@clazic/kordoc 2.3.0 → 2.3.1

package/dist/{batch-provider-FUCIIS4M.js → batch-provider-PCT4I4LK.js}

@@ -2,7 +2,7 @@
 import "./chunk-ZWE3DS7E.js";
 
 // src/ocr/batch-provider.ts
-import { spawnSync } from "child_process";
+import { spawn } from "child_process";
 import { writeFileSync, readFileSync, unlinkSync, mkdirSync } from "fs";
 import { join } from "path";
 import { tmpdir } from "os";
@@ -36,9 +36,9 @@ function createBatchCliProvider(mode, batchSize) {
         }
         let output;
         if (mode === "codex") {
-          output = callBatchCodexCli(tempFiles);
+          output = await callBatchCodexCli(tempFiles);
         } else {
-          output = callBatchCli(mode, tempFiles);
+          output = await callBatchCli(mode, tempFiles);
         }
         const cleaned = stripCodeFence(output.trim());
         const parts = cleaned.split(/<!--\s*PAGE_BREAK\s*-->/).map((p) => p.trim()).filter((p) => p.length > 0);
@@ -60,7 +60,48 @@ function createBatchCliProvider(mode, batchSize) {
     }
   };
 }
-function callBatchCli(mode, imagePaths) {
+function spawnAsync(cmd, args, opts) {
+  return new Promise((resolve, reject) => {
+    const child = spawn(cmd, args, {
+      cwd: opts.cwd,
+      env: process.env,
+      stdio: ["pipe", "pipe", "pipe"]
+    });
+    let stdout = "";
+    let stderr = "";
+    let killed = false;
+    child.stdout.setEncoding("utf-8");
+    child.stderr.setEncoding("utf-8");
+    child.stdout.on("data", (d) => {
+      stdout += d;
+    });
+    child.stderr.on("data", (d) => {
+      stderr += d;
+    });
+    const timer = setTimeout(() => {
+      killed = true;
+      child.kill("SIGTERM");
+    }, opts.timeoutMs);
+    if (opts.stdin !== void 0) {
+      child.stdin.end(opts.stdin);
+    } else {
+      child.stdin.end();
+    }
+    child.on("close", (code) => {
+      clearTimeout(timer);
+      if (killed) {
+        reject(new Error(`\uD0C0\uC784\uC544\uC6C3 (${Math.round(opts.timeoutMs / 1e3)}\uCD08)`));
+      } else {
+        resolve({ stdout, stderr, exitCode: code ?? 1 });
+      }
+    });
+    child.on("error", (err) => {
+      clearTimeout(timer);
+      reject(err);
+    });
+  });
+}
+async function callBatchCli(mode, imagePaths) {
   const fileRefs = imagePaths.map((p) => `@${p}`).join("\n");
   const prompt = `${BATCH_OCR_PROMPT}
 
@@ -76,24 +117,18 @@ ${fileRefs}`;
     if (model) args.push("--model", model);
   }
   const timeoutMs = 6e4 + imagePaths.length * 2e4;
-  const result = spawnSync(mode, args, {
-    encoding: "utf-8",
-    timeout: timeoutMs,
-    maxBuffer: 50 * 1024 * 1024,
-    // 50MB (large batch output)
+  const result = await spawnAsync(mode, args, {
+    timeoutMs,
     ...mode === "claude" ? { cwd: tmpdir() } : {}
   });
-  if (result.error) {
-    throw new Error(`${mode} \uBC30\uCE58 OCR \uC2E4\uD328: ${result.error.message}`);
-  }
-  if (result.status !== 0) {
-    const errMsg = result.stderr?.trim() || `exit code ${result.status}`;
+  if (result.exitCode !== 0) {
+    const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`;
     throw new Error(`${mode} \uBC30\uCE58 OCR \uC2E4\uD328: ${errMsg}`);
   }
   return result.stdout || "";
 }
-function callBatchCodexCli(imagePaths) {
-  const outPath = join(tmpdir(), `kordoc-codex-batch-${Date.now()}.txt`);
+async function callBatchCodexCli(imagePaths) {
+  const outPath = join(tmpdir(), `kordoc-codex-batch-${Date.now()}-${Math.random().toString(36).slice(2)}.txt`);
   try {
     const args = ["exec", BATCH_OCR_PROMPT];
     for (const p of imagePaths) {
@@ -103,17 +138,12 @@ function callBatchCodexCli(imagePaths) {
     const model = process.env.KORDOC_CODEX_MODEL;
     if (model) args.push("--model", model);
     const timeoutMs = 6e4 + imagePaths.length * 2e4;
-    const result = spawnSync("codex", args, {
-      encoding: "utf-8",
-      timeout: timeoutMs,
-      maxBuffer: 50 * 1024 * 1024,
-      input: ""
+    const result = await spawnAsync("codex", args, {
+      timeoutMs,
+      stdin: ""
     });
-    if (result.error) {
-      throw new Error(`codex \uBC30\uCE58 OCR \uC2E4\uD328: ${result.error.message}`);
-    }
