@clazic/kordoc 2.1.6 → 2.2.1

package/dist/mcp.js

@@ -7,22 +7,25 @@ import {
   extractHwp5MetadataOnly,
   extractHwpxMetadataOnly,
   extractPdfMetadataOnly,
+  markdownToHwpx,
+  markdownToXlsx,
   parse
-} from "./chunk-H7HMKSLX.js";
+} from "./chunk-3NF22UFF.js";
 import {
   KordocError,
   VERSION,
   sanitizeError,
   toArrayBuffer
-} from "./chunk-TFYOEQE2.js";
+} from "./chunk-7MXQWWUW.js";
 import "./chunk-MOL7MDBG.js";
+import "./chunk-ZWE3DS7E.js";
 
 // src/mcp.ts
 import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
 import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
 import { z } from "zod";
-import { readFileSync, realpathSync, openSync, readSync, closeSync, statSync } from "fs";
-import { resolve, isAbsolute, extname } from "path";
+import { readFileSync, realpathSync, openSync, readSync, closeSync, statSync, mkdirSync, writeFileSync, readdirSync, existsSync } from "fs";
+import { resolve, isAbsolute, extname, basename } from "path";
 var ALLOWED_EXTENSIONS = /* @__PURE__ */ new Set([".hwp", ".hwpx", ".pdf", ".xlsx", ".docx"]);
 var MAX_FILE_SIZE = 500 * 1024 * 1024;
 function safePath(filePath) {
@@ -62,11 +65,12 @@ server.tool(
   "parse_document",
   "\uD55C\uAD6D \uBB38\uC11C \uD30C\uC77C(HWP, HWPX, PDF, XLSX, DOCX)\uC744 \uB9C8\uD06C\uB2E4\uC6B4\uC73C\uB85C \uBCC0\uD658\uD569\uB2C8\uB2E4. \uD30C\uC77C \uACBD\uB85C\uB97C \uC785\uB825\uD558\uBA74 \uD3EC\uB9F7\uC744 \uC790\uB3D9 \uAC10\uC9C0\uD558\uC5EC \uD14D\uC2A4\uD2B8\uB97C \uCD94\uCD9C\uD569\uB2C8\uB2E4.",
   {
-    file_path: z.string().min(1).describe("\uD30C\uC2F1\uD560 \uBB38\uC11C \uD30C\uC77C\uC758 \uC808\uB300 \uACBD\uB85C (HWP, HWPX, PDF, XLSX, DOCX)")
+    file_path: z.string().min(1).describe("\uD30C\uC2F1\uD560 \uBB38\uC11C \uD30C\uC77C\uC758 \uC808\uB300 \uACBD\uB85C (HWP, HWPX, PDF, XLSX, DOCX)"),
+    image_dir: z.string().optional().describe("\uC774\uBBF8\uC9C0 \uC800\uC7A5 \uD3F4\uB354 \uACBD\uB85C (\uAE30\uBCF8: \uD30C\uC77C\uBA85\uACFC \uAC19\uC740 \uC774\uB984\uC758 \uD3F4\uB354)")
   },
-  async ({ file_path }) => {
+  async ({ file_path, image_dir }) => {
     try {
-      const { buffer } = readValidatedFile(file_path);
+      const { buffer, resolved: resolvedFilePath } = readValidatedFile(file_path);
       const format = detectFormat(buffer);
       if (format === "unknown") {
         return {
@@ -100,6 +104,20 @@ ${outlineText}`);
         parts.push(`
 \u26A0\uFE0F \uACBD\uACE0:
 ${warnText}`);
+      }
+      const savedImages = [];
+      if (result.images?.length) {
+        const defaultDir = resolve(resolvedFilePath, "..", basename(resolvedFilePath).replace(/\.[^.]+$/, ""));
+        const imgDir = image_dir ? resolve(image_dir) : defaultDir;
+        mkdirSync(imgDir, { recursive: true });
+        for (const img of result.images) {
+          const imgPath = resolve(imgDir, img.filename);
+          writeFileSync(imgPath, img.data);
+          savedImages.push(imgPath);
+        }
+        parts.push(`
+\u{1F4BE} \uC800\uC7A5\uB41C \uC774\uBBF8\uC9C0 (${savedImages.length}\uAC1C):
+${savedImages.map((p) => `  ${p}`).join("\n")}`);
       }
       parts.push(`
 
@@ -115,6 +133,81 @@ ${result.markdown}`);
     }
   }
 );
+server.tool(
+  "convert_document",
+  "\uB9C8\uD06C\uB2E4\uC6B4 \uD14D\uC2A4\uD2B8\uB97C HWPX \uB610\uB294 XLSX \uD30C\uC77C\uB85C \uBCC0\uD658\uD558\uC5EC \uC800\uC7A5\uD569\uB2C8\uB2E4.",
+  {
+    markdown: z.string().min(1).describe("\uBCC0\uD658\uD560 \uB9C8\uD06C\uB2E4\uC6B4 \uD14D\uC2A4\uD2B8"),
+    output_path: z.string().min(1).describe("\uC800\uC7A5\uD560 \uD30C\uC77C\uC758 \uC808\uB300 \uACBD\uB85C (.hwpx \uB610\uB294 .xlsx)"),
+    format: z.enum(["hwpx", "xlsx"]).optional().default("hwpx").describe("\uCD9C\uB825 \uD3EC\uB9F7 (\uAE30\uBCF8: hwpx)"),
+    image_dir: z.string().optional().describe("\uC774\uBBF8\uC9C0 \uD3F4\uB354 \uACBD\uB85C (\uAE30\uBCF8: output_path \uD30C\uC77C\uBA85 \uD3F4\uB354)"),
+    template_path: z.string().optional().describe("HWPX \uD15C\uD50C\uB9BF \uD30C\uC77C \uACBD\uB85C (hwpx \uC804\uC6A9)")
+  },
+  async ({ markdown, output_path, format, image_dir, template_path }) => {
+    try {
+      const resolvedOutput = resolve(output_path);
+      const outputExt = extname(resolvedOutput).toLowerCase();
+      if (outputExt !== ".hwpx" && outputExt !== ".xlsx") {
+        return {
+          content: [{ type: "text", text: `\uC624\uB958: \uCD9C\uB825 \uD30C\uC77C \uD655\uC7A5\uC790\uB294 .hwpx \uB610\uB294 .xlsx\uB9CC \uD5C8\uC6A9\uB429\uB2C8\uB2E4: ${outputExt || "(\uC5C6\uC74C)"}` }],
+          isError: true
+        };
+      }
+      const warnings = [];
+      const stem = basename(resolvedOutput).replace(/\.[^.]+$/, "");
+      const defaultImgDir = resolve(resolvedOutput, "..", stem);
+      const imgDir = image_dir ? resolve(image_dir) : defaultImgDir;
+      const images = [];
+      if (existsSync(imgDir)) {
+        const mimeMap = {
+          png: "image/png",
+          jpg: "image/jpeg",
+          jpeg: "image/jpeg",
+          gif: "image/gif",
+          bmp: "image/bmp"
+        };
+        for (const entry of readdirSync(imgDir, { withFileTypes: true })) {
+          if (!entry.isFile()) continue;
+          const fname = entry.name;
+          const ext = extname(fname).slice(1).toLowerCase();
+          if (!mimeMap[ext]) continue;
+          const data = readFileSync(resolve(imgDir, fname));
+          images.push({ filename: fname, data: new Uint8Array(data), mimeType: mimeMap[ext] });
+        }
+      }
+      let buf;
+      if (format === "xlsx") {
+        if (template_path) warnings.push("[warn] --template\uC740 hwpx \uC804\uC6A9\uC785\uB2C8\uB2E4. \uBB34\uC2DC\uB429\uB2C8\uB2E4.");
+        buf = await markdownToXlsx(markdown, { warnings, images: images.length ? images : void 0 });
+      } else {
+        let templateArrayBuffer;
+        if (template_path) {
+          const tmpl = readFileSync(safePath(template_path));
+          templateArrayBuffer = tmpl.buffer.slice(tmpl.byteOffset, tmpl.byteOffset + tmpl.byteLength);
+        }
+        buf = await markdownToHwpx(markdown, {
+          warnings,
+          images: images.length ? images : void 0,
+          templateArrayBuffer
+        });
+      }
+      writeFileSync(resolvedOutput, Buffer.from(buf));
+      const parts = [
+        `\u2705 \uBCC0\uD658 \uC644\uB8CC: ${resolvedOutput}`,
+        `\uD3EC\uB9F7: ${format.toUpperCase()}, \uD06C\uAE30: ${(buf.byteLength / 1024).toFixed(1)}KB`
+      ];
+      if (images.length) parts.push(`\uD3EC\uD568\uB41C \uC774\uBBF8\uC9C0: ${images.length}\uAC1C (${imgDir})`);
+      if (warnings.length) parts.push(`\uACBD\uACE0:
+${warnings.map((w) => `  ${w}`).join("\n")}`);
+      return { content: [{ type: "text", text: parts.join("\n") }] };
+    } catch (err) {
+      return {
+        content: [{ type: "text", text: `\uC624\uB958: ${sanitizeError(err)}` }],
+        isError: true
+      };
+    }
+  }
+);
 server.tool(
   "detect_format",
   "\uD30C\uC77C\uC758 \uD3EC\uB9F7\uC744 \uB9E4\uC9C1 \uBC14\uC774\uD2B8\uB85C \uAC10\uC9C0\uD569\uB2C8\uB2E4 (hwpx, hwp, pdf, unknown).",

