 }
 function startsWithMarker(line) {
   const t = line.trimStart();
@@ -2819,15 +3015,13 @@ function detectListBlocks(blocks) {
   for (let i = 0; i < blocks.length; i++) {
     const block = blocks[i];
     if (block.type === "paragraph" && block.text) {
-      const match = block.text.match(/^(\d+)\.\s/);
-      if (match) {
-        result.push({
-          ...block,
-          type: "list",
-          listType: "ordered",
-          // 원래 번호를 text에 보존 (blocksToMarkdown에서 그대로 출력)
-          text: block.text
-        });
+      const text = block.text.trim();
+      if (/^\d+\.\s/.test(text)) {
+        result.push({ ...block, type: "list", listType: "ordered", text: block.text });
+        continue;
+      }
+      if (/^[○●·※▶▷◆◇\-]\s/.test(text)) {
+        result.push({ ...block, type: "list", listType: "unordered", text: block.text });
         continue;
       }
     }
@@ -2986,11 +3180,20 @@ function mergeKoreanLines(text) {
   for (let i = 1; i < lines.length; i++) {
     const prev = result[result.length - 1];
     const curr = lines[i];
-    if (/^#{1,6}\s/.test(prev) || /^#{1,6}\s/.test(curr)) {
+    const currTrimmed = curr.trim();
+    if (/^#{1,6}\s/.test(prev) || /^#{1,6}\s/.test(curr) || /^\|/.test(currTrimmed) || /^---/.test(currTrimmed)) {
       result.push(curr);
       continue;
     }
-    if (/[가-힣·,\-]$/.test(prev) && /^[가-힣(]/.test(curr) && !curr.trimStart().startsWith("|") && !startsWithMarker(curr) && !isStandaloneHeader(prev)) {
+    if (/,$/.test(prev.trim()) && currTrimmed.length > 0) {
+      result[result.length - 1] = prev + "\n" + curr;
+      continue;
+    }
+    if (/^\(※/.test(currTrimmed)) {
+      result[result.length - 1] = prev + " " + currTrimmed;
+      continue;
+    }
+    if (/[가-힣·,\-]$/.test(prev) && /^[가-힣(]/.test(curr) && !startsWithMarker(curr) && !isStandaloneHeader(prev)) {
       result[result.length - 1] = prev + " " + curr;
     } else {
       result.push(curr);
@@ -3002,6 +3205,716 @@ function mergeKoreanLines(text) {
 // src/index.ts
 import { readFile } from "fs/promises";
 
+// src/xlsx/parser.ts
+import JSZip3 from "jszip";
+import { DOMParser as DOMParser2 } from "@xmldom/xmldom";
+var MAX_SHEETS = 100;
+var MAX_DECOMPRESS_SIZE3 = 100 * 1024 * 1024;
+var MAX_ROWS2 = 1e4;
+var MAX_COLS2 = 200;
+function cleanNumericValue(raw) {
+  if (!/^-?\d+\.\d+$/.test(raw)) return raw;
+  const num = parseFloat(raw);
+  if (!isFinite(num)) return raw;
+  const cleaned = parseFloat(num.toPrecision(15)).toString();
+  return cleaned;
+}
+function parseCellRef(ref) {
+  const m = ref.match(/^([A-Z]+)(\d+)$/);
+  if (!m) return null;
+  let col = 0;
+  for (const ch of m[1]) col = col * 26 + (ch.charCodeAt(0) - 64);
+  return { col: col - 1, row: parseInt(m[2], 10) - 1 };
+}
+function parseMergeRef(ref) {
+  const parts = ref.split(":");
+  if (parts.length !== 2) return null;
+  const start = parseCellRef(parts[0]);
+  const end = parseCellRef(parts[1]);
+  if (!start || !end) return null;
+  return { startCol: start.col, startRow: start.row, endCol: end.col, endRow: end.row };
+}
+function getElements(parent, tagName) {
+  const nodes = parent.getElementsByTagName(tagName);
+  const result = [];
+  for (let i = 0; i < nodes.length; i++) result.push(nodes[i]);
+  return result;
+}
+function getTextContent(el) {
+  return el.textContent?.trim() ?? "";
+}
+function parseXml(text) {
+  return new DOMParser2().parseFromString(text, "text/xml");
+}
+function parseSharedStrings(xml) {
+  const doc = parseXml(xml);
+  const strings = [];
+  const siList = getElements(doc.documentElement, "si");
+  for (const si of siList) {
+    const tElements = getElements(si, "t");
+    strings.push(tElements.map((t) => t.textContent ?? "").join(""));