-    if (result.status !== 0) {
-      const errMsg = result.stderr?.trim() || `exit code ${result.status}`;
+    if (result.exitCode !== 0) {
+      const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`;
       throw new Error(`codex \uBC30\uCE58 OCR \uC2E4\uD328: ${errMsg}`);
     }
     try {
@@ -136,4 +166,4 @@ export {
   DEFAULT_BATCH_SIZES,
   createBatchCliProvider
 };
-//# sourceMappingURL=batch-provider-FUCIIS4M.js.map
+//# sourceMappingURL=batch-provider-PCT4I4LK.js.map

package/dist/batch-provider-PCT4I4LK.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/ocr/batch-provider.ts"],"sourcesContent":["/**\n * CLI 배치 OCR 프로바이더\n *\n * 여러 페이지 이미지를 단일 CLI 호출로 처리하여 API 호출 수를 대폭 감소.\n * gemini/claude: @file 멀티 참조, codex: --image 멀티 플래그\n *\n * 299페이지 기준:\n * - 기존: CLI 299회 호출 (~30분)\n * - 배치: CLI 3~6회 호출 (~3분)\n */\n\nimport { spawn } from \"child_process\"\nimport { writeFileSync, readFileSync, unlinkSync, mkdirSync } from \"fs\"\nimport { join } from \"path\"\nimport { tmpdir } from \"os\"\nimport type { StructuredOcrResult, BatchOcrProvider } from \"../types.js\"\n\n/** 배치 OCR 프롬프트 */\nconst BATCH_OCR_PROMPT =\n  \"다음 문서 페이지 이미지들을 OCR하여 순수 Markdown으로 변환하세요.\\n\\n\" +\n  \"규칙:\\n\" +\n  \"- 각 페이지 결과 사이에 반드시 이 구분자를 삽입: <!-- PAGE_BREAK -->\\n\" +\n  \"- 테이블은 Markdown 테이블 문법 사용 (| 구분, |---|---| 헤더 구분선 포함)\\n\" +\n  \"- 병합된 셀은 해당 위치에 내용 기재\\n\" +\n  \"- 헤딩은 글자 크기에 따라 ## ~ ###### 사용\\n\" +\n  \"- 리스트는 - 또는 1. 사용\\n\" +\n  \"- 이미지, 도형 등 비텍스트 요소는 무시\\n\" +\n  \"- 원문의 읽기 순서와 구조를 유지\\n\" +\n  \"- ```로 감싸지 말고 순수 Markdown만 출력\"\n\n/** 모드별 기본 배치 크기 (CLI 내부 타임아웃 + 실측 기반)\n *\n * gemini CLI: 10장 이상에서 AbortError 발생 (내부 타임아웃).\n * 5장 배치가 안정적으로 동작 확인 (35초/배치).\n * 299페이지 = 60배치 = 기존 299회 대비 80% 감소.\n */\nexport const DEFAULT_BATCH_SIZES: Record<string, number> = {\n  gemini: 5,\n  claude: 5,\n  codex: 10,\n}\n\n/** 임시 디렉토리 — gemini CLI는 cwd 하위 + gitignore 밖만 @참조 가능 */\nlet _batchTempDir: string | null = null\nfunction getBatchTempDir(): string {\n  if (!_batchTempDir) {\n    _batchTempDir = join(process.cwd(), \"_kordoc_ocr_tmp\")\n    mkdirSync(_batchTempDir, { recursive: true })\n  }\n  return _batchTempDir\n}\n\n/**\n * 배치 CLI 프로바이더 생성\n */\nexport function createBatchCliProvider(\n  mode: \"gemini\" | \"claude\" | \"codex\",\n  batchSize: number\n): BatchOcrProvider {\n  return {\n    __batch: true as const,\n    batchSize,\n    async processBatch(pages) {\n      const results = new Map<number, StructuredOcrResult>()\n      const tempDir = getBatchTempDir()\n      const tempFiles: string[] = []\n\n      try {\n        // 1. Write all page images to temp files\n        for (const { image, pageNum } of pages) {\n          const path = join(tempDir, `batch-p${pageNum}.png`)\n          writeFileSync(path, image)\n          tempFiles.push(path)\n        }\n\n        // 2. Call CLI with all file references (비동기 — 병렬 배치 실행 가능)\n        let output: string\n        if (mode === \"codex\") {\n          output = await callBatchCodexCli(tempFiles)\n        } else {\n          output = await callBatchCli(mode, tempFiles)\n        }\n\n        // 3. Parse response by PAGE_BREAK separator\n        const cleaned = stripCodeFence(output.trim())\n        const parts = cleaned.split(/<!--\\s*PAGE_BREAK\\s*-->/)\n          .map(p => p.trim())\n          .filter(p => p.length > 0)\n\n        // 4. Map results to page numbers (best-effort if count mismatch)\n        for (let i = 0; i < pages.length; i++) {\n          const pageNum = pages[i].pageNum\n          if (i < parts.length) {\n            results.set(pageNum, { markdown: parts[i] })\n          }\n          // If fewer parts than pages, remaining pages get no result\n        }\n      } finally {\n        // 5. Clean up temp files\n        for (const f of tempFiles) {\n          try { unlinkSync(f) } catch { /* ignore */ }\n        }\n      }\n\n      return results\n    },\n  }\n}\n\n/**\n * 비동기 CLI 실행 헬퍼 — spawn + Promise 래핑.\n * spawnSync는 이벤트 루프를 차단하여 병렬 배치 실행 불가.