package/dist/mcp.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/mcp.ts"],"sourcesContent":["/** kordoc MCP 서버 — Claude/Cursor에서 문서 파싱 도구로 사용 */\n\nimport { McpServer } from \"@modelcontextprotocol/sdk/server/mcp.js\"\nimport { StdioServerTransport } from \"@modelcontextprotocol/sdk/server/stdio.js\"\nimport { z } from \"zod\"\nimport { readFileSync, realpathSync, openSync, readSync, closeSync, statSync } from \"fs\"\nimport { resolve, isAbsolute, extname } from \"path\"\nimport { parse, detectFormat, blocksToMarkdown, compare, extractFormFields } from \"./index.js\"\nimport { VERSION, toArrayBuffer, sanitizeError, KordocError } from \"./utils.js\"\nimport { extractHwp5MetadataOnly } from \"./hwp5/parser.js\"\nimport { extractHwpxMetadataOnly } from \"./hwpx/parser.js\"\nimport { extractPdfMetadataOnly } from \"./pdf/parser.js\"\n\n/** 허용 파일 확장자 */\nconst ALLOWED_EXTENSIONS = new Set([\".hwp\", \".hwpx\", \".pdf\", \".xlsx\", \".docx\"])\n/** 최대 파일 크기 (500MB) */\nconst MAX_FILE_SIZE = 500 * 1024 * 1024\n\n/** 경로 정규화 및 보안 검증 */\nfunction safePath(filePath: string): string {\n  if (!filePath) throw new KordocError(\"파일 경로가 비어있습니다\")\n  const resolved = resolve(filePath)\n  const real = realpathSync(resolved)\n  if (!isAbsolute(real)) throw new KordocError(\"절대 경로만 허용됩니다\")\n  const ext = extname(real).toLowerCase()\n  if (!ALLOWED_EXTENSIONS.has(ext)) throw new KordocError(`지원하지 않는 확장자입니다: ${ext} (허용: ${[...ALLOWED_EXTENSIONS].join(\", \")})`)\n  return real\n}\n\n/** 최대 파일 크기 — metadata 전용 (50MB, 전체 파싱보다 보수적) */\nconst MAX_METADATA_FILE_SIZE = 50 * 1024 * 1024\n\n/** 파일 읽기 + 크기 검증 공통 로직 */\nfunction readValidatedFile(filePath: string, maxSize = MAX_FILE_SIZE): { buffer: ArrayBuffer; resolved: string } {\n  const resolved = safePath(filePath)\n  const fileSize = statSync(resolved).size\n  if (fileSize > maxSize) {\n    throw new KordocError(`파일이 너무 큽니다: ${(fileSize / 1024 / 1024).toFixed(1)}MB (최대 ${maxSize / 1024 / 1024}MB)`)\n  }\n  const raw = readFileSync(resolved)\n  return { buffer: toArrayBuffer(raw), resolved }\n}\n\n/** 파일 헤더(16바이트)만 읽어 포맷 감지 — 전체 파일 로드 불필요 */\nfunction detectFormatFromHeader(resolved: string): ReturnType<typeof detectFormat> {\n  const fd = openSync(resolved, \"r\")\n  try {\n    const headerBuf = Buffer.alloc(16)\n    readSync(fd, headerBuf, 0, 16, 0)\n    return detectFormat(toArrayBuffer(headerBuf))\n  } finally {\n    closeSync(fd)\n  }\n}\n\nconst server = new McpServer({\n  name: \"kordoc\",\n  version: VERSION,\n})\n\n// ─── 도구: parse_document ────────────────────────────\n\nserver.tool(\n  \"parse_document\",\n  \"한국 문서 파일(HWP, HWPX, PDF, XLSX, DOCX)을 마크다운으로 변환합니다. 파일 경로를 입력하면 포맷을 자동 감지하여 텍스트를 추출합니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로 (HWP, HWPX, PDF, XLSX, DOCX)\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const meta = [\n        `포맷: ${result.fileType.toUpperCase()}`,\n        result.pageCount ? `페이지: ${result.pageCount}` : null,\n        result.metadata?.title ? `제목: ${result.metadata.title}` : null,\n        result.metadata?.author ? `작성자: ${result.metadata.author}` : null,\n        result.isImageBased ? \"이미지 기반 PDF (텍스트 추출 불가)\" : null,\n      ].filter(Boolean).join(\" | \")\n\n      // outline/warnings 부가 정보 추가\n      const parts: string[] = [`[${meta}]`]\n\n      if (result.outline && result.outline.length > 0) {\n        const outlineText = result.outline.map(o => `${\"  \".repeat(o.level - 1)}- ${o.text}`).join(\"\\n\")\n        parts.push(`\\n📑 문서 구조:\\n${outlineText}`)\n      }\n\n      if (result.warnings && result.warnings.length > 0) {\n        const warnText = result.warnings.map(w => `- [p${w.page || \"?\"}] ${w.message}`).join(\"\\n\")\n        parts.push(`\\n⚠️ 경고:\\n${warnText}`)\n      }\n\n      parts.push(`\\n\\n${result.markdown}`)\n\n      return {\n        content: [{ type: \"text\", text: parts.join(\"\") }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: detect_format ─────────────────────────────\n\nserver.tool(\n  \"detect_format\",\n  \"파일의 포맷을 매직 바이트로 감지합니다 (hwpx, hwp, pdf, unknown).\",\n  {\n    file_path: z.string().min(1).describe(\"감지할 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const resolved = safePath(file_path)\n      const format = detectFormatFromHeader(resolved)\n      return {\n        content: [{ type: \"text\", text: `${file_path}: ${format}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_metadata ────────────────────────────\n\nserver.tool(\n  \"parse_metadata\",\n  \"문서의 메타데이터(제목, 작성자, 날짜 등)만 빠르게 추출합니다. 전체 파싱 없이 헤더/매니페스트만 읽습니다.\",\n  {\n    file_path: z.string().min(1).describe(\"메타데이터를 추출할 문서 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const resolved = safePath(file_path)\n      const format = detectFormatFromHeader(resolved)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      // metadata 전용 크기 제한 (50MB)\n      const { buffer } = readValidatedFile(file_path, MAX_METADATA_FILE_SIZE)\n\n      let metadata\n      switch (format) {\n        case \"hwp\":\n          metadata = extractHwp5MetadataOnly(Buffer.from(buffer))\n          break\n        case \"hwpx\":\n          metadata = await extractHwpxMetadataOnly(buffer)\n          break\n        case \"pdf\":\n          metadata = await extractPdfMetadataOnly(buffer)\n          break\n      }\n\n      return {\n        content: [{ type: \"text\", text: JSON.stringify({ format, ...metadata }, null, 2) }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_pages ──────────────────────────────\n\nserver.tool(\n  \"parse_pages\",\n  \"문서의 특정 페이지/섹션 범위만 파싱합니다. PDF는 정확한 페이지, HWP/HWPX는 섹션 단위 근사치입니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로\"),\n    pages: z.string().min(1).describe(\"페이지 범위 (예: '1-3', '1,3,5-7')\"),\n  },\n  async ({ file_path, pages }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer, { pages })\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const meta = [\n        `포맷: ${result.fileType.toUpperCase()}`,\n        `범위: ${pages}`,\n        result.pageCount ? `페이지: ${result.pageCount}` : null,\n      ].filter(Boolean).join(\" | \")\n\n      return {\n        content: [{ type: \"text\", text: `[${meta}]\\n\\n${result.markdown}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_table ──────────────────────────────\n\nserver.tool(\n  \"parse_table\",\n  \"문서에서 N번째 테이블만 추출합니다 (0-based index). 테이블이 없거나 인덱스 범위를 초과하면 오류를 반환합니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로\"),\n    table_index: z.number().int().min(0).describe(\"추출할 테이블 인덱스 (0부터 시작)\"),\n  },\n  async ({ file_path, table_index }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const tableBlocks = result.blocks.filter(b => b.type === \"table\" && b.table)\n      if (tableBlocks.length === 0) {\n        return {\n          content: [{ type: \"text\", text: `문서에 테이블이 없습니다.` }],\n          isError: true,\n        }\n      }\n\n      if (table_index >= tableBlocks.length) {\n        return {\n          content: [{ type: \"text\", text: `테이블 인덱스 초과: ${table_index} (총 ${tableBlocks.length}개 테이블)` }],\n          isError: true,\n        }\n      }\n\n      const tableBlock = tableBlocks[table_index]\n      const tableMarkdown = blocksToMarkdown([tableBlock])\n\n      return {\n        content: [{ type: \"text\", text: `[테이블 #${table_index} / 총 ${tableBlocks.length}개]\\n\\n${tableMarkdown}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: compare_documents ─────────────────────────\n\nserver.tool(\n  \"compare_documents\",\n  \"두 한국 문서 파일을 비교하여 추가/삭제/변경된 블록을 표시합니다. 신구대조표 생성에 활용됩니다. 크로스 포맷(HWP↔HWPX) 비교 가능.\",\n  {\n    file_path_a: z.string().min(1).describe(\"비교 원본 문서의 절대 경로\"),\n    file_path_b: z.string().min(1).describe(\"비교 대상 문서의 절대 경로\"),\n  },\n  async ({ file_path_a, file_path_b }) => {\n    try {\n      const { buffer: bufA } = readValidatedFile(file_path_a)\n      const { buffer: bufB } = readValidatedFile(file_path_b)\n\n      const result = await compare(bufA, bufB)\n      const { stats, diffs } = result\n\n      const lines: string[] = [\n        `## 문서 비교 결과`,\n        `추가: ${stats.added} | 삭제: ${stats.removed} | 변경: ${stats.modified} | 동일: ${stats.unchanged}`,\n        \"\",\n      ]\n\n      for (const d of diffs) {\n        const prefix = d.type === \"added\" ? \"+\" : d.type === \"removed\" ? \"-\" : d.type === \"modified\" ? \"~\" : \" \"\n        const text = d.after?.text || d.before?.text || (d.after?.table ? \"[테이블]\" : d.before?.table ? \"[테이블]\" : \"\")\n        const sim = d.similarity !