+  }
+  return strings;
+}
+function parseWorkbook(xml) {
+  const doc = parseXml(xml);
+  const sheets = [];
+  const sheetElements = getElements(doc.documentElement, "sheet");
+  for (const el of sheetElements) {
+    sheets.push({
+      name: el.getAttribute("name") ?? `Sheet${sheets.length + 1}`,
+      sheetId: el.getAttribute("sheetId") ?? "",
+      rId: el.getAttribute("r:id") ?? ""
+    });
+  }
+  return sheets;
+}
+function parseRels(xml) {
+  const doc = parseXml(xml);
+  const map = /* @__PURE__ */ new Map();
+  const rels = getElements(doc.documentElement, "Relationship");
+  for (const rel of rels) {
+    const id = rel.getAttribute("Id");
+    const target = rel.getAttribute("Target");
+    if (id && target) map.set(id, target);
+  }
+  return map;
+}
+function parseWorksheet(xml, sharedStrings) {
+  const doc = parseXml(xml);
+  const grid = [];
+  let maxRow = 0;
+  let maxCol = 0;
+  const rows = getElements(doc.documentElement, "row");
+  for (const rowEl of rows) {
+    const rowNum = parseInt(rowEl.getAttribute("r") ?? "0", 10) - 1;
+    if (rowNum < 0 || rowNum >= MAX_ROWS2) continue;
+    const cells = getElements(rowEl, "c");
+    for (const cellEl of cells) {
+      const ref = cellEl.getAttribute("r");
+      if (!ref) continue;
+      const pos = parseCellRef(ref);
+      if (!pos || pos.col >= MAX_COLS2) continue;
+      const type = cellEl.getAttribute("t");
+      const vElements = getElements(cellEl, "v");
+      const fElements = getElements(cellEl, "f");
+      let value = "";
+      if (vElements.length > 0) {
+        const raw = getTextContent(vElements[0]);
+        if (type === "s") {
+          const idx = parseInt(raw, 10);
+          value = sharedStrings[idx] ?? "";
+        } else if (type === "b") {
+          value = raw === "1" ? "TRUE" : "FALSE";
+        } else {
+          value = cleanNumericValue(raw);
+        }
+      } else if (type === "inlineStr") {
+        const isEl = getElements(cellEl, "is");
+        if (isEl.length > 0) {
+          const tElements = getElements(isEl[0], "t");
+          value = tElements.map((t) => t.textContent ?? "").join("");
+        }
+      }
+      if (!value && fElements.length > 0) {
+        value = `=${getTextContent(fElements[0])}`;
+      }
+      while (grid.length <= pos.row) grid.push([]);
+      while (grid[pos.row].length <= pos.col) grid[pos.row].push("");
+      grid[pos.row][pos.col] = value;
+      if (pos.row > maxRow) maxRow = pos.row;
+      if (pos.col > maxCol) maxCol = pos.col;
+    }
+  }
+  const merges = [];
+  const mergeCellElements = getElements(doc.documentElement, "mergeCell");
+  for (const el of mergeCellElements) {
+    const ref = el.getAttribute("ref");
+    if (!ref) continue;
+    const m = parseMergeRef(ref);
+    if (m) merges.push(m);
+  }
+  return { grid, merges, maxRow, maxCol };
+}
+function sheetToBlocks(sheetName, grid, merges, maxRow, maxCol, sheetIndex) {
+  const blocks = [];
+  if (sheetName) {
+    blocks.push({
+      type: "heading",
+      text: sheetName,
+      level: 2,
+      pageNumber: sheetIndex + 1
+    });
+  }
+  if (maxRow < 0 || maxCol < 0 || grid.length === 0) return blocks;
+  const mergeMap = /* @__PURE__ */ new Map();
+  const mergeSkip = /* @__PURE__ */ new Set();
+  for (const m of merges) {
+    const colSpan = m.endCol - m.startCol + 1;
+    const rowSpan = m.endRow - m.startRow + 1;
+    mergeMap.set(`${m.startRow},${m.startCol}`, { colSpan, rowSpan });
+    for (let r = m.startRow; r <= m.endRow; r++) {
+      for (let c = m.startCol; c <= m.endCol; c++) {
+        if (r !== m.startRow || c !== m.startCol) {
+          mergeSkip.add(`${r},${c}`);
+        }
+      }
+    }
+  }
+  let firstRow = -1;