\n */\nfunction spawnAsync(\n  cmd: string,\n  args: string[],\n  opts: { timeoutMs: number; cwd?: string; stdin?: string }\n): Promise<{ stdout: string; stderr: string; exitCode: number }> {\n  return new Promise((resolve, reject) => {\n    const child = spawn(cmd, args, {\n      cwd: opts.cwd,\n      env: process.env,\n      stdio: [\"pipe\", \"pipe\", \"pipe\"],\n    })\n\n    let stdout = \"\"\n    let stderr = \"\"\n    let killed = false\n\n    child.stdout.setEncoding(\"utf-8\")\n    child.stderr.setEncoding(\"utf-8\")\n    child.stdout.on(\"data\", (d: string) => { stdout += d })\n    child.stderr.on(\"data\", (d: string) => { stderr += d })\n\n    const timer = setTimeout(() => {\n      killed = true\n      child.kill(\"SIGTERM\")\n    }, opts.timeoutMs)\n\n    if (opts.stdin !== undefined) {\n      child.stdin.end(opts.stdin)\n    } else {\n      child.stdin.end()\n    }\n\n    child.on(\"close\", (code) => {\n      clearTimeout(timer)\n      if (killed) {\n        reject(new Error(`타임아웃 (${Math.round(opts.timeoutMs / 1000)}초)`))\n      } else {\n        resolve({ stdout, stderr, exitCode: code ?? 1 })\n      }\n    })\n    child.on(\"error\", (err) => {\n      clearTimeout(timer)\n      reject(err)\n    })\n  })\n}\n\n/** gemini/claude 배치 호출 (비동기) */\nasync function callBatchCli(mode: \"gemini\" | \"claude\", imagePaths: string[]): Promise<string> {\n  const fileRefs = imagePaths.map(p => `@${p}`).join(\"\\n\")\n  const prompt = `${BATCH_OCR_PROMPT}\\n\\n${fileRefs}`\n\n  let args: string[]\n  if (mode === \"gemini\") {\n    args = [\"--prompt\", prompt, \"--yolo\"]\n    const model = process.env.KORDOC_GEMINI_MODEL\n    if (model) args.push(\"--model\", model)\n  } else {\n    args = [\"--print\", prompt]\n    const model = process.env.KORDOC_CLAUDE_MODEL\n    if (model) args.push(\"--model\", model)\n  }\n\n  const timeoutMs = 60_000 + imagePaths.length * 20_000\n  const result = await spawnAsync(mode, args, {\n    timeoutMs,\n    ...(mode === \"claude\" ? { cwd: tmpdir() } : {}),\n  })\n\n  if (result.exitCode !== 0) {\n    const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`\n    throw new Error(`${mode} 배치 OCR 실패: ${errMsg}`)\n  }\n\n  return result.stdout || \"\"\n}\n\n/** codex 배치 호출 (비동기) — --image를 여러 번 지정 */\nasync function callBatchCodexCli(imagePaths: string[]): Promise<string> {\n  const outPath = join(tmpdir(), `kordoc-codex-batch-${Date.now()}-${Math.random().toString(36).slice(2)}.txt`)\n  try {\n    const args = [\"exec\", BATCH_OCR_PROMPT]\n    for (const p of imagePaths) {\n      args.push(\"--image\", p)\n    }\n    args.push(\"--output-last-message\", outPath)\n    const model = process.env.KORDOC_CODEX_MODEL\n    if (model) args.push(\"--model\", model)\n\n    const timeoutMs = 60_000 + imagePaths.length * 20_000\n    const result = await spawnAsync(\"codex\", args, {\n      timeoutMs,\n      stdin: \"\",\n    })\n\n    if (result.exitCode !== 0) {\n      const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`\n      throw new Error(`codex 배치 OCR 실패: ${errMsg}`)\n    }\n\n    try {\n      return readFileSync(outPath, \"utf-8\")\n    } catch {\n      return result.stdout || \"\"\n    }\n  } finally {\n    try { unlinkSync(outPath) } catch { /* ignore */ }\n  }\n}\n\n/** LLM 출력에서 코드 펜스 제거 (cli-provider.ts와 동일 로직) */\nfunction stripCodeFence(text: string): string {\n  const match = text.match(/^```(?:markdown|md)?\\s*\\n([\\s\\S]*?)\\n```\\s*$/m)\n  return match ? match[1].trim() : text\n}\n"],"mappings":";;;;AAWA,SAAS,aAAa;AACtB,SAAS,eAAe,cAAc,YAAY,iBAAiB;AACnE,SAAS,YAAY;AACrB,SAAS,cAAc;AAIvB,IAAM,mBACJ;AAiBK,IAAM,sBAA8C;AAAA,EACzD,QAAQ;AAAA,EACR,QAAQ;AAAA,EACR,OAAO;AACT;AAGA,IAAI,gBAA+B;AACnC,SAAS,kBAA0B;AACjC,MAAI,CAAC,eAAe;AAClB,oBAAgB,KAAK,QAAQ,IAAI,GAAG,iBAAiB;AACrD,cAAU,eAAe,EAAE,WAAW,KAAK,CAAC;AAAA,EAC9C;AACA,SAAO;AACT;AAKO,SAAS,uBACd,MACA,WACkB;AAClB,SAAO;AAAA,IACL,SAAS;AAAA,IACT;AAAA,IACA,MAAM,aAAa,OAAO;AACxB,YAAM,UAAU,oBAAI,IAAiC;AACrD,YAAM,UAAU,gBAAgB;AAChC,YAAM,YAAsB,CAAC;AAE7B,UAAI;AAEF,mBAAW,EAAE,OAAO,QAAQ,KAAK,OAAO;AACtC,gBAAM,OAAO,KAAK,SAAS,UAAU,OAAO,MAAM;AAClD,wBAAc,MAAM,KAAK;AACzB,oBAAU,KAAK,IAAI;AAAA,QACrB;AAGA,YAAI;AACJ,YAAI,SAAS,SAAS;AACpB,mBAAS,MAAM,kBAAkB,SAAS;AAAA,QAC5C,OAAO;AACL,mBAAS,MAAM,aAAa,MAAM,SAAS;AAAA,QAC7C;AAGA,cAAM,UAAU,eAAe,OAAO,KAAK,CAAC;AAC5C,cAAM,QAAQ,QAAQ,MAAM,yBAAyB,EAClD