== undefined ? ` (${(d.similarity * 100).toFixed(0)}%)` : \"\"\n        lines.push(`${prefix} ${text.substring(0, 200)}${sim}`)\n      }\n\n      return {\n        content: [{ type: \"text\", text: lines.join(\"\\n\") }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_form ───────────────────────────────\n\nserver.tool(\n  \"parse_form\",\n  \"한국 서식 문서에서 레이블-값 쌍을 구조화된 JSON으로 추출합니다. 양식/서식 문서에 최적화.\",\n  {\n    file_path: z.string().min(1).describe(\"서식 문서 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패: ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const form = extractFormFields(result.blocks)\n      return {\n        content: [{ type: \"text\", text: JSON.stringify(form, null, 2) }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 서버 시작 ───────────────────────────────────────\n\nasync function main() {\n  const transport = new StdioServerTransport()\n  await server.connect(transport)\n}\n\nmain().catch((err) => { console.error(err); process.exit(1) })\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAEA,SAAS,iBAAiB;AAC1B,SAAS,4BAA4B;AACrC,SAAS,SAAS;AAClB,SAAS,cAAc,cAAc,UAAU,UAAU,WAAW,gBAAgB;AACpF,SAAS,SAAS,YAAY,eAAe;AAQ7C,IAAM,qBAAqB,oBAAI,IAAI,CAAC,QAAQ,SAAS,QAAQ,SAAS,OAAO,CAAC;AAE9E,IAAM,gBAAgB,MAAM,OAAO;AAGnC,SAAS,SAAS,UAA0B;AAC1C,MAAI,CAAC,SAAU,OAAM,IAAI,YAAY,sEAAe;AACpD,QAAM,WAAW,QAAQ,QAAQ;AACjC,QAAM,OAAO,aAAa,QAAQ;AAClC,MAAI,CAAC,WAAW,IAAI,EAAG,OAAM,IAAI,YAAY,gEAAc;AAC3D,QAAM,MAAM,QAAQ,IAAI,EAAE,YAAY;AACtC,MAAI,CAAC,mBAAmB,IAAI,GAAG,EAAG,OAAM,IAAI,YAAY,+EAAmB,GAAG,mBAAS,CAAC,GAAG,kBAAkB,EAAE,KAAK,IAAI,CAAC,GAAG;AAC5H,SAAO;AACT;AAGA,IAAM,yBAAyB,KAAK,OAAO;AAG3C,SAAS,kBAAkB,UAAkB,UAAU,eAA0D;AAC/G,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,WAAW,SAAS,QAAQ,EAAE;AACpC,MAAI,WAAW,SAAS;AACtB,UAAM,IAAI,YAAY,wDAAgB,WAAW,OAAO,MAAM,QAAQ,CAAC,CAAC,oBAAU,UAAU,OAAO,IAAI,KAAK;AAAA,EAC9G;AACA,QAAM,MAAM,aAAa,QAAQ;AACjC,SAAO,EAAE,QAAQ,cAAc,GAAG,GAAG,SAAS;AAChD;AAGA,SAAS,uBAAuB,UAAmD;AACjF,QAAM,KAAK,SAAS,UAAU,GAAG;AACjC,MAAI;AACF,UAAM,YAAY,OAAO,MAAM,EAAE;AACjC,aAAS,IAAI,WAAW,GAAG,IAAI,CAAC;AAChC,WAAO,aAAa,cAAc,SAAS,CAAC;AAAA,EAC9C,UAAE;AACA,cAAU,EAAE;AAAA,EACd;AACF;AAEA,IAAM,SAAS,IAAI,UAAU;AAAA,EAC3B,MAAM;AAAA,EACN,SAAS;AACX,CAAC;AAID,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,2GAA+C;AAAA,EACvF;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO;AAAA,QACX,iBAAO,OAAO,SAAS,YAAY,CAAC;AAAA,QACpC,OAAO,YAAY,uBAAQ,OAAO,SAAS,KAAK;AAAA,QAChD,OAAO,UAAU,QAAQ,iBAAO,OAAO,SAAS,KAAK,KAAK;AAAA,QAC1D,OAAO,UAAU,SAAS,uBAAQ,OAAO,SAAS,MAAM,KAAK;AAAA,QAC7D,OAAO,eAAe,uFAA2B;AAAA,MACnD,EAAE,OAAO,OAAO,EAAE,KAAK,KAAK;AAG5B,YAAM,QAAkB,CAAC,IAAI,IAAI,GAAG;AAEpC,UAAI,OAAO,WAAW,OAAO,QAAQ,SAAS,GAAG;AAC/C,cAAM,cAAc,OAAO,QAAQ,IAAI,OAAK,GAAG,KAAK,OAAO,EAAE,QAAQ,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,KAAK,IAAI;AAC/F,cAAM,KAAK;AAAA;AAAA,EAAgB,WAAW,EAAE;AAAA,MAC1C;AAEA,UAAI,OAAO,YAAY,OAAO,SAAS,SAAS,GAAG;AACjD,cAAM,WAAW,OAAO,SAAS,IAAI,OAAK,OAAO,EAAE,QAAQ,GAAG,KAAK,EAAE,OAAO,EAAE,EAAE,KAAK,IAAI;AACzF,cAAM,KAAK;AAAA;AAAA,EAAa,QAAQ,EAAE;AAAA,MACpC;AAEA,YAAM,KAAK;AAAA;AAAA,EAAO,OAAO,QAAQ,EAAE;AAEnC,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,MAAM,KAAK,EAAE,EAAE,CAAC;AAAA,MAClD;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,iEAAe;AAAA,EACvD;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,WAAW,SAAS,SAAS;AACnC,YAAM,SAAS,uBAAuB,QAAQ;AAC9C,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,GAAG,SAAS,KAAK,MAAM,GAAG,CAAC;AAAA,MAC7D;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,mHAAyB;AAAA,EACjE;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,WAAW,SAAS,SAAS;AACnC,YAAM,SAAS,uBAAuB,QAAQ;AAE9C,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAGA,YAAM,EAAE,OAAO,IAAI,kBAAkB,WAAW,sBAAsB;AAEtE,UAAI;AACJ,cAAQ,QAAQ;AAAA,QACd,KAAK;AACH,qBAAW,wBAAwB,OAAO,KAAK,MAAM,CAAC;AACtD;AAAA,QACF,KAAK;AACH,qBAAW,MAAM,wBAAwB,MAAM;AAC/C;AAAA,QACF,KAAK;AACH,qBAAW,MAAM,uBAAuB,MAAM;AAC9C;AAAA,MACJ;AAEA,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,KAAK,UAAU,EAAE,QAAQ,GAAG,SAAS,GAAG,MAAM,CAAC,EAAE,CAAC;AAAA,MACpF;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,8EAAkB;AAAA,IACxD,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,4DAA8B;AAAA,EAClE;AAAA,EACA,OAAO,EAAE,WAAW,MAAM,MAAM;AAC9B,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,QAAQ,EAAE,MAAM,CAAC;AAE5C,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO;AAAA,QACX,iBAAO,OAAO,SAAS,YAAY,CAAC;AAAA,QACpC,iBAAO,KAAK;AAAA,QACZ,OAAO,YAAY,uBAAQ,OAAO,SAAS,KAAK;AAAA,MAClD,EAAE,OAAO,OAAO,EAAE,KAAK,KAAK;AAE5B,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,IAAI,IAAI;AAAA;AAAA,EAAQ,OAAO,QAAQ,GAAG,CAAC;AAAA,MACrE;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,8EAAkB;AAAA,IACxD,aAAa,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,CAAC,EAAE,SAAS,uFAAsB;AAAA,EACtE;AAAA,EACA,OAAO,EAAE,WAAW,YAAY,MAAM;AACpC,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,cAAc,OAAO,OAAO,OAAO,OAAK,EAAE,SAAS,WAAW,EAAE,KAAK;AAC3E,UAAI,YAAY,WAAW,GAAG;AAC5B,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,wEAAiB,CAAC;AAAA,UAClD,SAAS;AAAA,QACX;AAAA,MACF;AAEA,UAAI,eAAe,YAAY,QAAQ;AACrC,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,uDAAe,WAAW,YAAO,YAAY,MAAM,6BAAS,CAAC;AAAA,UAC7F,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,aAAa,YAAY,WAAW;AAC1C,YAAM,gBAAgB,iBAAiB,CAAC,UAAU,CAAC;AAEnD,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,wBAAS,WAAW,aAAQ,YAAY,MAAM;AAAA;AAAA,EAAS,aAAa,GAAG,CAAC;AAAA,MAC1G;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,aAAa,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,IACzD,aAAa,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,EAC3D;AAAA,EACA,OAAO,EAAE,aAAa,YAAY,MAAM;AACtC,QAAI;AACF,YAAM,EAAE,QAAQ,KAAK,IAAI,kBAAkB,WAAW;AACtD,YAAM,EAAE,QAAQ,KAAK,IAAI,kBAAkB,WAAW;AAEtD,YAAM,SAAS,MAAM,QAAQ,MAAM,IAAI;AACvC,YAAM,EAAE,OAAO,MAAM,IAAI;AAEzB,YAAM,QAAkB;AAAA,QACtB;AAAA,QACA,iBAAO,MAAM,KAAK,oBAAU,MAAM,OAAO,oBAAU,MAAM,QAAQ,oBAAU,MAAM,SAAS;AAAA,QAC1F;AAAA,MACF;AAEA,iBAAW,KAAK,OAAO;AACrB,cAAM,SAAS,EAAE,SAAS,UAAU,MAAM,EAAE,SAAS,YAAY,MAAM,EAAE,SAAS,aAAa,MAAM;AACrG,cAAM,OAAO,EAAE,OAAO,QAAQ,EAAE,QAAQ,SAAS,EAAE,OAAO,QAAQ,yBAAU,EAAE,QAAQ,QAAQ,yBAAU;AACxG,cAAM,MAAM,EAAE,eAAe,SAAY,MAAM,EAAE,aAAa,KAAK,QAAQ,CAAC,CAAC,OAAO;AACpF,cAAM,KAAK,GAAG,MAAM,IAAI,KAAK,UAAU,GAAG,GAAG,CAAC,GAAG,GAAG,EAAE;AAAA,MACxD;AAEA,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,MAAM,KAAK,IAAI,EAAE,CAAC;AAAA,MACpD;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,EACzD;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,KAAK,GAAG,CAAC;AAAA,UAC1D,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO,kBAAkB,OAAO,MAAM;AAC5C,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,KAAK,UAAU,MAAM,MAAM,CAAC,EAAE,CAAC;AAAA,MACjE;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,eAAe,OAAO;AACpB,QAAM,YAAY,IAAI,qBAAqB;AAC3C,QAAM,OAAO,QAAQ,SAAS;AAChC;AAEA,KAAK,EAAE,MAAM,CAAC,QAAQ;AAAE,UAAQ,MAAM,GAAG;AAAG,UAAQ,KAAK,CAAC;AAAE,CAAC;","names":[]}