+  let lastRow = -1;
+  for (let r = 0; r <= maxRow; r++) {
+    const row = grid[r];
+    if (row && row.some((cell) => cell !== "")) {
+      if (firstRow === -1) firstRow = r;
+      lastRow = r;
+    }
+  }
+  if (firstRow === -1) return blocks;
+  const cellRows = [];
+  for (let r = firstRow; r <= lastRow; r++) {
+    const row = [];
+    for (let c = 0; c <= maxCol; c++) {
+      const key = `${r},${c}`;
+      if (mergeSkip.has(key)) continue;
+      const text = (grid[r] && grid[r][c]) ?? "";
+      const merge = mergeMap.get(key);
+      row.push({
+        text,
+        colSpan: merge?.colSpan ?? 1,
+        rowSpan: merge?.rowSpan ?? 1
+      });
+    }
+    cellRows.push(row);
+  }
+  if (cellRows.length > 0) {
+    const table = buildTable(cellRows);
+    if (table.rows > 0) {
+      blocks.push({ type: "table", table, pageNumber: sheetIndex + 1 });
+    }
+  }
+  return blocks;
+}
+async function parseXlsxDocument(buffer, options) {
+  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE3);
+  const zip = await JSZip3.loadAsync(buffer);
+  const warnings = [];
+  const workbookFile = zip.file("xl/workbook.xml");
+  if (!workbookFile) {
+    throw new KordocError("\uC720\uD6A8\uD558\uC9C0 \uC54A\uC740 XLSX \uD30C\uC77C: xl/workbook.xml\uC774 \uC5C6\uC2B5\uB2C8\uB2E4");
+  }
+  let sharedStrings = [];
+  const ssFile = zip.file("xl/sharedStrings.xml");
+  if (ssFile) {
+    sharedStrings = parseSharedStrings(await ssFile.async("text"));
+  }
+  const sheets = parseWorkbook(await workbookFile.async("text"));
+  if (sheets.length === 0) {
+    throw new KordocError("XLSX \uD30C\uC77C\uC5D0 \uC2DC\uD2B8\uAC00 \uC5C6\uC2B5\uB2C8\uB2E4");
+  }
+  let relsMap = /* @__PURE__ */ new Map();
+  const relsFile = zip.file("xl/_rels/workbook.xml.rels");
+  if (relsFile) {
+    relsMap = parseRels(await relsFile.async("text"));
+  }
+  let pageFilter = null;
+  if (options?.pages) {
+    const { parsePageRange: parsePageRange2 } = await import("./page-range-737B4EZW.js");
+    pageFilter = parsePageRange2(options.pages, sheets.length);
+  }
+  const blocks = [];
+  const processedSheets = Math.min(sheets.length, MAX_SHEETS);
+  for (let i = 0; i < processedSheets; i++) {
+    if (pageFilter && !pageFilter.has(i + 1)) continue;
+    const sheet = sheets[i];
+    options?.onProgress?.(i + 1, processedSheets);
+    let sheetPath = relsMap.get(sheet.rId);
+    if (sheetPath) {
+      if (!sheetPath.startsWith("xl/") && !sheetPath.startsWith("/")) {
+        sheetPath = `xl/${sheetPath}`;
+      } else if (sheetPath.startsWith("/")) {
+        sheetPath = sheetPath.slice(1);
+      }
+    } else {
+      sheetPath = `xl/worksheets/sheet${i + 1}.xml`;
+    }
+    const sheetFile = zip.file(sheetPath);
+    if (!sheetFile) {
+      warnings.push({
+        page: i + 1,
+        message: `\uC2DC\uD2B8 "${sheet.name}" \uD30C\uC77C\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4: ${sheetPath}`,
+        code: "PARTIAL_PARSE"
+      });
+      continue;
+    }
+    try {
+      const sheetXml = await sheetFile.async("text");
+      const { grid, merges, maxRow, maxCol } = parseWorksheet(sheetXml, sharedStrings);
+      const sheetBlocks = sheetToBlocks(sheet.name, grid, merges, maxRow, maxCol, i);
+      blocks.push(...sheetBlocks);
+    } catch (err) {
+      warnings.push({
+        page: i + 1,
+        message: `\uC2DC\uD2B8 "${sheet.name}" \uD30C\uC2F1 \uC2E4\uD328: ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
+        code: "PARTIAL_PARSE"
+      });
+    }
+  }
+  const metadata = {
+    pageCount: processedSheets
+  };
+  const coreFile = zip.file("docProps/core.xml");
+  if (coreFile) {
+    try {
+      const coreXml = await coreFile.async("text");