,IAAI,OAAK,EAAE,KAAK,CAAC,EACjB,OAAO,OAAK,EAAE,SAAS,CAAC;AAG3B,iBAAS,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;AACrC,gBAAM,UAAU,MAAM,CAAC,EAAE;AACzB,cAAI,IAAI,MAAM,QAAQ;AACpB,oBAAQ,IAAI,SAAS,EAAE,UAAU,MAAM,CAAC,EAAE,CAAC;AAAA,UAC7C;AAAA,QAEF;AAAA,MACF,UAAE;AAEA,mBAAW,KAAK,WAAW;AACzB,cAAI;AAAE,uBAAW,CAAC;AAAA,UAAE,QAAQ;AAAA,UAAe;AAAA,QAC7C;AAAA,MACF;AAEA,aAAO;AAAA,IACT;AAAA,EACF;AACF;AAMA,SAAS,WACP,KACA,MACA,MAC+D;AAC/D,SAAO,IAAI,QAAQ,CAAC,SAAS,WAAW;AACtC,UAAM,QAAQ,MAAM,KAAK,MAAM;AAAA,MAC7B,KAAK,KAAK;AAAA,MACV,KAAK,QAAQ;AAAA,MACb,OAAO,CAAC,QAAQ,QAAQ,MAAM;AAAA,IAChC,CAAC;AAED,QAAI,SAAS;AACb,QAAI,SAAS;AACb,QAAI,SAAS;AAEb,UAAM,OAAO,YAAY,OAAO;AAChC,UAAM,OAAO,YAAY,OAAO;AAChC,UAAM,OAAO,GAAG,QAAQ,CAAC,MAAc;AAAE,gBAAU;AAAA,IAAE,CAAC;AACtD,UAAM,OAAO,GAAG,QAAQ,CAAC,MAAc;AAAE,gBAAU;AAAA,IAAE,CAAC;AAEtD,UAAM,QAAQ,WAAW,MAAM;AAC7B,eAAS;AACT,YAAM,KAAK,SAAS;AAAA,IACtB,GAAG,KAAK,SAAS;AAEjB,QAAI,KAAK,UAAU,QAAW;AAC5B,YAAM,MAAM,IAAI,KAAK,KAAK;AAAA,IAC5B,OAAO;AACL,YAAM,MAAM,IAAI;AAAA,IAClB;AAEA,UAAM,GAAG,SAAS,CAAC,SAAS;AAC1B,mBAAa,KAAK;AAClB,UAAI,QAAQ;AACV,eAAO,IAAI,MAAM,6BAAS,KAAK,MAAM,KAAK,YAAY,GAAI,CAAC,SAAI,CAAC;AAAA,MAClE,OAAO;AACL,gBAAQ,EAAE,QAAQ,QAAQ,UAAU,QAAQ,EAAE,CAAC;AAAA,MACjD;AAAA,IACF,CAAC;AACD,UAAM,GAAG,SAAS,CAAC,QAAQ;AACzB,mBAAa,KAAK;AAClB,aAAO,GAAG;AAAA,IACZ,CAAC;AAAA,EACH,CAAC;AACH;AAGA,eAAe,aAAa,MAA2B,YAAuC;AAC5F,QAAM,WAAW,WAAW,IAAI,OAAK,IAAI,CAAC,EAAE,EAAE,KAAK,IAAI;AACvD,QAAM,SAAS,GAAG,gBAAgB;AAAA;AAAA,EAAO,QAAQ;AAEjD,MAAI;AACJ,MAAI,SAAS,UAAU;AACrB,WAAO,CAAC,YAAY,QAAQ,QAAQ;AACpC,UAAM,QAAQ,QAAQ,IAAI;AAC1B,QAAI,MAAO,MAAK,KAAK,WAAW,KAAK;AAAA,EACvC,OAAO;AACL,WAAO,CAAC,WAAW,MAAM;AACzB,UAAM,QAAQ,QAAQ,IAAI;AAC1B,QAAI,MAAO,MAAK,KAAK,WAAW,KAAK;AAAA,EACvC;AAEA,QAAM,YAAY,MAAS,WAAW,SAAS;AAC/C,QAAM,SAAS,MAAM,WAAW,MAAM,MAAM;AAAA,IAC1C;AAAA,IACA,GAAI,SAAS,WAAW,EAAE,KAAK,OAAO,EAAE,IAAI,CAAC;AAAA,EAC/C,CAAC;AAED,MAAI,OAAO,aAAa,GAAG;AACzB,UAAM,SAAS,OAAO,QAAQ,KAAK,KAAK,aAAa,OAAO,QAAQ;AACpE,UAAM,IAAI,MAAM,GAAG,IAAI,mCAAe,MAAM,EAAE;AAAA,EAChD;AAEA,SAAO,OAAO,UAAU;AAC1B;AAGA,eAAe,kBAAkB,YAAuC;AACtE,QAAM,UAAU,KAAK,OAAO,GAAG,sBAAsB,KAAK,IAAI,CAAC,IAAI,KAAK,OAAO,EAAE,SAAS,EAAE,EAAE,MAAM,CAAC,CAAC,MAAM;AAC5G,MAAI;AACF,UAAM,OAAO,CAAC,QAAQ,gBAAgB;AACtC,eAAW,KAAK,YAAY;AAC1B,WAAK,KAAK,WAAW,CAAC;AAAA,IACxB;AACA,SAAK,KAAK,yBAAyB,OAAO;AAC1C,UAAM,QAAQ,QAAQ,IAAI;AAC1B,QAAI,MAAO,MAAK,KAAK,WAAW,KAAK;AAErC,UAAM,YAAY,MAAS,WAAW,SAAS;AAC/C,UAAM,SAAS,MAAM,WAAW,SAAS,MAAM;AAAA,MAC7C;AAAA,MACA,OAAO;AAAA,IACT,CAAC;AAED,QAAI,OAAO,aAAa,GAAG;AACzB,YAAM,SAAS,OAAO,QAAQ,KAAK,KAAK,aAAa,OAAO,QAAQ;AACpE,YAAM,IAAI,MAAM,wCAAoB,MAAM,EAAE;AAAA,IAC9C;AAEA,QAAI;AACF,aAAO,aAAa,SAAS,OAAO;AAAA,IACtC,QAAQ;AACN,aAAO,OAAO,UAAU;AAAA,IAC1B;AAAA,EACF,UAAE;AACA,QAAI;AAAE,iBAAW,OAAO;AAAA,IAAE,QAAQ;AAAA,IAAe;AAAA,EACnD;AACF;AAGA,SAAS,eAAe,MAAsB;AAC5C,QAAM,QAAQ,KAAK,MAAM,+CAA+C;AACxE,SAAO,QAAQ,MAAM,CAAC,EAAE,KAAK,IAAI;AACnC;","names":[]}

package/dist/{chunk-WWILSVMJ.js → chunk-W5KUC23B.js}

@@ -1,7 +1,7 @@
 #!/usr/bin/env node
 
 // src/utils.ts
-var VERSION = true ? "2.3.0" : "0.0.0-dev";
+var VERSION = true ? "2.3.1" : "0.0.0-dev";
 function toArrayBuffer(buf) {
   if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
     return buf.buffer;
@@ -90,4 +90,4 @@ export {
   sanitizeHref,
   classifyError
 };
-//# sourceMappingURL=chunk-WWILSVMJ.js.map
+//# sourceMappingURL=chunk-W5KUC23B.js.map

package/dist/{chunk-2ZGLFZCN.js → chunk-ZOEUKD77.js}

@@ -6,7 +6,7 @@ import {
   precheckZipSize,
   sanitizeHref,
   toArrayBuffer
-} from "./chunk-WWILSVMJ.js";
+} from "./chunk-W5KUC23B.js";
 import {
   parsePageRange
 } from "./chunk-MOL7MDBG.js";
@@ -5406,7 +5406,7 @@ async function parsePdfDocument(buffer, options) {
       const ocrMode = options?.ocrMode;
       if (!ocrProvider && ocrMode && ocrMode !== "off") {
         try {
-          const { resolveOcrProvider } = await import("./resolve-LBFYRHJI.js");
+          const { resolveOcrProvider } = await import("./resolve-4FSAQF2S.js");
           const concurrency = options?.ocrConcurrency ?? 1;
           const batchSize = options?.ocrBatchSize;
           ocrProvider = await resolveOcrProvider(ocrMode, warnings, concurrency, batchSize);
@@ -5422,7 +5422,7 @@ async function parsePdfDocument(buffer, options) {