+{"version":3,"sources":["../src/mcp.ts"],"sourcesContent":["/** kordoc MCP 서버 — Claude/Cursor에서 문서 파싱 도구로 사용 */\n\nimport { McpServer } from \"@modelcontextprotocol/sdk/server/mcp.js\"\nimport { StdioServerTransport } from \"@modelcontextprotocol/sdk/server/stdio.js\"\nimport { z } from \"zod\"\nimport { readFileSync, realpathSync, openSync, readSync, closeSync, statSync, mkdirSync, writeFileSync, readdirSync, existsSync } from \"fs\"\nimport { resolve, isAbsolute, extname, basename } from \"path\"\nimport { parse, detectFormat, blocksToMarkdown, compare, extractFormFields, markdownToHwpx, markdownToXlsx } from \"./index.js\"\nimport type { ExtractedImage } from \"./types.js\"\nimport { VERSION, toArrayBuffer, sanitizeError, KordocError } from \"./utils.js\"\nimport { extractHwp5MetadataOnly } from \"./hwp5/parser.js\"\nimport { extractHwpxMetadataOnly } from \"./hwpx/parser.js\"\nimport { extractPdfMetadataOnly } from \"./pdf/parser.js\"\n\n/** 허용 파일 확장자 */\nconst ALLOWED_EXTENSIONS = new Set([\".hwp\", \".hwpx\", \".pdf\", \".xlsx\", \".docx\"])\n/** 최대 파일 크기 (500MB) */\nconst MAX_FILE_SIZE = 500 * 1024 * 1024\n\n/** 경로 정규화 및 보안 검증 */\nfunction safePath(filePath: string): string {\n  if (!filePath) throw new KordocError(\"파일 경로가 비어있습니다\")\n  const resolved = resolve(filePath)\n  const real = realpathSync(resolved)\n  if (!isAbsolute(real)) throw new KordocError(\"절대 경로만 허용됩니다\")\n  const ext = extname(real).toLowerCase()\n  if (!ALLOWED_EXTENSIONS.has(ext)) throw new KordocError(`지원하지 않는 확장자입니다: ${ext} (허용: ${[...ALLOWED_EXTENSIONS].join(\", \")})`)\n  return real\n}\n\n/** 최대 파일 크기 — metadata 전용 (50MB, 전체 파싱보다 보수적) */\nconst MAX_METADATA_FILE_SIZE = 50 * 1024 * 1024\n\n/** 파일 읽기 + 크기 검증 공통 로직 */\nfunction readValidatedFile(filePath: string, maxSize = MAX_FILE_SIZE): { buffer: ArrayBuffer; resolved: string } {\n  const resolved = safePath(filePath)\n  const fileSize = statSync(resolved).size\n  if (fileSize > maxSize) {\n    throw new KordocError(`파일이 너무 큽니다: ${(fileSize / 1024 / 1024).toFixed(1)}MB (최대 ${maxSize / 1024 / 1024}MB)`)\n  }\n  const raw = readFileSync(resolved)\n  return { buffer: toArrayBuffer(raw), resolved }\n}\n\n/** 파일 헤더(16바이트)만 읽어 포맷 감지 — 전체 파일 로드 불필요 */\nfunction detectFormatFromHeader(resolved: string): ReturnType<typeof detectFormat> {\n  const fd = openSync(resolved, \"r\")\n  try {\n    const headerBuf = Buffer.alloc(16)\n    readSync(fd, headerBuf, 0, 16, 0)\n    return detectFormat(toArrayBuffer(headerBuf))\n  } finally {\n    closeSync(fd)\n  }\n}\n\nconst server = new McpServer({\n  name: \"kordoc\",\n  version: VERSION,\n})\n\n// ─── 도구: parse_document ────────────────────────────\n\nserver.tool(\n  \"parse_document\",\n  \"한국 문서 파일(HWP, HWPX, PDF, XLSX, DOCX)을 마크다운으로 변환합니다. 파일 경로를 입력하면 포맷을 자동 감지하여 텍스트를 추출합니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로 (HWP, HWPX, PDF, XLSX, DOCX)\"),\n    image_dir: z.string().optional().describe(\"이미지 저장 폴더 경로 (기본: 파일명과 같은 이름의 폴더)\"),\n  },\n  async ({ file_path, image_dir }) => {\n    try {\n      const { buffer, resolved: resolvedFilePath } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const meta = [\n        `포맷: ${result.fileType.toUpperCase()}`,\n        result.pageCount ? `페이지: ${result.pageCount}` : null,\n        result.metadata?.title ? `제목: ${result.metadata.title}` : null,\n        result.metadata?.author ? `작성자: ${result.metadata.author}` : null,\n        result.isImageBased ? \"이미지 기반 PDF (텍스트 추출 불가)\" : null,\n      ].filter(Boolean).join(\" | \")\n\n      // outline/warnings 부가 정보 추가\n      const parts: string[] = [`[${meta}]`]\n\n      if (result.outline && result.outline.length > 0) {\n        const outlineText = result.outline.map(o => `${\"  \".repeat(o.level - 1)}- ${o.text}`).join(\"\\n\")\n        parts.push(`\\n📑 문서 구조:\\n${outlineText}`)\n      }\n\n      if (result.warnings && result.warnings.length > 0) {\n        const warnText = result.warnings.map(w => `- [p${w.page || \"?\"}] ${w.message}`).join(\"\\n\")\n        parts.push(`\\n⚠️ 경고:\\n${warnText}`)\n      }\n\n      // 이미지 저장\n      const savedImages: string[] = []\n      if (result.images?.length) {\n        const defaultDir = resolve(resolvedFilePath, \"..\", basename(resolvedFilePath).replace(/\\.[^.]+$/, \"\"))\n        const imgDir = image_dir ? resolve(image_dir) : defaultDir\n        mkdirSync(imgDir, { recursive: true })\n        for (const img of result.images) {\n          const imgPath = resolve(imgDir, img.filename)\n          writeFileSync(imgPath, img.data)\n          savedImages.push(imgPath)\n        }\n        parts.push(`\\n💾 저장된 이미지 (${savedImages.length}개):\\n${savedImages.map(p => `  ${p}`).join(\"\\n\")}`)\n      }\n\n      parts.push(`\\n\\n${result.markdown}`)\n\n      return {\n        content: [{ type: \"text\", text: parts.join(\"\") }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: convert_document ──────────────────────────\n\nserver.tool(\n  \"convert_document\",\n  \"마크다운 텍스트를 HWPX 또는 XLSX 파일로 변환하여 저장합니다.\",\n  {\n    markdown:      z.string().min(1).describe(\"변환할 마크다운 텍스트\"),\n    output_path:   z.string().min(1).describe(\"저장할 파일의 절대 경로 (.hwpx 또는 .xlsx)\"),\n    format:        z.enum([\"hwpx\", \"xlsx\"]).optional().default(\"hwpx\").describe(\"출력 포맷 (기본: hwpx)\"),\n    image_dir:     z.string().optional().describe(\"이미지 폴더 경로 (기본: output_path 파일명 폴더)\"),\n    template_path: z.string().optional().describe(\"HWPX 템플릿 파일 경로 (hwpx 전용)\"),\n  },\n  async ({ markdown, output_path, format, image_dir, template_path }) => {\n    try {\n      const resolvedOutput = resolve(output_path)\n      const outputExt = extname(resolvedOutput).toLowerCase()\n      if (outputExt !== \".hwpx\" && outputExt !== \".xlsx\") {\n        return {\n          content: [{ type: \"text\", text: `오류: 출력 파일 확장자는 .hwpx 또는 .xlsx만 허용됩니다: ${outputExt || \"(없음)\"}` }],\n          isError: true,\n        }\n      }\n      const warnings: string[] = []\n\n      // 이미지 폴더에서 이미지 로드\n      const stem = basename(resolvedOutput).replace(/\\.[^.]+$/, \"\")\n      const defaultImgDir = resolve(resolvedOutput, \"..\", stem)\n      const imgDir = image_dir ? resolve(image_dir) : defaultImgDir\n      const images: ExtractedImage[] = []\n\n      if (existsSync(imgDir)) {\n        const mimeMap: Record<string, string> = {\n          png: \"image/png\", jpg: \"image/jpeg\", jpeg: \"image/jpeg\",\n          gif: \"image/gif\", bmp: \"image/bmp\",\n        }\n        for (const entry of readdirSync(imgDir, { withFileTypes: true })) {\n          if (!entry.isFile()) continue\n          const fname = entry.name\n          const ext = extname(fname).slice(1).toLowerCase()\n          if (!mimeMap[ext]) continue\n          const data = readFileSync(resolve(imgDir, fname))\n          images.push({ filename: fname, data: new Uint8Array(data), mimeType: mimeMap[ext] })\n        }\n      }\n\n      let buf: ArrayBuffer\n      if (format === \"xlsx\") {\n        if (template_path) warnings.push(\"[warn] --template은 hwpx 전용입니다. 무시됩니다.\")\n        buf = await markdownToXlsx(markdown, { warnings, images: images.length ? images : undefined })\n      } else {\n        let templateArrayBuffer: ArrayBuffer | undefined\n        if (template_path) {\n          const tmpl = readFileSync(safePath(template_path))\n          templateArrayBuffer = tmpl.buffer.slice(tmpl.byteOffset, tmpl.byteOffset + tmpl.byteLength)\n        }\n        buf = await markdownToHwpx(markdown, {\n          warnings,\n          images: images.length ? images : undefined,\n          templateArrayBuffer,\n        })\n      }\n\n      writeFileSync(resolvedOutput, Buffer.from(buf))\n\n      const parts = [\n        `✅ 변환 완료: ${resolvedOutput}`,\n        `포맷: ${format.toUpperCase()}, 크기: ${(buf.byteLength / 1024).toFixed(1)}KB`,\n      ]\n      if (images.length) parts.push(`포함된 이미지: ${images.length}개 (${imgDir})`)\n      if (warnings.length) parts.push(`경고:\\n${warnings.map(w => `  ${w}`).join(\"\\n\")}`)\n\n      return { content: [{ type: \"text\", text: parts.join(\"\\n\") }] }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: detect_format ─────────────────────────────\n\nserver.tool(\n  \"detect_format\",\n  \"파일의 포맷을 매직 바이트로 감지합니다 (hwpx, hwp, pdf, unknown).\",\n  {\n    file_path: z.string().min(1).describe(\"감지할 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const resolved = safePath(file_path)\n      const format = detectFormatFromHeader(resolved)\n      return {\n        content: [{ type: \"text\", text: `${file_path}: ${format}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_metadata ────────────────────────────\n\nserver.tool(\n  \"parse_metadata\",\n  \"문서의 메타데이터(제목, 작성자, 날짜 등)만 빠르게 추출합니다. 전체 파싱 없이 헤더/매니페스트만 읽습니다.