+      const doc = parseXml(coreXml);
+      const getFirst = (tag) => {
+        const els = doc.getElementsByTagName(tag);
+        return els.length > 0 ? (els[0].textContent ?? "").trim() : void 0;
+      };
+      metadata.title = getFirst("dc:title") || getFirst("dcterms:title");
+      metadata.author = getFirst("dc:creator");
+      metadata.description = getFirst("dc:description");
+      const created = getFirst("dcterms:created");
+      if (created) metadata.createdAt = created;
+      const modified = getFirst("dcterms:modified");
+      if (modified) metadata.modifiedAt = modified;
+    } catch {
+    }
+  }
+  const markdown = blocksToMarkdown(blocks);
+  return { markdown, blocks, metadata, warnings: warnings.length > 0 ? warnings : void 0 };
+}
+
+// src/docx/parser.ts
+import JSZip4 from "jszip";
+import { DOMParser as DOMParser3 } from "@xmldom/xmldom";
+var MAX_DECOMPRESS_SIZE4 = 100 * 1024 * 1024;
+function getChildElements(parent, localName) {
+  const result = [];
+  const children = parent.childNodes;
+  for (let i = 0; i < children.length; i++) {
+    const node = children[i];
+    if (node.nodeType === 1) {
+      const el = node;
+      if (el.localName === localName || el.tagName?.endsWith(`:${localName}`)) {
+        result.push(el);
+      }
+    }
+  }
+  return result;
+}
+function findElements(parent, localName) {
+  const result = [];
+  const walk = (node) => {
+    const children = node.childNodes;
+    for (let i = 0; i < children.length; i++) {
+      const child = children[i];
+      if (child.nodeType === 1) {
+        const el = child;
+        if (el.localName === localName || el.tagName?.endsWith(`:${localName}`)) {
+          result.push(el);
+        }
+        walk(el);
+      }
+    }
+  };
+  walk(parent);
+  return result;
+}
+function getAttr(el, localName) {
+  const attrs = el.attributes;
+  for (let i = 0; i < attrs.length; i++) {
+    const attr = attrs[i];
+    if (attr.localName === localName || attr.name === localName) return attr.value;
+  }
+  return null;
+}
+function parseXml2(text) {
+  return new DOMParser3().parseFromString(text, "text/xml");
+}
+function parseStyles(xml) {
+  const doc = parseXml2(xml);
+  const styles = /* @__PURE__ */ new Map();
+  const styleElements = findElements(doc, "style");
+  for (const el of styleElements) {
+    const styleId = getAttr(el, "styleId");
+    if (!styleId) continue;
+    const nameEls = getChildElements(el, "name");
+    const name = nameEls.length > 0 ? getAttr(nameEls[0], "val") ?? "" : "";
+    const basedOnEls = getChildElements(el, "basedOn");
+    const basedOn = basedOnEls.length > 0 ? getAttr(basedOnEls[0], "val") ?? void 0 : void 0;
+    const pPrEls = getChildElements(el, "pPr");
+    let outlineLevel;
+    if (pPrEls.length > 0) {
+      const outlineEls = getChildElements(pPrEls[0], "outlineLvl");
+      if (outlineEls.length > 0) {
+        const val = getAttr(outlineEls[0], "val");
+        if (val !== null) outlineLevel = parseInt(val, 10);
+      }
+    }
+    if (outlineLevel === void 0) {
+      const headingMatch = name.match(/^(?:heading|Heading)\s*(\d+)$/i);
+      if (headingMatch) outlineLevel = parseInt(headingMatch[1], 10) - 1;
+    }
+    styles.set(styleId, { name, basedOn, outlineLevel });
+  }
+  return styles;
+}
+function parseNumbering(xml) {
+  const doc = parseXml2(xml);
+  const abstractNums = /* @__PURE__ */ new Map();
+  const abstractElements = findElements(doc, "abstractNum");
+  for (const el of abstractElements) {
+    const abstractNumId = getAttr(el, "abstractNumId");
+    if (!abstractNumId) continue;
+    const levels = /* @__PURE__ */ new Map();
+    const lvlElements = getChildElements(el, "lvl");