       if (ocrProvider) {
         let ocrBlocks = [];
         try {
-          const { ocrPages } = await import("./provider-OBY3XFSZ.js");
+          const { ocrPages } = await import("./provider-WYHC4NHI.js");
           const concurrency = options?.ocrConcurrency ?? 1;
           ocrBlocks = await ocrPages(doc, ocrProvider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
         } catch {
@@ -9624,4 +9624,4 @@ export {
 cfb/cfb.js:
   (*! crc32.js (C) 2014-present SheetJS -- http://sheetjs.com *)
 */
-//# sourceMappingURL=chunk-2ZGLFZCN.js.map
+//# sourceMappingURL=chunk-ZOEUKD77.js.map

package/dist/cli.js

@@ -4,11 +4,11 @@ import {
   markdownToHwpx,
   markdownToXlsx,
   parse
-} from "./chunk-2ZGLFZCN.js";
+} from "./chunk-ZOEUKD77.js";
 import {
   VERSION,
   toArrayBuffer
-} from "./chunk-WWILSVMJ.js";
+} from "./chunk-W5KUC23B.js";
 import "./chunk-MOL7MDBG.js";
 import "./chunk-ZWE3DS7E.js";
 
@@ -66,6 +66,8 @@ async function runParse(files, opts) {
         if (n > 0) parseOptions.ocrConcurrency = n;
       } else if (parseOptions.ocrMode === "tesseract") {
         parseOptions.ocrConcurrency = cpus().length;
+      } else {
+        parseOptions.ocrConcurrency = 4;
       }
       if (opts.ocrBatchSize) {
         const n = parseInt(opts.ocrBatchSize, 10);
@@ -132,7 +134,7 @@ async function runParse(files, opts) {
         saveImages(absPath);
       }
     } catch (err) {
-      const { sanitizeError } = await import("./utils-QAK24RJS.js");
+      const { sanitizeError } = await import("./utils-HSF5HI5T.js");
       process.stderr.write(`
 [kordoc] ERROR: ${fileName} \u2014 ${sanitizeError(err)}
 `);
@@ -216,7 +218,7 @@ program.command("convert <input>").description("\uB9C8\uD06C\uB2E4\uC6B4 \uD30C\
 `));
     }
   } catch (err) {
-    const { sanitizeError } = await import("./utils-QAK24RJS.js");
+    const { sanitizeError } = await import("./utils-HSF5HI5T.js");
     process.stderr.write(` FAIL
 `);
     process.stderr.write(`  \u2192 ${sanitizeError(err)}
@@ -225,7 +227,7 @@ program.command("convert <input>").description("\uB9C8\uD06C\uB2E4\uC6B4 \uD30C\
   }
 });
 program.command("watch <dir>").description("\uB514\uB809\uD1A0\uB9AC \uAC10\uC2DC \u2014 \uC0C8 \uBB38\uC11C \uC790\uB3D9 \uBCC0\uD658").option("--webhook <url>", "\uACB0\uACFC \uC804\uC1A1 \uC6F9\uD6C5 URL").option("-d, --out-dir <dir>", "\uBCC0\uD658 \uACB0\uACFC \uCD9C\uB825 \uB514\uB809\uD1A0\uB9AC").option("-p, --pages <range>", "\uD398\uC774\uC9C0/\uC139\uC158 \uBC94\uC704").option("--format <type>", "\uCD9C\uB825 \uD615\uC2DD: markdown \uB610\uB294 json", "markdown").option("--silent", "\uC9C4\uD589 \uBA54\uC2DC\uC9C0 \uC228\uAE30\uAE30").action(async (dir, opts) => {
-  const { watchDirectory } = await import("./watch-MPHX3QIH.js");
+  const { watchDirectory } = await import("./watch-R2JHXDGF.js");
   await watchDirectory({
     dir,
     outDir: opts.outDir,

package/dist/cli.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/cli.ts"],"sourcesContent":["/** kordoc CLI — 모두 파싱해버리겠다 */\n\nimport { readFileSync, writeFileSync, mkdirSync, statSync, existsSync, readdirSync } from \"fs\"\nimport { basename, resolve, extname } from \"path\"\nimport { cpus } from \"os\"\nimport { Command } from \"commander\"\nimport { parse, detectFormat, markdownToHwpx, markdownToXlsx } from \"./index.js\"\nimport type { ParseOptions, OcrMode } from \"./types.js\"\nimport { VERSION, toArrayBuffer } from \"./utils.js\"\n\nconst program = new Command()\n\n/** 공통 parse 옵션 타입 */\ninterface ParseOpts {\n  output?: string\n  outDir?: string\n  pages?: string\n  format: string\n  headerFooter: boolean\n  imageDir?: string\n  silent?: boolean\n  ocr?: string\n  ocrJobs?: string\n  ocrBatchSize?: string\n}\n\n/** parse 액션 공통 구현 — 루트 커맨드와 `parse` 서브커맨드가 공유 */\nasync function runParse(files: string[], opts: ParseOpts) {\n  const validFormats = [\"markdown\", \"json\"]\n  if (!validFormats.includes(opts.format)) {\n    process.stderr.write(`[kordoc] 지원하지 않는 형식: ${opts.format} (markdown 또는 json)\\n`)\n    process.exit(1)\n  }\n  for (let fi = 0; fi < files.length; fi++) {\n    const