\",\n  {\n    file_path: z.string().min(1).describe(\"메타데이터를 추출할 문서 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const resolved = safePath(file_path)\n      const format = detectFormatFromHeader(resolved)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      // metadata 전용 크기 제한 (50MB)\n      const { buffer } = readValidatedFile(file_path, MAX_METADATA_FILE_SIZE)\n\n      let metadata\n      switch (format) {\n        case \"hwp\":\n          metadata = extractHwp5MetadataOnly(Buffer.from(buffer))\n          break\n        case \"hwpx\":\n          metadata = await extractHwpxMetadataOnly(buffer)\n          break\n        case \"pdf\":\n          metadata = await extractPdfMetadataOnly(buffer)\n          break\n      }\n\n      return {\n        content: [{ type: \"text\", text: JSON.stringify({ format, ...metadata }, null, 2) }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_pages ──────────────────────────────\n\nserver.tool(\n  \"parse_pages\",\n  \"문서의 특정 페이지/섹션 범위만 파싱합니다. PDF는 정확한 페이지, HWP/HWPX는 섹션 단위 근사치입니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로\"),\n    pages: z.string().min(1).describe(\"페이지 범위 (예: '1-3', '1,3,5-7')\"),\n  },\n  async ({ file_path, pages }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer, { pages })\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const meta = [\n        `포맷: ${result.fileType.toUpperCase()}`,\n        `범위: ${pages}`,\n        result.pageCount ? `페이지: ${result.pageCount}` : null,\n      ].filter(Boolean).join(\" | \")\n\n      return {\n        content: [{ type: \"text\", text: `[${meta}]\\n\\n${result.markdown}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_table ──────────────────────────────\n\nserver.tool(\n  \"parse_table\",\n  \"문서에서 N번째 테이블만 추출합니다 (0-based index). 테이블이 없거나 인덱스 범위를 초과하면 오류를 반환합니다.\",\n  {\n    file_path: z.string().min(1).describe(\"파싱할 문서 파일의 절대 경로\"),\n    table_index: z.number().int().min(0).describe(\"추출할 테이블 인덱스 (0부터 시작)\"),\n  },\n  async ({ file_path, table_index }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const format = detectFormat(buffer)\n\n      if (format === \"unknown\") {\n        return {\n          content: [{ type: \"text\", text: `지원하지 않는 파일 형식입니다: ${file_path}` }],\n          isError: true,\n        }\n      }\n\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패 (${result.fileType}): ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const tableBlocks = result.blocks.filter(b => b.type === \"table\" && b.table)\n      if (tableBlocks.length === 0) {\n        return {\n          content: [{ type: \"text\", text: `문서에 테이블이 없습니다.` }],\n          isError: true,\n        }\n      }\n\n      if (table_index >= tableBlocks.length) {\n        return {\n          content: [{ type: \"text\", text: `테이블 인덱스 초과: ${table_index} (총 ${tableBlocks.length}개 테이블)` }],\n          isError: true,\n        }\n      }\n\n      const tableBlock = tableBlocks[table_index]\n      const tableMarkdown = blocksToMarkdown([tableBlock])\n\n      return {\n        content: [{ type: \"text\", text: `[테이블 #${table_index} / 총 ${tableBlocks.length}개]\\n\\n${tableMarkdown}` }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: compare_documents ─────────────────────────\n\nserver.tool(\n  \"compare_documents\",\n  \"두 한국 문서 파일을 비교하여 추가/삭제/변경된 블록을 표시합니다. 신구대조표 생성에 활용됩니다. 크로스 포맷(HWP↔HWPX) 비교 가능.\",\n  {\n    file_path_a: z.string().min(1).describe(\"비교 원본 문서의 절대 경로\"),\n    file_path_b: z.string().min(1).describe(\"비교 대상 문서의 절대 경로\"),\n  },\n  async ({ file_path_a, file_path_b }) => {\n    try {\n      const { buffer: bufA } = readValidatedFile(file_path_a)\n      const { buffer: bufB } = readValidatedFile(file_path_b)\n\n      const result = await compare(bufA, bufB)\n      const { stats, diffs } = result\n\n      const lines: string[] = [\n        `## 문서 비교 결과`,\n        `추가: ${stats.added} | 삭제: ${stats.removed} | 변경: ${stats.modified} | 동일: ${stats.unchanged}`,\n        \"\",\n      ]\n\n      for (const d of diffs) {\n        const prefix = d.type === \"added\" ? \"+\" : d.type === \"removed\" ? \"-\" : d.type === \"modified\" ? \"~\" : \" \"\n        const text = d.after?.text || d.before?.text || (d.after?.table ? \"[테이블]\" : d.before?.table ? \"[테이블]\" : \"\")\n        const sim = d.similarity !== undefined ? ` (${(d.similarity * 100).toFixed(0)}%)` : \"\"\n        lines.push(`${prefix} ${text.substring(0, 200)}${sim}`)\n      }\n\n      return {\n        content: [{ type: \"text\", text: lines.join(\"\\n\") }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 도구: parse_form ───────────────────────────────\n\nserver.tool(\n  \"parse_form\",\n  \"한국 서식 문서에서 레이블-값 쌍을 구조화된 JSON으로 추출합니다. 양식/서식 문서에 최적화.\",\n  {\n    file_path: z.string().min(1).describe(\"서식 문서 파일의 절대 경로\"),\n  },\n  async ({ file_path }) => {\n    try {\n      const { buffer } = readValidatedFile(file_path)\n      const result = await parse(buffer)\n\n      if (!result.success) {\n        return {\n          content: [{ type: \"text\", text: `파싱 실패: ${result.error}` }],\n          isError: true,\n        }\n      }\n\n      const form = extractFormFields(result.blocks)\n      return {\n        content: [{ type: \"text\", text: JSON.stringify(form, null, 2) }],\n      }\n    } catch (err) {\n      return {\n        content: [{ type: \"text\", text: `오류: ${sanitizeError(err)}` }],\n        isError: true,\n      }\n    }\n  }\n)\n\n// ─── 서버 시작 ───────────────────────────────────────\n\nasync function main() {\n  const transport = new StdioServerTransport()\n  await server.connect(transport)\n}\n\nmain().catch((err) => { console.error(err); process.exit(1) })\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;AAEA,SAAS,iBAAiB;AAC1B,SAAS,4BAA4B;AACrC,SAAS,SAAS;AAClB,SAAS,cAAc,cAAc,UAAU,UAAU,WAAW,UAAU,WAAW,eAAe,aAAa,kBAAkB;AACvI,SAAS,SAAS,YAAY,SAAS,gBAAgB;AASvD,IAAM,qBAAqB,oBAAI,IAAI,CAAC,QAAQ,SAAS,QAAQ,SAAS,OAAO,CAAC;AAE9E,IAAM,gBAAgB,MAAM,OAAO;AAGnC,SAAS,SAAS,UAA0B;AAC1C,MAAI,CAAC,SAAU,OAAM,IAAI,YAAY,sEAAe;AACpD,QAAM,WAAW,QAAQ,QAAQ;AACjC,QAAM,OAAO,aAAa,QAAQ;AAClC,MAAI,CAAC,WAAW,IAAI,EAAG,OAAM,IAAI,YAAY,gEAAc;AAC3D,QAAM,MAAM,QAAQ,IAAI,EAAE,YAAY;AACtC,MAAI,CAAC,mBAAmB,IAAI,GAAG,EAAG,OAAM,IAAI,YAAY,+EAAmB,GAAG,mBAAS,CAAC,GAAG,kBAAkB,EAAE,KAAK,IAAI,CAAC,GAAG;AAC5H,SAAO;AACT;AAGA,IAAM,yBAAyB,KAAK,OAAO;AAG3C,SAAS,kBAAkB,UAAkB,UAAU,eAA0D;AAC/G,QAAM,WAAW,SAAS,QAAQ;AAClC,QAAM,WAAW,SAAS,QAAQ,EAAE;AACpC,MAAI,WAAW,SAAS;AACtB,UAAM,IAAI,YAAY,wDAAgB,WAAW,OAAO,MAAM,QAAQ,CAAC,CAAC,oBAAU,UAAU,OAAO,IAAI,KAAK;AAAA,EAC9G;AACA,QAAM,MAAM,aAAa,QAAQ;AACjC,SAAO,EAAE,QAAQ,cAAc,GAAG,GAAG,SAAS;AAChD;AAGA,SAAS,uBAAuB,UAAmD;AACjF,QAAM,KAAK,SAAS,UAAU,GAAG;AACjC,MAAI;AACF,UAAM,YAAY,OAAO,MAAM,EAAE;AACjC,aAAS,IAAI,WAAW,GAAG,IAAI,CAAC;AAChC,WAAO,aAAa,cAAc,SAAS,CAAC;AAAA,EAC9C,UAAE;AACA,cAAU,EAAE;AAAA,EACd;AACF;AAEA,IAAM,SAAS,IAAI,UAAU;AAAA,EAC3B,MAAM;AAAA,EACN,SAAS;AACX,CAAC;AAID,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,2GAA+C;AAAA,IACrF,WAAW,EAAE,OAAO,EAAE,SAAS,EAAE,SAAS,iJAAmC;AAAA,EAC/E;AAAA,EACA,OAAO,EAAE,WAAW,UAAU,MAAM;AAClC,QAAI;AACF,YAAM,EAAE,QAAQ,UAAU,iBAAiB,IAAI,kBAAkB,SAAS;AAC1E,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO;AAAA,QACX,iBAAO,OAAO,SAAS,YAAY,CAAC;AAAA,QACpC,OAAO,YAAY,uBAAQ,OAAO,SAAS,KAAK;AAAA,QAChD,OAAO,UAAU,QAAQ,iBAAO,OAAO,SAAS,KAAK,KAAK;AAAA,QAC1D,OAAO,UAAU,SAAS,uBAAQ,OAAO,SAAS,MAAM,KAAK;AAAA,QAC7D,OAAO,eAAe,uFAA2B;AAAA,MACnD,EAAE,OAAO,OAAO,EAAE,KAAK,KAAK;AAG5B,YAAM,QAAkB,CAAC,IAAI,IAAI,GAAG;AAEpC,UAAI,OAAO,WAAW,OAAO,QAAQ,SAAS,GAAG;AAC/C,cAAM,cAAc,OAAO,QAAQ,IAAI,OAAK,GAAG,KAAK,OAAO,EAAE,QAAQ,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,KAAK,IAAI;AAC/F,cAAM,KAAK;AAAA;AAAA,EAAgB,WAAW,EAAE;AAAA,MAC1C;AAEA,UAAI,OAAO,YAAY,OAAO,SAAS,SAAS,GAAG;AACjD,cAAM,WAAW,OAAO,SAAS,IAAI,OAAK,OAAO,EAAE,QAAQ,GAAG,KAAK,EAAE,OAAO,EAAE,EAAE,KAAK,IAAI;AACzF,cAAM,KAAK;AAAA;AAAA,EAAa,QAAQ,EAAE;AAAA,MACpC;AAGA,YAAM,cAAwB,CAAC;AAC/B,UAAI,OAAO,QAAQ,QAAQ;AACzB,cAAM,aAAa,QAAQ,kBAAkB,MAAM,SAAS,gBAAgB,EAAE,QAAQ,YAAY,EAAE,CAAC;AACrG,cAAM,SAAS,YAAY,QAAQ,SAAS,IAAI;AAChD,kBAAU,QAAQ,EAAE,WAAW,KAAK,CAAC;AACrC,mBAAW,OAAO,OAAO,QAAQ;AAC/B,gBAAM,UAAU,QAAQ,QAAQ,IAAI,QAAQ;AAC5C,wBAAc,SAAS