+    for (const lvl of lvlElements) {
+      const ilvl = parseInt(getAttr(lvl, "ilvl") ?? "0", 10);
+      const numFmtEls = getChildElements(lvl, "numFmt");
+      const numFmt = numFmtEls.length > 0 ? getAttr(numFmtEls[0], "val") ?? "bullet" : "bullet";
+      levels.set(ilvl, { numFmt, level: ilvl });
+    }
+    abstractNums.set(abstractNumId, levels);
+  }
+  const nums = /* @__PURE__ */ new Map();
+  const numElements = findElements(doc, "num");
+  for (const el of numElements) {
+    const numId = getAttr(el, "numId");
+    if (!numId) continue;
+    const abstractRefs = getChildElements(el, "abstractNumId");
+    if (abstractRefs.length > 0) {
+      const ref = getAttr(abstractRefs[0], "val");
+      if (ref && abstractNums.has(ref)) {
+        nums.set(numId, abstractNums.get(ref));
+      }
+    }
+  }
+  return nums;
+}
+function parseRels2(xml) {
+  const doc = parseXml2(xml);
+  const map = /* @__PURE__ */ new Map();
+  const rels = findElements(doc, "Relationship");
+  for (const rel of rels) {
+    const id = getAttr(rel, "Id");
+    const target = getAttr(rel, "Target");
+    if (id && target) map.set(id, target);
+  }
+  return map;
+}
+function parseFootnotes(xml) {
+  const doc = parseXml2(xml);
+  const notes = /* @__PURE__ */ new Map();
+  const fnElements = findElements(doc, "footnote");
+  for (const fn of fnElements) {
+    const id = getAttr(fn, "id");
+    if (!id || id === "0" || id === "-1") continue;
+    const texts = [];
+    const pElements = findElements(fn, "p");
+    for (const p of pElements) {
+      const runs = findElements(p, "r");
+      for (const r of runs) {
+        const tElements = getChildElements(r, "t");
+        for (const t of tElements) texts.push(t.textContent ?? "");
+      }
+    }
+    notes.set(id, texts.join("").trim());
+  }
+  return notes;
+}
+function extractRun(r) {
+  const tElements = getChildElements(r, "t");
+  const text = tElements.map((t) => t.textContent ?? "").join("");
+  let bold = false;
+  let italic = false;
+  const rPrEls = getChildElements(r, "rPr");
+  if (rPrEls.length > 0) {
+    bold = getChildElements(rPrEls[0], "b").length > 0;
+    italic = getChildElements(rPrEls[0], "i").length > 0;
+  }
+  return { text, bold, italic };
+}
+function parseParagraph(p, styles, numbering, footnotes, rels) {
+  const pPrEls = getChildElements(p, "pPr");
+  let styleId = "";
+  let numId = "";
+  let ilvl = 0;
+  if (pPrEls.length > 0) {
+    const pStyleEls = getChildElements(pPrEls[0], "pStyle");
+    if (pStyleEls.length > 0) styleId = getAttr(pStyleEls[0], "val") ?? "";
+    const numPrEls = getChildElements(pPrEls[0], "numPr");
+    if (numPrEls.length > 0) {
+      const numIdEls = getChildElements(numPrEls[0], "numId");
+      const ilvlEls = getChildElements(numPrEls[0], "ilvl");
+      numId = numIdEls.length > 0 ? getAttr(numIdEls[0], "val") ?? "" : "";
+      ilvl = ilvlEls.length > 0 ? parseInt(getAttr(ilvlEls[0], "val") ?? "0", 10) : 0;
+    }
+  }
+  const parts = [];
+  let hasBold = false;
+  let hasItalic = false;
+  let href;
+  let footnoteText;
+  const hyperlinks = getChildElements(p, "hyperlink");
+  const hyperlinkTexts = /* @__PURE__ */ new Set();
+  for (const hl of hyperlinks) {
+    const rId = getAttr(hl, "id");
+    const hlText = [];
+    const runs2 = findElements(hl, "r");
+    for (const r of runs2) {
+      const result = extractRun(r);
+      hlText.push(result.text);
+    }
+    const text2 = hlText.join("");
+    if (text2) {
+      hyperlinkTexts.add(text2);
+      if (rId && rels.has(rId)) {
+        href = rels.get(rId);
+        parts.push(text2);
+      } else {
+        parts.push(text2);
+      }
+    }
+  }
+  const runs = getChildElements(p, "r");
+  for (const r of runs) {