filePath = files[fi]\n    const absPath = resolve(filePath)\n    const fileName = basename(absPath)\n    const filePrefix = files.length > 1 ? `[${fi + 1}/${files.length}] ` : \"\"\n\n    try {\n      const fileSize = statSync(absPath).size\n      if (fileSize > 500 * 1024 * 1024) {\n        process.stderr.write(`\\n[kordoc] SKIP: ${fileName} — 파일이 너무 큽니다 (${(fileSize / 1024 / 1024).toFixed(1)}MB)\\n`)\n        process.exitCode = 1\n        continue\n      }\n      const buffer = readFileSync(absPath)\n      const arrayBuffer = toArrayBuffer(buffer)\n      const format = detectFormat(arrayBuffer)\n\n      if (!opts.silent) {\n        process.stderr.write(`[kordoc] ${filePrefix}${fileName} (${format}) ...`)\n      }\n\n      const parseOptions: ParseOptions = {}\n      if (opts.pages) parseOptions.pages = opts.pages as string\n      if (opts.headerFooter === false) parseOptions.removeHeaderFooter = false\n\n      // OCR 모드: CLI 기본값 \"auto\" (라이브러리 API는 undefined 유지)\n      const validOcrModes = [\"auto\", \"gemini\", \"claude\", \"codex\", \"ollama\", \"tesseract\", \"off\"]\n      if (opts.ocr) {\n        if (!validOcrModes.includes(opts.ocr)) {\n          process.stderr.write(`[kordoc] 지원하지 않는 OCR 모드: ${opts.ocr}\\n`)\n          process.stderr.write(`  사용 가능: ${validOcrModes.join(\", \")}\\n`)\n          process.exit(1)\n        }\n        parseOptions.ocrMode = opts.ocr as OcrMode\n      } else {\n        parseOptions.ocrMode = \"auto\"\n      }\n\n      // OCR 병렬 처리 수 (--ocr-jobs): tesseract 기본값은 CPU 코어 수\n      if (opts.ocrJobs) {\n        const n = parseInt(opts.ocrJobs, 10)\n        if (n > 0) parseOptions.ocrConcurrency = n\n      } else if (parseOptions.ocrMode === \"tesseract\") {\n        parseOptions.ocrConcurrency = cpus().length\n      }\n\n      // OCR 배치 크기 (--ocr-batch-size)\n      if (opts.ocrBatchSize) {\n        const n = parseInt(opts.ocrBatchSize, 10)\n        if (n > 0) parseOptions.ocrBatchSize = n\n      }\n\n      if (!opts.silent) {\n        parseOptions.onProgress = (current: number, total: number) => {\n          process.stderr.write(`\\r[kordoc] ${filePrefix}${fileName} (${format}) [${current}/${total}]`)\n        }\n      }\n      const result = await parse(arrayBuffer, parseOptions)\n\n      if (!result.success) {\n        process.stderr.write(` FAIL\\n`)\n        process.stderr.write(`  → ${result.error}\\n`)\n        process.exitCode = 1\n        continue\n      }\n\n      if (!opts.silent) process.stderr.write(` OK\\n`)\n\n      // 이미지 기반 PDF OCR 결과 표시\n      if (!opts.silent && result.success && result.isImageBased) {\n        process.stderr.write(`  → 이미지 기반 PDF — OCR 처리됨\\n`)\n      }\n\n      // 경고 표시\n      if (!opts.silent && result.success && result.warnings?.length) {\n        for (const w of result.warnings) {\n          process.stderr.write(`  ⚠ ${w.message}\\n`)\n        }\n      }\n\n      let markdown = result.markdown\n      // --out-dir 시 이미지 참조 경로에 images/ 접두사 추가\n      if (opts.outDir && result.images?.length) {\n        markdown = markdown.replace(/!\\[image\\]\\(image_/g, \"![image](images/image_\")\n      }\n      const output = opts.format === \"json\"\n        ? JSON.stringify(result, null, 2)\n        : markdown\n\n      // 이미지 저장: 출력 MD 파일 기준 폴더 사용 (convert와 일치)\n      const saveImages = (outFilePath: string) => {\n        if (!result.images?.length) return\n        const stem = basename(outFilePath).replace(/\\.[^.]+$/, \"\")\n        const defaultDir = resolve(outFilePath, \"..\", stem + \"_images\")\n        const imgDir = opts.imageDir ? resolve(opts.imageDir) : defaultDir\n        mkdirSync(imgDir, { recursive: true })\n        for (const img of result.images) {\n          writeFileSync(resolve(imgDir, img.filename), img.data)\n        }\n        if (!opts.silent) process.stderr.write(`  → ${result.images.length}개 이미지 → ${imgDir}\\n`)\n      }\n\n      if (opts.output && files.length === 1) {\n        writeFileSync(opts.output, output, \"utf-8\")\n        if (!opts.silent) process.stderr.write(`  → ${opts.output}\\n`)\n        saveImages(resolve(opts.output))\n      } else if (opts.outDir) {\n        mkdirSync(opts.outDir, { recursive: true })\n        const outExt = opts.format === \"json\" ? \".json\" : \".md\"\n        const outPath = resolve(opts.outDir, fileName.replace(/\\.