,IAAI,IAAI;AAC/B,sBAAY,KAAK,OAAO;AAAA,QAC1B;AACA,cAAM,KAAK;AAAA,mDAAiB,YAAY,MAAM;AAAA,EAAQ,YAAY,IAAI,OAAK,KAAK,CAAC,EAAE,EAAE,KAAK,IAAI,CAAC,EAAE;AAAA,MACnG;AAEA,YAAM,KAAK;AAAA;AAAA,EAAO,OAAO,QAAQ,EAAE;AAEnC,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,MAAM,KAAK,EAAE,EAAE,CAAC;AAAA,MAClD;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,UAAe,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,gEAAc;AAAA,IACxD,aAAe,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,4FAAgC;AAAA,IAC1E,QAAe,EAAE,KAAK,CAAC,QAAQ,MAAM,CAAC,EAAE,SAAS,EAAE,QAAQ,MAAM,EAAE,SAAS,gDAAkB;AAAA,IAC9F,WAAe,EAAE,OAAO,EAAE,SAAS,EAAE,SAAS,0GAAoC;AAAA,IAClF,eAAe,EAAE,OAAO,EAAE,SAAS,EAAE,SAAS,uEAA0B;AAAA,EAC1E;AAAA,EACA,OAAO,EAAE,UAAU,aAAa,QAAQ,WAAW,cAAc,MAAM;AACrE,QAAI;AACF,YAAM,iBAAiB,QAAQ,WAAW;AAC1C,YAAM,YAAY,QAAQ,cAAc,EAAE,YAAY;AACtD,UAAI,cAAc,WAAW,cAAc,SAAS;AAClD,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,mIAAyC,aAAa,gBAAM,GAAG,CAAC;AAAA,UAChG,SAAS;AAAA,QACX;AAAA,MACF;AACA,YAAM,WAAqB,CAAC;AAG5B,YAAM,OAAO,SAAS,cAAc,EAAE,QAAQ,YAAY,EAAE;AAC5D,YAAM,gBAAgB,QAAQ,gBAAgB,MAAM,IAAI;AACxD,YAAM,SAAS,YAAY,QAAQ,SAAS,IAAI;AAChD,YAAM,SAA2B,CAAC;AAElC,UAAI,WAAW,MAAM,GAAG;AACtB,cAAM,UAAkC;AAAA,UACtC,KAAK;AAAA,UAAa,KAAK;AAAA,UAAc,MAAM;AAAA,UAC3C,KAAK;AAAA,UAAa,KAAK;AAAA,QACzB;AACA,mBAAW,SAAS,YAAY,QAAQ,EAAE,eAAe,KAAK,CAAC,GAAG;AAChE,cAAI,CAAC,MAAM,OAAO,EAAG;AACrB,gBAAM,QAAQ,MAAM;AACpB,gBAAM,MAAM,QAAQ,KAAK,EAAE,MAAM,CAAC,EAAE,YAAY;AAChD,cAAI,CAAC,QAAQ,GAAG,EAAG;AACnB,gBAAM,OAAO,aAAa,QAAQ,QAAQ,KAAK,CAAC;AAChD,iBAAO,KAAK,EAAE,UAAU,OAAO,MAAM,IAAI,WAAW,IAAI,GAAG,UAAU,QAAQ,GAAG,EAAE,CAAC;AAAA,QACrF;AAAA,MACF;AAEA,UAAI;AACJ,UAAI,WAAW,QAAQ;AACrB,YAAI,cAAe,UAAS,KAAK,8FAAuC;AACxE,cAAM,MAAM,eAAe,UAAU,EAAE,UAAU,QAAQ,OAAO,SAAS,SAAS,OAAU,CAAC;AAAA,MAC/F,OAAO;AACL,YAAI;AACJ,YAAI,eAAe;AACjB,gBAAM,OAAO,aAAa,SAAS,aAAa,CAAC;AACjD,gCAAsB,KAAK,OAAO,MAAM,KAAK,YAAY,KAAK,aAAa,KAAK,UAAU;AAAA,QAC5F;AACA,cAAM,MAAM,eAAe,UAAU;AAAA,UACnC;AAAA,UACA,QAAQ,OAAO,SAAS,SAAS;AAAA,UACjC;AAAA,QACF,CAAC;AAAA,MACH;AAEA,oBAAc,gBAAgB,OAAO,KAAK,GAAG,CAAC;AAE9C,YAAM,QAAQ;AAAA,QACZ,qCAAY,cAAc;AAAA,QAC1B,iBAAO,OAAO,YAAY,CAAC,oBAAU,IAAI,aAAa,MAAM,QAAQ,CAAC,CAAC;AAAA,MACxE;AACA,UAAI,OAAO,OAAQ,OAAM,KAAK,0CAAY,OAAO,MAAM,WAAM,MAAM,GAAG;AACtE,UAAI,SAAS,OAAQ,OAAM,KAAK;AAAA,EAAQ,SAAS,IAAI,OAAK,KAAK,CAAC,EAAE,EAAE,KAAK,IAAI,CAAC,EAAE;AAEhF,aAAO,EAAE,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,MAAM,KAAK,IAAI,EAAE,CAAC,EAAE;AAAA,IAC/D,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,iEAAe;AAAA,EACvD;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,WAAW,SAAS,SAAS;AACnC,YAAM,SAAS,uBAAuB,QAAQ;AAC9C,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,GAAG,SAAS,KAAK,MAAM,GAAG,CAAC;AAAA,MAC7D;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,mHAAyB;AAAA,EACjE;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,WAAW,SAAS,SAAS;AACnC,YAAM,SAAS,uBAAuB,QAAQ;AAE9C,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAGA,YAAM,EAAE,OAAO,IAAI,kBAAkB,WAAW,sBAAsB;AAEtE,UAAI;AACJ,cAAQ,QAAQ;AAAA,QACd,KAAK;AACH,qBAAW,wBAAwB,OAAO,KAAK,MAAM,CAAC;AACtD;AAAA,QACF,KAAK;AACH,qBAAW,MAAM,wBAAwB,MAAM;AAC/C;AAAA,QACF,KAAK;AACH,qBAAW,MAAM,uBAAuB,MAAM;AAC9C;AAAA,MACJ;AAEA,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,KAAK,UAAU,EAAE,QAAQ,GAAG,SAAS,GAAG,MAAM,CAAC,EAAE,CAAC;AAAA,MACpF;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,8EAAkB;AAAA,IACxD,OAAO,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,4DAA8B;AAAA,EAClE;AAAA,EACA,OAAO,EAAE,WAAW,MAAM,MAAM;AAC9B,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,QAAQ,EAAE,MAAM,CAAC;AAE5C,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO;AAAA,QACX,iBAAO,OAAO,SAAS,YAAY,CAAC;AAAA,QACpC,iBAAO,KAAK;AAAA,QACZ,OAAO,YAAY,uBAAQ,OAAO,SAAS,KAAK;AAAA,MAClD,EAAE,OAAO,OAAO,EAAE,KAAK,KAAK;AAE5B,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,IAAI,IAAI;AAAA;AAAA,EAAQ,OAAO,QAAQ,GAAG,CAAC;AAAA,MACrE;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,8EAAkB;AAAA,IACxD,aAAa,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,CAAC,EAAE,SAAS,uFAAsB;AAAA,EACtE;AAAA,EACA,OAAO,EAAE,WAAW,YAAY,MAAM;AACpC,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,aAAa,MAAM;AAElC,UAAI,WAAW,WAAW;AACxB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,sFAAqB,SAAS,GAAG,CAAC;AAAA,UAClE,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,QAAQ,MAAM,OAAO,KAAK,GAAG,CAAC;AAAA,UAC/E,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,cAAc,OAAO,OAAO,OAAO,OAAK,EAAE,SAAS,WAAW,EAAE,KAAK;AAC3E,UAAI,YAAY,WAAW,GAAG;AAC5B,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,wEAAiB,CAAC;AAAA,UAClD,SAAS;AAAA,QACX;AAAA,MACF;AAEA,UAAI,eAAe,YAAY,QAAQ;AACrC,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,uDAAe,WAAW,YAAO,YAAY,MAAM,6BAAS,CAAC;AAAA,UAC7F,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,aAAa,YAAY,WAAW;AAC1C,YAAM,gBAAgB,iBAAiB,CAAC,UAAU,CAAC;AAEnD,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,wBAAS,WAAW,aAAQ,YAAY,MAAM;AAAA;AAAA,EAAS,aAAa,GAAG,CAAC;AAAA,MAC1G;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,aAAa,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,IACzD,aAAa,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,EAC3D;AAAA,EACA,OAAO,EAAE,aAAa,YAAY,MAAM;AACtC,QAAI;AACF,YAAM,EAAE,QAAQ,KAAK,IAAI,kBAAkB,WAAW;AACtD,YAAM,EAAE,QAAQ,KAAK,IAAI,kBAAkB,WAAW;AAEtD,YAAM,SAAS,MAAM,QAAQ,MAAM,IAAI;AACvC,YAAM,EAAE,OAAO,MAAM,IAAI;AAEzB,YAAM,QAAkB;AAAA,QACtB;AAAA,QACA,iBAAO,MAAM,KAAK,oBAAU,MAAM,OAAO,oBAAU,MAAM,QAAQ,oBAAU,MAAM,SAAS;AAAA,QAC1F;AAAA,MACF;AAEA,iBAAW,KAAK,OAAO;AACrB,cAAM,SAAS,EAAE,SAAS,UAAU,MAAM,EAAE,SAAS,YAAY,MAAM,EAAE,SAAS,aAAa,MAAM;AACrG,cAAM,OAAO,EAAE,OAAO,QAAQ,EAAE,QAAQ,SAAS,EAAE,OAAO,QAAQ,yBAAU,EAAE,QAAQ,QAAQ,yBAAU;AACxG,cAAM,MAAM,EAAE,eAAe,SAAY,MAAM,EAAE,aAAa,KAAK,QAAQ,CAAC,CAAC,OAAO;AACpF,cAAM,KAAK,GAAG,MAAM,IAAI,KAAK,UAAU,GAAG,GAAG,CAAC,GAAG,GAAG,EAAE;AAAA,MACxD;AAEA,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,MAAM,KAAK,IAAI,EAAE,CAAC;AAAA,MACpD;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,OAAO;AAAA,EACL;AAAA,EACA;AAAA,EACA;AAAA,IACE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,EAAE,SAAS,wEAAiB;AAAA,EACzD;AAAA,EACA,OAAO,EAAE,UAAU,MAAM;AACvB,QAAI;AACF,YAAM,EAAE,OAAO,IAAI,kBAAkB,SAAS;AAC9C,YAAM,SAAS,MAAM,MAAM,MAAM;AAEjC,UAAI,CAAC,OAAO,SAAS;AACnB,eAAO;AAAA,UACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,8BAAU,OAAO,KAAK,GAAG,CAAC;AAAA,UAC1D,SAAS;AAAA,QACX;AAAA,MACF;AAEA,YAAM,OAAO,kBAAkB,OAAO,MAAM;AAC5C,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,KAAK,UAAU,MAAM,MAAM,CAAC,EAAE,CAAC;AAAA,MACjE;AAAA,IACF,SAAS,KAAK;AACZ,aAAO;AAAA,QACL,SAAS,CAAC,EAAE,MAAM,QAAQ,MAAM,iBAAO,cAAc,GAAG,CAAC,GAAG,CAAC;AAAA,QAC7D,SAAS;AAAA,MACX;AAAA,IACF;AAAA,EACF;AACF;AAIA,eAAe,OAAO;AACpB,QAAM,YAAY,IAAI,qBAAqB;AAC3C,QAAM,OAAO,QAAQ,SAAS;AAChC;AAEA,KAAK,EAAE,MAAM,CAAC,QAAQ;AAAE,UAAQ,MAAM,GAAG;AAAG,UAAQ,KAAK,CAAC;AAAE,CAAC;","names":[]}

package/dist/{page-range-737B4EZW.js → page-range-ALIRXAL5.js}

@@ -2,7 +2,8 @@
 import {
   parsePageRange
 } from "./chunk-MOL7MDBG.js";
+import "./chunk-ZWE3DS7E.js";
 export {
   parsePageRange
 };
-//# sourceMappingURL=page-range-737B4EZW.js.map
+//# sourceMappingURL=page-range-ALIRXAL5.js.map

package/dist/provider-XVKP5OGI.js

@@ -0,0 +1,167 @@
+#!/usr/bin/env node
+import "./chunk-ZWE3DS7E.js";
+
+// src/ocr/markdown-to-blocks.ts
+function markdownToBlocks(markdown, pageNumber) {
+  const blocks = [];
+  const lines = markdown.split("\n");
+  let i = 0;
+  while (i < lines.length) {
+    const line = lines[i];
+    if (line.trim() === "") {
+      i++;
+      continue;
+    }
+    const headingMatch = line.match(/^(#{1,6})\s+(.+)$/);
+    if (headingMatch) {
+      blocks.push({
+        type: "heading",
+        level: headingMatch[1].length,
+        text: headingMatch[2].trim(),
+        pageNumber
+      });
+      i++;
+      continue;
+    }
+    if (/^[-*_]{3,}\s*$/.test(line.trim())) {
+      blocks.push({ type: "separator", pageNumber });