+    if (r.parentNode && r.parentNode.localName === "hyperlink") continue;
+    const result = extractRun(r);
+    if (result.bold) hasBold = true;
+    if (result.italic) hasItalic = true;
+    const fnRefEls = getChildElements(r, "footnoteReference");
+    if (fnRefEls.length > 0) {
+      const fnId = getAttr(fnRefEls[0], "id");
+      if (fnId && footnotes.has(fnId)) {
+        footnoteText = footnotes.get(fnId);
+      }
+    }
+    if (result.text) parts.push(result.text);
+  }
+  const text = parts.join("").trim();
+  if (!text) return null;
+  const style = styles.get(styleId);
+  if (style?.outlineLevel !== void 0 && style.outlineLevel >= 0 && style.outlineLevel <= 5) {
+    return {
+      type: "heading",
+      text,
+      level: style.outlineLevel + 1
+    };
+  }
+  if (numId && numId !== "0") {
+    const numDef = numbering.get(numId);
+    const levelInfo = numDef?.get(ilvl);
+    const listType = levelInfo?.numFmt === "bullet" ? "unordered" : "ordered";
+    return { type: "list", text, listType };
+  }
+  const block = { type: "paragraph", text };
+  if (hasBold || hasItalic) {
+    block.style = { bold: hasBold || void 0, italic: hasItalic || void 0 };
+  }
+  if (href) block.href = href;
+  if (footnoteText) block.footnoteText = footnoteText;
+  return block;
+}
+function parseTable(tbl, styles, numbering, footnotes, rels) {
+  const trElements = getChildElements(tbl, "tr");
+  if (trElements.length === 0) return null;
+  const rows = [];
+  let maxCols = 0;
+  for (const tr of trElements) {
+    const tcElements = getChildElements(tr, "tc");
+    const row = [];
+    for (const tc of tcElements) {
+      let colSpan = 1;
+      let rowSpan = 1;
+      const tcPrEls = getChildElements(tc, "tcPr");
+      if (tcPrEls.length > 0) {
+        const gridSpanEls = getChildElements(tcPrEls[0], "gridSpan");
+        if (gridSpanEls.length > 0) {
+          colSpan = parseInt(getAttr(gridSpanEls[0], "val") ?? "1", 10);
+        }
+        const vMergeEls = getChildElements(tcPrEls[0], "vMerge");
+        if (vMergeEls.length > 0) {
+          const val = getAttr(vMergeEls[0], "val");
+          if (val !== "restart" && val !== null) {
+            row.push({ text: "", colSpan, rowSpan: 0 });
+            continue;
+          }
+        }
+      }
+      const cellTexts = [];
+      const pElements = getChildElements(tc, "p");
+      for (const p of pElements) {
+        const block = parseParagraph(p, styles, numbering, footnotes, rels);
+        if (block?.text) cellTexts.push(block.text);
+      }
+      row.push({ text: cellTexts.join("\n"), colSpan, rowSpan });
+    }
+    rows.push(row);
+    if (row.length > maxCols) maxCols = row.length;
+  }
+  for (let c = 0; c < maxCols; c++) {
+    for (let r = 0; r < rows.length; r++) {
+      const cell = rows[r][c];
+      if (!cell || cell.rowSpan === 0) continue;
+      let span = 1;
+      for (let nr = r + 1; nr < rows.length; nr++) {
+        if (rows[nr][c]?.rowSpan === 0) span++;
+        else break;
+      }
+      cell.rowSpan = span;
+    }
+  }
+  const cleanRows = [];
+  for (const row of rows) {
+    const clean = row.filter((cell) => cell.rowSpan !== 0);
+    cleanRows.push(clean);
+  }
+  if (cleanRows.length === 0) return null;
+  let cols = 0;
+  for (const row of cleanRows) {
+    let c = 0;
+    for (const cell of row) c += cell.colSpan;
+    if (c > cols) cols = c;
+  }
+  const table = {
+    rows: cleanRows.length,
+    cols,
+    cells: cleanRows,
+    hasHeader: cleanRows.length > 1
+  };
+  return { type: "table", table };
+}
+async function extractImages(zip, rels, doc) {
+  const blocks = [];
+  const images = [];
+  const drawingElements = findElements(doc.documentElement, "drawing");