[^.]+$/, outExt))\n        writeFileSync(outPath, output, \"utf-8\")\n        if (!opts.silent) process.stderr.write(`  → ${outPath}\\n`)\n        saveImages(outPath)\n      } else {\n        process.stdout.write(output + \"\\n\")\n        saveImages(absPath)  // stdout 출력 시 입력 파일 기준\n      }\n    } catch (err) {\n      const { sanitizeError } = await import(\"./utils.js\")\n      process.stderr.write(`\\n[kordoc] ERROR: ${fileName} — ${sanitizeError(err)}\\n`)\n      process.exitCode = 1\n    }\n  }\n}\n\n/** 공통 parse 옵션 등록 헬퍼 */\nfunction addParseOptions(cmd: Command): Command {\n  return cmd\n    .option(\"-o, --output <path>\", \"출력 파일 경로 (단일 파일 시)\")\n    .option(\"-d, --out-dir <dir>\", \"출력 디렉토리 (다중 파일 시)\")\n    .option(\"-p, --pages <range>\", \"페이지/섹션 범위 (예: 1-3, 1,3,5)\")\n    .option(\"--format <type>\", \"출력 형식: markdown (기본) 또는 json\", \"markdown\")\n    .option(\"--no-header-footer\", \"PDF 머리글/바닥글 자동 제거\")\n    .option(\"--image-dir <dir>\", \"이미지 저장 폴더 (기본: 입력 파일명_images 폴더)\")\n    .option(\"--silent\", \"진행 메시지 숨기기\")\n    .option(\"--ocr <mode>\", \"OCR 모드: auto(기본), gemini, claude, codex, ollama, tesseract, off\")\n    .option(\"--ocr-jobs <n>\", \"OCR 병렬 처리 수 (기본: CPU 코어 수, tesseract 전용)\")\n    .option(\"--ocr-batch-size <n>\", \"OCR 배치 크기 — CLI당 페이지 수 (기본: gemini/claude 50, codex 100)\")\n}\n\nprogram\n  .enablePositionalOptions()\n  .name(\"kordoc\")\n  .description(\"모두 파싱해버리겠다 — HWP, HWPX, PDF, XLSX, DOCX → Markdown\")\n  .version(VERSION)\n\n// `kordoc parse <files>` 서브커맨드 (권장)\naddParseOptions(\n  program\n    .command(\"parse\")\n    .description(\"파일을 마크다운으로 파싱 (HWP, HWPX, PDF, XLSX, DOCX)\")\n    .argument(\"<files...>\", \"변환할 파일 경로\")\n).action(runParse)\n\n// `kordoc <files>` 루트 커맨드 (하위 호환)\naddParseOptions(\n  program\n    .argument(\"<files...>\", \"변환할 파일 경로 (HWP, HWPX, PDF, XLSX, DOCX)\")\n).action(runParse)\n\nprogram\n  .command(\"convert <input>\")\n  .description(\"마크다운 파일을 HWPX 또는 XLSX로 변환\")\n  .option(\"-f, --format <type>\", \"출력 포맷: hwpx | xlsx\", \"hwpx\")\n  .option(\"-o, --output <path>\", \"출력 파일 경로 (기본: 입력명 + 포맷 확장자)\")\n  .option(\"--image-dir <dir>\", \"이미지 폴더 경로 (기본: 입력 MD 파일명_images 폴더)\")\n  .option(\"--images\", \"이미지 포함 (기본: 생략, 레이아웃 문제 방지)\")\n  .option(\"--template <path>\", \"HWPX 템플릿 파일 경로 (hwpx 전용)\")\n  .option(\"--silent\", \"진행 메시지 숨기기\")\n  .action(async (input: string, opts) => {\n    const validFormats = [\"hwpx\", \"xlsx\"]\n    if (!validFormats.includes(opts.format)) {\n      process.stderr.write(`[kordoc] 지원하지 않는 포맷: ${opts.format} (hwpx 또는 xlsx)\\n`)\n      process.exit(1)\n    }\n\n    const absInput = resolve(input)\n    if (!existsSync(absInput)) {\n      process.stderr.write(`[kordoc] 파일을 찾을 수 없습니다: ${input}\\n`)\n      process.exit(1)\n    }\n\n    const stem = basename(absInput).replace(/\\.[^.]+$/, \"\")\n    const outPath = opts.output\n      ? resolve(opts.output)\n      : resolve(absInput, \"..\", `${stem}.${opts.format}`)\n\n    if (!opts.silent) process.stderr.write(`[kordoc] ${basename(absInput)} → ${basename(outPath)} ...`)\n\n    try {\n      const markdown = readFileSync(absInput, \"utf-8\")\n\n      // 이미지 폴더에서 이미지 로드 (--images 플래그 필요)\n      const imgDir = opts.imageDir ? resolve(opts.imageDir) : resolve(absInput, \"..\", stem + \"_images\")\n      const images: import(\"./types.js\").ExtractedImage[] = []\n      if (opts.images && 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        if (!opts.silent) process.stderr.write(`  → 이미지 ${images.length}개 로드\\n`)\n      }\n\n      const warnings: string[] = []\n\n      let buf: ArrayBuffer\n      if (opts.format === \"xlsx\") {\n        if (opts.template && !opts.silent) {\n          process.stderr.write(`\\n[kordoc] 경고: --template은 hwpx 전용입니다. 무시됩니다.