+      i++;
+      continue;
+    }
+    if (line.trim().startsWith("|")) {
+      const tableLines = [];
+      while (i < lines.length && lines[i].trim().startsWith("|")) {
+        tableLines.push(lines[i]);
+        i++;
+      }
+      const table = parseMarkdownTable(tableLines);
+      if (table) {
+        blocks.push({ type: "table", table, pageNumber });
+      }
+      continue;
+    }
+    const ulMatch = line.match(/^(\s*)[-*+]\s+(.+)$/);
+    if (ulMatch) {
+      blocks.push({
+        type: "list",
+        listType: "unordered",
+        text: ulMatch[2].trim(),
+        pageNumber
+      });
+      i++;
+      continue;
+    }
+    const olMatch = line.match(/^(\s*)\d+\.\s+(.+)$/);
+    if (olMatch) {
+      blocks.push({
+        type: "list",
+        listType: "ordered",
+        text: olMatch[2].trim(),
+        pageNumber
+      });
+      i++;
+      continue;
+    }
+    const paraLines = [];
+    while (i < lines.length && lines[i].trim() !== "" && !isStructuralLine(lines[i])) {
+      paraLines.push(lines[i].trim());
+      i++;
+    }
+    if (paraLines.length > 0) {
+      blocks.push({
+        type: "paragraph",
+        text: paraLines.join("\n"),
+        pageNumber
+      });
+    }
+  }
+  return blocks;
+}
+function isStructuralLine(line) {
+  if (/^#{1,6}\s+/.test(line)) return true;
+  if (line.trim().startsWith("|")) return true;
+  if (/^[-*_]{3,}\s*$/.test(line.trim())) return true;
+  if (/^\s*[-*+]\s+/.test(line)) return true;
+  if (/^\s*\d+\.\s+/.test(line)) return true;
+  return false;
+}
+function parseMarkdownTable(lines) {
+  const hasSeparator = lines.some((line) => /^\|[\s:|-]+\|$/.test(line.trim()));
+  const rows = [];
+  let maxCols = 0;
+  for (const line of lines) {
+    if (/^\|\s*:?-+:?\s*(\|\s*:?-+:?\s*)+\|?\s*$/.test(line.trim())) continue;
+    const parts = line.split("|");
+    const cells = parts.slice(1, parts[parts.length - 1].trim() === "" ? -1 : void 0).map((cell) => ({
+      text: cell.trim(),
+      colSpan: 1,
+      rowSpan: 1
+    }));
+    if (cells.length > 0) {
+      rows.push(cells);
+      maxCols = Math.max(maxCols, cells.length);
+    }
+  }
+  if (rows.length === 0) return null;
+  for (const row of rows) {
+    while (row.length < maxCols) {
+      row.push({ text: "", colSpan: 1, rowSpan: 1 });
+    }
+  }
+  return {
+    rows: rows.length,
+    cols: maxCols,
+    cells: rows,
+    hasHeader: hasSeparator && rows.length > 1
+  };
+}
+
+// src/ocr/provider.ts
+async function ocrPages(doc, provider, pageFilter, effectivePageCount, warnings) {
+  const blocks = [];
+  for (let i = 1; i <= effectivePageCount; i++) {
+    if (pageFilter && !pageFilter.has(i)) continue;
+    const page = await doc.getPage(i);
+    try {
+      const imageData = await renderPageToPng(page);
+      const result = await provider(imageData, i, "image/png");
+      if (typeof result === "string") {
+        if (result.trim()) {
+          blocks.push({ type: "paragraph", text: result.trim(), pageNumber: i });
+        }
+      } else if (result && typeof result === "object" && "markdown" in result) {
+        const structured = result;
+        if (structured.markdown.trim()) {
+          const pageBlocks = markdownToBlocks(structured.markdown, i);
+          for (const b of pageBlocks) blocks.push(b);
+        }
+      }
+    } catch (err) {
+      warnings?.push({
+        page: i,
+        message: `\uD398\uC774\uC9C0 ${i} OCR \uC2E4\uD328: ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
+        code: "OCR_PAGE_FAILED"
+      });
+    }
+  }
+  return blocks;
+}
+async function renderPageToPng(page) {
+  let createCanvas;
+  try {
+    const canvasModule = await import("canvas");
+    createCanvas = canvasModule.createCanvas;
+  } catch {
+    throw new Error("OCR\uC744 \uC0AC\uC6A9\uD558\uB824\uBA74 'canvas' \uD328\uD0A4\uC9C0\uB97C \uC124\uCE58\uD558\uC138\uC694: npm install canvas");
+  }
+  const scale = 2;
+  const viewport = page.getViewport({ scale });
+  const canvas = createCanvas(Math.floor(viewport.width), Math.floor(viewport.height));
+  const ctx = canvas.getContext("2d");
+  await page.render({ canvasContext: ctx, viewport }).promise;
+  return new Uint8Array(canvas.toBuffer("image/png"));
+}
+export {
+  ocrPages
+};
+//# sourceMappingURL=provider-XVKP5OGI.js.map

package/dist/provider-XVKP5OGI.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/ocr/markdown-to-blocks.ts","../src/ocr/provider.ts"],"sourcesContent":["/**\n * Markdown → IRBlock[] 역파싱\n *\n * Vision LLM(gemini/claude/codex 등)이 반환한 Markdown 문자열을\n * kordoc의 IRBlock[] 중간 표현으로 변환.\n * 기존 blocksToMarkdown()의 역방향 처리.\n */\n\nimport type { IRBlock, IRTable, IRCell } from \"../types.js\"\n\n/**\n * Markdown 문자열을 IRBlock[] 배열로 변환.\n *\n * 지원 요소:\n * - 헤딩: # ~ ######\n * - 테이블: | col1 | col2 | (파이프 구분, |---|---| 구분선 포함)\n * - 순서/비순서 리스트: - / 1.\n * - 구분선: ---, ***, ___\n * - 일반 텍스트 (paragraph)\n */\nexport function markdownToBlocks(markdown: string, pageNumber: number): IRBlock[] {\n  const blocks: IRBlock[] = []\n  const lines = markdown.split(\"\\n\")\n  let i = 0\n\n  while (i < lines.length) {\n    const line = lines[i]\n\n    // 빈 줄 스킵\n    if (line.trim() === \"\") {\n      i++\n      continue\n    }\n\n    // 1. 헤딩: # ~ ######\n    const headingMatch = line.match(/^(#{1,6})\\s+(.+)$/)\n    if (headingMatch) {\n      blocks.push({\n        type: \"heading\",\n        level: headingMatch[1].length,\n        text: headingMatch[2].trim(),\n        pageNumber,\n      })\n      i++\n      continue\n    }\n\n    // 2. 구분선: ---, ***, ___\n    if (/^[-*_]{3,}\\s*$/.test(line.trim())) {\n      blocks.push({ type: \"separator\", pageNumber })\n      i++\n      continue\n    }\n\n    // 3. 테이블: | 로 시작하는 연속 행 수집\n    if (line.trim().startsWith(\"|\")) {\n      const tableLines: string[] = []\n      while (i < lines.length && lines[i].trim().startsWith(\"|\")) {\n        tableLines.push(lines[i])\n        i++\n      }\n      const table = parseMarkdownTable(tableLines)\n      if (table) {\n        blocks.push({ type: \"table\", table, pageNumber })\n      }\n      continue\n    }\n\n    // 4. 비순서 리스트: -, *, +\n    const ulMatch = line.match(/^(\\s*)[-*+]\\s+(.+)$/)\n    if (ulMatch) {\n      blocks.push({\n        type: \"list\",\n        listType: \"unordered\",\n        text: ulMatch[2].trim(),\n        pageNumber,\n      })\n      i++\n      continue\n    }\n\n    // 5. 순서 리스트: 1.\n    const olMatch = line.match(/^(\\s*)\\d+\\.\\s+(.+)$/)\n    if (olMatch) {\n      blocks.push({\n        type: \"list\",\n        listType: \"ordered\",\n        text: olMatch[2].trim(),\n        pageNumber,\n      })\n      i++\n      continue\n    }\n\n    // 6. 일반 텍스트 — 구조적 행이 나올 때까지 병합\n    const paraLines: string[] = []\n    while (i < lines.length && lines[i].trim() !== \"\" && !isStructuralLine(lines[i])) {\n      paraLines.push(lines[i].trim())\n      i++\n    }\n    if (paraLines.length > 0) {\n      blocks.push({\n        type: \"paragraph\",\n        text: paraLines.join(\"\\n\"),\n        pageNumber,\n      })\n    }\n  }\n\n  return blocks\n}\n\n/**\n * 구조적 행 판별 — paragraph 병합 중단 트리거.\n */\nfunction isStructuralLine(line: string): boolean {\n  if (/^#{1,6}\\s+/.test(line)) return true\n  if (line.trim().startsWith(\"|\")) return true\n  if (/^[-*_]{3,}\\s*$/.test(line.trim())) return true\n  if (/^\\s*[-*+]\\s+/.test(line)) return true\n  if (/^\\s*\\d+\\.\\s+/.test(line)) return true\n  return false\n}\n\n/**\n * Markdown 테이블 행 배열을 IRTable로 변환.\n *\n * 구분선 행(|---|---|)은 제거 후 데이터 행만 파싱.\n * hasHeader: 구분선이 있었으면 true.\n */\nfunction parseMarkdownTable(lines: string[]): IRTable | null {\n  const hasSeparator = lines.some(line => /^\\|[\\s:|-]+\\|$/.test(line.trim()))\n\n  const rows: IRCell[][] = []\n  let maxCols = 0\n\n  for (const line of lines) {\n    // 구분선 행 스킵: |---|---| 패턴\n    if (/^\\|\\s*:?