+  let imgIdx = 0;
+  for (const drawing of drawingElements) {
+    const blips = findElements(drawing, "blip");
+    for (const blip of blips) {
+      const embedId = getAttr(blip, "embed");
+      if (!embedId) continue;
+      const target = rels.get(embedId);
+      if (!target) continue;
+      const imgPath = target.startsWith("/") ? target.slice(1) : target.startsWith("word/") ? target : `word/${target}`;
+      const imgFile = zip.file(imgPath);
+      if (!imgFile) continue;
+      try {
+        const data = await imgFile.async("uint8array");
+        imgIdx++;
+        const ext = imgPath.split(".").pop()?.toLowerCase() ?? "png";
+        const mimeMap = {
+          png: "image/png",
+          jpg: "image/jpeg",
+          jpeg: "image/jpeg",
+          gif: "image/gif",
+          bmp: "image/bmp",
+          wmf: "image/wmf",
+          emf: "image/emf"
+        };
+        const filename = `image_${String(imgIdx).padStart(3, "0")}.${ext}`;
+        images.push({ filename, data, mimeType: mimeMap[ext] ?? "image/png" });
+        blocks.push({ type: "image", text: filename });
+      } catch {
+      }
+    }
+  }
+  return { blocks, images };
+}
+async function parseDocxDocument(buffer, options) {
+  precheckZipSize(buffer, MAX_DECOMPRESS_SIZE4);
+  const zip = await JSZip4.loadAsync(buffer);
+  const warnings = [];
+  const docFile = zip.file("word/document.xml");
+  if (!docFile) {
+    throw new KordocError("\uC720\uD6A8\uD558\uC9C0 \uC54A\uC740 DOCX \uD30C\uC77C: word/document.xml\uC774 \uC5C6\uC2B5\uB2C8\uB2E4");
+  }
+  let rels = /* @__PURE__ */ new Map();
+  const relsFile = zip.file("word/_rels/document.xml.rels");
+  if (relsFile) {
+    rels = parseRels2(await relsFile.async("text"));
+  }
+  let styles = /* @__PURE__ */ new Map();
+  const stylesFile = zip.file("word/styles.xml");
+  if (stylesFile) {
+    try {
+      styles = parseStyles(await stylesFile.async("text"));
+    } catch {
+    }
+  }
+  let numbering = /* @__PURE__ */ new Map();
+  const numFile = zip.file("word/numbering.xml");
+  if (numFile) {
+    try {
+      numbering = parseNumbering(await numFile.async("text"));
+    } catch {
+    }
+  }
+  let footnotes = /* @__PURE__ */ new Map();
+  const fnFile = zip.file("word/footnotes.xml");
+  if (fnFile) {
+    try {
+      footnotes = parseFootnotes(await fnFile.async("text"));
+    } catch {
+    }
+  }
+  const docXml = await docFile.async("text");
+  const doc = parseXml2(docXml);
+  const body = findElements(doc, "body");
+  if (body.length === 0) {
+    throw new KordocError("DOCX \uBCF8\uBB38(w:body)\uC744 \uCC3E\uC744 \uC218 \uC5C6\uC2B5\uB2C8\uB2E4");
+  }
+  const blocks = [];
+  const bodyEl = body[0];
+  const children = bodyEl.childNodes;
+  for (let i = 0; i < children.length; i++) {
+    const node = children[i];
+    if (node.nodeType !== 1) continue;
+    const el = node;
+    const localName = el.localName ?? el.tagName?.split(":").pop();
+    if (localName === "p") {
+      const block = parseParagraph(el, styles, numbering, footnotes, rels);
+      if (block) blocks.push(block);
+    } else if (localName === "tbl") {
+      const block = parseTable(el, styles, numbering, footnotes, rels);
+      if (block) blocks.push(block);
+    }
+  }
+  const { blocks: imgBlocks, images } = await extractImages(zip, rels, doc);
+  const metadata = {};
+  const coreFile = zip.file("docProps/core.xml");
+  if (coreFile) {
+    try {
+      const coreXml = await coreFile.async("text");
+      const coreDoc = parseXml2(coreXml);
+      const getFirst = (tag) => {
+        const els = coreDoc.getElementsByTagName(tag);
+        return els.length > 0 ? (els[0].textContent ?? "").trim() : void 0;
+      };