-+:?\\s*(\\|\\s*:?-+:?\\s*)+\\|?\\s*$/.test(line.trim())) continue\n\n    const parts = line.split(\"|\")\n    // 앞뒤 빈 요소 제거 (| 로 시작/종료하는 행)\n    const cells: IRCell[] = parts\n      .slice(1, parts[parts.length - 1].trim() === \"\" ? -1 : undefined)\n      .map(cell => ({\n        text: cell.trim(),\n        colSpan: 1,\n        rowSpan: 1,\n      }))\n\n    if (cells.length > 0) {\n      rows.push(cells)\n      maxCols = Math.max(maxCols, cells.length)\n    }\n  }\n\n  if (rows.length === 0) return null\n\n  // 열 수 통일 (부족한 셀은 빈 셀로 채움)\n  for (const row of rows) {\n    while (row.length < maxCols) {\n      row.push({ text: \"\", colSpan: 1, rowSpan: 1 })\n    }\n  }\n\n  return {\n    rows: rows.length,\n    cols: maxCols,\n    cells: rows,\n    hasHeader: hasSeparator && rows.length > 1,\n  }\n}\n","/**\n * OCR 프로바이더 브릿지 — PDF 페이지를 이미지로 렌더링하여 OCR 호출\n *\n * kordoc은 OCR 라이브러리를 번들하지 않음.\n * 사용자가 OcrProvider 함수를 제공하면 이미지 기반 PDF도 텍스트 추출 가능.\n *\n * @example\n * ```ts\n * import { parse } from \"kordoc\"\n *\n * const result = await parse(buffer, {\n *   ocr: async (pageImage, pageNumber, mimeType) => {\n *     // Tesseract, Claude Vision, Google Vision 등 사용\n *     return await myOcrService.recognize(pageImage)\n *   }\n * })\n * ```\n */\n\nimport type { OcrProvider, IRBlock, ParseWarning, StructuredOcrResult } from \"../types.js\"\nimport { markdownToBlocks } from \"./markdown-to-blocks.js\"\n\n/**\n * 이미지 기반 PDF 페이지에 OCR을 적용하여 IRBlock[] 반환.\n *\n * pdfjs page 객체에서 viewport + render를 통해 PNG 생성 후\n * 사용자 제공 OcrProvider 호출.\n *\n * - string 반환: 단순 텍스트 → paragraph 블록\n * - StructuredOcrResult 반환: Markdown → markdownToBlocks()로 구조화\n *\n * canvas 미설치 시 pdfjs render 불가하므로 에러 반환.\n */\nexport async function ocrPages(\n  doc: { numPages: number; getPage(n: number): Promise<PdfPageProxy> },\n  provider: OcrProvider,\n  pageFilter: Set<number> | null,\n  effectivePageCount: number,\n  warnings?: ParseWarning[]\n): Promise<IRBlock[]> {\n  const blocks: IRBlock[] = []\n\n  for (let i = 1; i <= effectivePageCount; i++) {\n    if (pageFilter && !pageFilter.has(i)) continue\n    const page = await doc.getPage(i)\n    try {\n      const imageData = await renderPageToPng(page)\n      const result = await provider(imageData, i, \"image/png\")\n\n      if (typeof result === \"string\") {\n        // 기존 동작: 순수 텍스트 → paragraph 블록\n        if (result.trim()) {\n          blocks.push({ type: \"paragraph\", text: result.trim(), pageNumber: i })\n        }\n      } else if (result && typeof result === \"object\" && \"markdown\" in result) {\n        // 신규: 구조화된 결과 → Markdown → IRBlock[]\n        const structured = result as StructuredOcrResult\n        if (structured.markdown.trim()) {\n          const pageBlocks = markdownToBlocks(structured.markdown, i)\n          for (const b of pageBlocks) blocks.push(b)\n        }\n      }\n    } catch (err) {\n      // 개별 페이지 실패 시 경고 발행 후 계속 진행\n      warnings?.push({\n        page: i,\n        message: `페이지 ${i} OCR 실패: ${err instanceof Error ? err.message : \"알 수 없는 오류\"}`,\n        code: \"OCR_PAGE_FAILED\",\n      })\n    }\n  }\n\n  return blocks\n}\n\ninterface PdfPageProxy {\n  getViewport(params: { scale: number }): { width: number; height: number }\n  render(params: { canvasContext: unknown; viewport: unknown }): { promise: Promise<void> }\n}\n\n/**\n * PDF 페이지를 PNG로 렌더링.\n * node-canvas가 설치되어 있어야 동작.\n * 미설치 시 에러 throw → 호출측에서 catch.\n */\nasync function renderPageToPng(page: PdfPageProxy): Promise<Uint8Array> {\n  // node-canvas 동적 로드 (선택적 의존성)\n  let createCanvas: (w: number, h: number) => { getContext(t: string): unknown; toBuffer(t: string): Buffer }\n  try {\n    const canvasModule = await import(\"canvas\")\n    createCanvas = canvasModule.createCanvas\n  } catch {\n    throw new Error(\"OCR을 사용하려면 'canvas' 패키지를 설치하세요: npm install canvas\")\n  }\n\n  const scale = 2.0 // 300 DPI 근사\n  const viewport = page.getViewport({ scale })\n  const canvas = createCanvas(Math.floor(viewport.width), Math.floor(viewport.height))\n  const ctx = canvas.getContext(\"2d\")\n\n  await page.render({ canvasContext: ctx, viewport }).promise\n  return new Uint8Array(canvas.toBuffer(\"image/png\"))\n}\n"],"mappings":";;;;AAoBO,SAAS,iBAAiB,UAAkB,YAA+B;AAChF,QAAM,SAAoB,CAAC;AAC3B,QAAM,QAAQ,SAAS,MAAM,IAAI;AACjC,MAAI,IAAI;AAER,SAAO,IAAI,MAAM,QAAQ;AACvB,UAAM,OAAO,MAAM,CAAC;AAGpB,QAAI,KAAK,KAAK,MAAM,IAAI;AACtB;AACA;AAAA,IACF;AAGA,UAAM,eAAe,KAAK,MAAM,mBAAmB;AACnD,QAAI,cAAc;AAChB,aAAO,KAAK;AAAA,QACV,MAAM;AAAA,QACN,OAAO,aAAa,CAAC,EAAE;AAAA,QACvB,MAAM,aAAa,CAAC,EAAE,KAAK;AAAA,QAC3B;AAAA,MACF,CAAC;AACD;AACA;AAAA,IACF;AAGA,QAAI,iBAAiB,KAAK,KAAK,KAAK,CAAC,GAAG;AACtC,aAAO,KAAK,EAAE,MAAM,aAAa,WAAW,CAAC;AAC7C;AACA;AAAA,IACF;AAGA,QAAI,KAAK,KAAK,EAAE,WAAW,GAAG,GAAG;AAC/B,YAAM,aAAuB,CAAC;AAC9B,aAAO,IAAI,MAAM,UAAU,MAAM,CAAC,EAAE,KAAK,EAAE,WAAW,GAAG,GAAG;AAC1D,mBAAW,KAAK,MAAM,CAAC,CAAC;AACxB;AAAA,MACF;AACA,YAAM,QAAQ,mBAAmB,UAAU;AAC3C,UAAI,OAAO;AACT,eAAO,KAAK,EAAE,MAAM,SAAS,OAAO,WAAW,CAAC;AAAA,MAClD;AACA;AAAA,IACF;AAGA,UAAM,UAAU,KAAK,MAAM,qBAAqB;AAChD,QAAI,SAAS;AACX,aAAO,KAAK;AAAA,QACV,MAAM;AAAA,QACN,UAAU;AAAA,QACV,MAAM,QAAQ,CAAC,EAAE,KAAK;AAAA,QACtB;AAAA,MACF,CAAC;AACD;AACA;AAAA,IACF;AAGA,UAAM,UAAU,KAAK,MAAM,qBAAqB;AAChD,QAAI,SAAS;AACX,aAAO,KAAK;AAAA,QACV,MAAM;AAAA,QACN,UAAU;AAAA,QACV,MAAM,QAAQ,CAAC,EAAE,KAAK;AAAA,QACtB;AAAA,MACF,CAAC;AACD;AACA;AAAA,IACF;AAGA,UAAM,YAAsB,CAAC;AAC7B,WAAO,IAAI,MAAM,UAAU,MAAM,CAAC,EAAE,KAAK,MAAM,MAAM,CAAC,iBAAiB,MAAM,CAAC,CAAC,GAAG;AAChF,gBAAU,KAAK,MAAM,CAAC,EAAE,KAAK,CAAC;AAC9B;AAAA,IACF;AACA,QAAI,UAAU,SAAS,GAAG;AACxB,aAAO,KAAK;AAAA,QACV,MAAM;AAAA,QACN,MAAM,UAAU,KAAK,IAAI;AAAA,QACzB;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAKA,SAAS,iBAAiB,MAAuB;AAC/C,MAAI,aAAa,KAAK,IAAI,EAAG,QAAO;AACpC,MAAI,KAAK,KAAK,EAAE,WAAW,GAAG,EAAG,QAAO;AACxC,MAAI,iBAAiB,KAAK,KAAK,KAAK,CAAC,EAAG,QAAO;AAC/C,MAAI,eAAe,KAAK,IAAI,EAAG,QAAO;AACtC,MAAI,eAAe,KAAK,IAAI,EAAG,QAAO;AACtC,SAAO;AACT;AAQA,SAAS,mBAAmB,OAAiC;AAC3D,QAAM,eAAe,MAAM,KAAK,UAAQ,iBAAiB,KAAK,KAAK,KAAK,CAAC,CAAC;AAE1E,QAAM,OAAmB,CAAC;AAC1B,MAAI,UAAU;AAEd,aAAW,QAAQ,OAAO;AAExB,QAAI,0CAA0C,KAAK,KAAK,KAAK,CAAC,EAAG;AAEjE,UAAM,QAAQ,KAAK,MAAM,GAAG;AAE5B,UAAM,QAAkB,MACrB,MAAM,GAAG,MAAM,MAAM,SAAS,CAAC,EAAE,KAAK,MAAM,KAAK,KAAK,MAAS,EAC/D,IAAI,WAAS;AAAA,MACZ,MAAM,KAAK,KAAK;AAAA,MAChB,SAAS;AAAA,MACT,SAAS;AAAA,IACX,EAAE;AAEJ,QAAI,MAAM,SAAS,GAAG;AACpB,WAAK,KAAK,KAAK;AACf,gBAAU,KAAK,IAAI,SAAS,MAAM,MAAM;AAAA,IAC1C;AAAA,EACF;AAEA,MAAI,KAAK,WAAW,EAAG,QAAO;AAG9B,aAAW,OAAO,MAAM;AACtB,WAAO,IAAI,SAAS,SAAS;AAC3B,UAAI,KAAK,EAAE,MAAM,IAAI,SAAS,GAAG,SAAS,EAAE,CAAC;AAAA,IAC/C;AAAA,EACF;AAEA,SAAO;AAAA,IACL,MAAM,KAAK;AAAA,IACX,MAAM;AAAA,IACN,OAAO;AAAA,IACP,WAAW,gBAAgB,KAAK,SAAS;AAAA,EAC3C;AACF;;;AC1IA,eAAsB,SACpB,KACA,UACA,YACA,oBACA,UACoB;AACpB,QAAM,SAAoB,CAAC;AAE3B,WAAS,IAAI,GAAG,KAAK,oBAAoB,KAAK;AAC5C,QAAI,cAAc,CAAC,WAAW,IAAI,CAAC,EAAG;AACtC,UAAM,OAAO,MAAM,IAAI,QAAQ,CAAC;AAChC,QAAI;AACF,YAAM,YAAY,MAAM,gBAAgB,IAAI;AAC5C,YAAM,SAAS,MAAM,SAAS,WAAW,GAAG,WAAW;AAEvD,UAAI,OAAO,WAAW,UAAU;AAE9B,YAAI,OAAO,KAAK,GAAG;AACjB,iBAAO,KAAK,EAAE,MAAM,aAAa,MAAM,OAAO,KAAK,GAAG,YAAY,EAAE,CAAC;AAAA,QACvE;AAAA,MACF,WAAW,UAAU,OAAO,WAAW,YAAY,cAAc,QAAQ;AAEvE,cAAM,aAAa;AACnB,YAAI,WAAW,SAAS,KAAK,GAAG;AAC9B,gBAAM,aAAa,iBAAiB,WAAW,UAAU,CAAC;AAC1D,qBAAW,KAAK,WAAY,QAAO,KAAK,CAAC;AAAA,QAC3C;AAAA,MACF;AAAA,IACF,SAAS,KAAK;AAEZ,gBAAU,KAAK;AAAA,QACb,MAAM;AAAA,QACN,SAAS,sBAAO,CAAC,sBAAY,eAAe,QAAQ,IAAI,UAAU,yCAAW;AAAA,QAC7E,MAAM;AAAA,MACR,CAAC;AAAA,IACH;AAAA,EACF;AAEA,SAAO;AACT;AAYA,eAAe,gBAAgB,MAAyC;AAEtE,MAAI;AACJ,MAAI;AACF,UAAM,eAAe,MAAM,OAAO,QAAQ;AAC1C,mBAAe,aAAa;AAAA,EAC9B,QAAQ;AACN,UAAM,IAAI,MAAM,+HAAoD;AAAA,EACtE;AAEA,QAAM,QAAQ;AACd,QAAM,WAAW,KAAK,YAAY,EAAE,MAAM,CAAC;AAC3C,QAAM,SAAS,aAAa,KAAK,MAAM,SAAS,KAAK,GAAG,KAAK,MAAM,SAAS,MAAM,CAAC;AACnF,QAAM,MAAM,OAAO,WAAW,IAAI;AAElC,QAAM,KAAK,OAAO,EAAE,eAAe,KAAK,SAAS,CAAC,EAAE;AACpD,SAAO,IAAI,WAAW,OAAO,SAAS,WAAW,CAAC;AACpD;","names":[]}