+      metadata.title = getFirst("dc:title") || getFirst("dcterms:title");
+      metadata.author = getFirst("dc:creator");
+      metadata.description = getFirst("dc:description");
+      const created = getFirst("dcterms:created");
+      if (created) metadata.createdAt = created;
+      const modified = getFirst("dcterms:modified");
+      if (modified) metadata.modifiedAt = modified;
+    } catch {
+    }
+  }
+  const outline = blocks.filter((b) => b.type === "heading").map((b) => ({ level: b.level ?? 2, text: b.text ?? "" }));
+  const markdown = blocksToMarkdown(blocks);
+  return {
+    markdown,
+    blocks,
+    metadata,
+    outline: outline.length > 0 ? outline : void 0,
+    warnings: warnings.length > 0 ? warnings : void 0,
+    images: images.length > 0 ? images : void 0
+  };
+}
+
 // src/form/recognize.ts
 var LABEL_KEYWORDS = /* @__PURE__ */ new Set([
   "\uC131\uBA85",
@@ -3129,7 +4042,7 @@ function extractInlineFields(text) {
 }
 
 // src/hwpx/generator.ts
-import JSZip2 from "jszip";
+import JSZip5 from "jszip";
 
 // src/index.ts
 async function parse(input, options) {
@@ -3152,8 +4065,12 @@ async function parse(input, options) {
   }
   const format = detectFormat(buffer);
   switch (format) {
-    case "hwpx":
+    case "hwpx": {
+      const zipFormat = await detectZipFormat(buffer);
+      if (zipFormat === "xlsx") return parseXlsx(buffer, options);
+      if (zipFormat === "docx") return parseDocx(buffer, options);
       return parseHwpx(buffer, options);
+    }
     case "hwp":
       return parseHwp(buffer, options);
     case "pdf":
@@ -3180,9 +4097,27 @@ async function parseHwp(buffer, options) {
 }
 async function parsePdf(buffer, options) {
   try {
-    return await parsePdfDocument(buffer, options);
+    const { markdown, blocks, metadata, outline, warnings, isImageBased } = await parsePdfDocument(buffer, options);
+    return { success: true, fileType: "pdf", markdown, blocks, metadata, outline, warnings, isImageBased };
+  } catch (err) {
+    const isImageBased = err instanceof Error && "isImageBased" in err ? true : void 0;
+    return { success: false, fileType: "pdf", error: err instanceof Error ? err.message : "PDF \uD30C\uC2F1 \uC2E4\uD328", code: classifyError(err), isImageBased };
+  }
+}
+async function parseXlsx(buffer, options) {
+  try {
+    const { markdown, blocks, metadata, warnings } = await parseXlsxDocument(buffer, options);
+    return { success: true, fileType: "xlsx", markdown, blocks, metadata, warnings };
   } catch (err) {
-    return { success: false, fileType: "pdf", error: err instanceof Error ? err.message : "PDF \uD30C\uC2F1 \uC2E4\uD328", code: classifyError(err) };
+    return { success: false, fileType: "xlsx", error: err instanceof Error ? err.message : "XLSX \uD30C\uC2F1 \uC2E4\uD328", code: classifyError(err) };
+  }
+}
+async function parseDocx(buffer, options) {
+  try {
+    const { markdown, blocks, metadata, outline, warnings, images } = await parseDocxDocument(buffer, options);
+    return { success: true, fileType: "docx", markdown, blocks, metadata, outline, warnings, images: images?.length ? images : void 0 };
+  } catch (err) {
+    return { success: false, fileType: "docx", error: err instanceof Error ? err.message : "DOCX \uD30C\uC2F1 \uC2E4\uD328", code: classifyError(err) };
   }
 }
 
@@ -3360,10 +4295,6 @@ function diffTableCells(a, b) {
 export {
   detectFormat,
   blocksToMarkdown,
-  VERSION,
-  toArrayBuffer,
-  KordocError,
-  sanitizeError,
   extractHwpxMetadataOnly,
   extractHwp5MetadataOnly,
   extractPdfMetadataOnly,
@@ -3371,4 +4302,4 @@ export {
   extractFormFields,
   parse
 };
-//# sourceMappingURL=chunk-NJ3R7LNR.js.map
+//# sourceMappingURL=chunk-MDRW3HYC.js.map