\\n`)\n        }\n        buf = await markdownToXlsx(markdown, { warnings, images: images.length ? images : undefined })\n      } else {\n        let templateArrayBuffer: ArrayBuffer | undefined\n        if (opts.template) {\n          const tmplBuf = readFileSync(resolve(opts.template))\n          templateArrayBuffer = tmplBuf.buffer.slice(tmplBuf.byteOffset, tmplBuf.byteOffset + tmplBuf.byteLength)\n        }\n        buf = await markdownToHwpx(markdown, { warnings, images: images.length ? images : undefined, templateArrayBuffer })\n      }\n\n      writeFileSync(outPath, Buffer.from(buf))\n\n      if (!opts.silent) {\n        process.stderr.write(` OK\\n`)\n        process.stderr.write(`  → ${outPath}\\n`)\n        if (warnings.length) warnings.forEach(w => process.stderr.write(`  ${w}\\n`))\n      }\n    } catch (err) {\n      const { sanitizeError } = await import(\"./utils.js\")\n      process.stderr.write(` FAIL\\n`)\n      process.stderr.write(`  → ${sanitizeError(err)}\\n`)\n      process.exit(1)\n    }\n  })\n\nprogram\n  .command(\"watch <dir>\")\n  .description(\"디렉토리 감시 — 새 문서 자동 변환\")\n  .option(\"--webhook <url>\", \"결과 전송 웹훅 URL\")\n  .option(\"-d, --out-dir <dir>\", \"변환 결과 출력 디렉토리\")\n  .option(\"-p, --pages <range>\", \"페이지/섹션 범위\")\n  .option(\"--format <type>\", \"출력 형식: markdown 또는 json\", \"markdown\")\n  .option(\"--silent\", \"진행 메시지 숨기기\")\n  .action(async (dir: string, opts) => {\n    const { watchDirectory } = await import(\"./watch.js\")\n    await watchDirectory({\n      dir,\n      outDir: opts.outDir,\n      webhook: opts.webhook,\n      format: opts.format,\n      pages: opts.pages,\n      silent: opts.silent,\n    })\n  })\n\nprogram.parse()\n"],"mappings":";;;;;;;;;;;;;;;AAEA,SAAS,cAAc,eAAe,WAAW,UAAU,YAAY,mBAAmB;AAC1F,SAAS,UAAU,SAAS,eAAe;AAC3C,SAAS,YAAY;AACrB,SAAS,eAAe;AAKxB,IAAM,UAAU,IAAI,QAAQ;AAiB5B,eAAe,SAAS,OAAiB,MAAiB;AACxD,QAAM,eAAe,CAAC,YAAY,MAAM;AACxC,MAAI,CAAC,aAAa,SAAS,KAAK,MAAM,GAAG;AACvC,YAAQ,OAAO,MAAM,gEAAwB,KAAK,MAAM;AAAA,CAAuB;AAC/E,YAAQ,KAAK,CAAC;AAAA,EAChB;AACA,WAAS,KAAK,GAAG,KAAK,MAAM,QAAQ,MAAM;AACxC,UAAM,WAAW,MAAM,EAAE;AACzB,UAAM,UAAU,QAAQ,QAAQ;AAChC,UAAM,WAAW,SAAS,OAAO;AACjC,UAAM,aAAa,MAAM,SAAS,IAAI,IAAI,KAAK,CAAC,IAAI,MAAM,MAAM,OAAO;AAEvE,QAAI;AACF,YAAM,WAAW,SAAS,OAAO,EAAE;AACnC,UAAI,WAAW,MAAM,OAAO,MAAM;AAChC,gBAAQ,OAAO,MAAM;AAAA,iBAAoB,QAAQ,gEAAmB,WAAW,OAAO,MAAM,QAAQ,CAAC,CAAC;AAAA,CAAO;AAC7G,gBAAQ,WAAW;AACnB;AAAA,MACF;AACA,YAAM,SAAS,aAAa,OAAO;AACnC,YAAM,cAAc,cAAc,MAAM;AACxC,YAAM,SAAS,aAAa,WAAW;AAEvC,UAAI,CAAC,KAAK,QAAQ;AAChB,gBAAQ,OAAO,MAAM,YAAY,UAAU,GAAG,QAAQ,KAAK,MAAM,OAAO;AAAA,MAC1E;AAEA,YAAM,eAA6B,CAAC;AACpC,UAAI,KAAK,MAAO,cAAa,QAAQ,KAAK;AAC1C,UAAI,KAAK,iBAAiB,MAAO,cAAa,qBAAqB;AAGnE,YAAM,gBAAgB,CAAC,QAAQ,UAAU,UAAU,SAAS,UAAU,aAAa,KAAK;AACxF,UAAI,KAAK,KAAK;AACZ,YAAI,CAAC,cAAc,SAAS,KAAK,GAAG,GAAG;AACrC,kBAAQ,OAAO,MAAM,oEAA4B,KAAK,GAAG;AAAA,CAAI;AAC7D,kBAAQ,OAAO,MAAM,gCAAY,cAAc,KAAK,IAAI,CAAC;AAAA,CAAI;AAC7D,kBAAQ,KAAK,CAAC;AAAA,QAChB;AACA,qBAAa,UAAU,KAAK;AAAA,MAC9B,OAAO;AACL,qBAAa,UAAU;AAAA,MACzB;AAGA,UAAI,KAAK,SAAS;AAChB,cAAM,IAAI,SAAS,KAAK,SAAS,EAAE;AACnC,YAAI,IAAI,EAAG,cAAa,iBAAiB;AAAA,MAC3C,WAAW,aAAa,YAAY,aAAa;AAC/C,qBAAa,iBAAiB,KAAK,EAAE;AAAA,MACvC;AAGA,UAAI,KAAK,cAAc;AACrB,cAAM,IAAI,SAAS,KAAK,cAAc,EAAE;AACxC,YAAI,IAAI,EAAG,cAAa,eAAe;AAAA,MACzC;AAEA,UAAI,CAAC,KAAK,QAAQ;AAChB,qBAAa,aAAa,CAAC,SAAiB,UAAkB;AAC5D,kBAAQ,OAAO,MAAM,cAAc,UAAU,GAAG,QAAQ,KAAK,MAAM,MAAM,OAAO,IAAI,KAAK,GAAG;AAAA,QAC9F;AAAA,MACF;AACA,YAAM,SAAS,MAAM,MAAM,aAAa,YAAY;AAEpD,UAAI,CAAC,OAAO,SAAS;AACnB,gBAAQ,OAAO,MAAM;AAAA,CAAS;AAC9B,gBAAQ,OAAO,MAAM,YAAO,OAAO,KAAK;AAAA,CAAI;AAC5C,gBAAQ,WAAW;AACnB;AAAA,MACF;AAEA,UAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM;AAAA,CAAO;AAG9C,UAAI,CAAC,KAAK,UAAU,OAAO,WAAW,OAAO,cAAc;AACzD,gBAAQ,OAAO,MAAM;AAAA,CAA4B;AAAA,MACnD;AAGA,UAAI,CAAC,KAAK,UAAU,OAAO,WAAW,OAAO,UAAU,QAAQ;AAC7D,mBAAW,KAAK,OAAO,UAAU;AAC/B,kBAAQ,OAAO,MAAM,YAAO,EAAE,OAAO;AAAA,CAAI;AAAA,QAC3C;AAAA,MACF;AAEA,UAAI,WAAW,OAAO;AAEtB,UAAI,KAAK,UAAU,OAAO,QAAQ,QAAQ;AACxC,mBAAW,SAAS,QAAQ,uBAAuB,wBAAwB;AAAA,MAC7E;AACA,YAAM,SAAS,KAAK,WAAW,SAC3B,KAAK,UAAU,QAAQ,MAAM,CAAC,IAC9B;AAGJ,YAAM,aAAa,CAAC,gBAAwB;AAC1C,YAAI,CAAC,OAAO,QAAQ,OAAQ;AAC5B,cAAM,OAAO,SAAS,WAAW,EAAE,QAAQ,YAAY,EAAE;AACzD,cAAM,aAAa,QAAQ,aAAa,MAAM,OAAO,SAAS;AAC9D,cAAM,SAAS,KAAK,WAAW,QAAQ,KAAK,QAAQ,IAAI;AACxD,kBAAU,QAAQ,EAAE,WAAW,KAAK,CAAC;AACrC,mBAAW,OAAO,OAAO,QAAQ;AAC/B,wBAAc,QAAQ,QAAQ,IAAI,QAAQ,GAAG,IAAI,IAAI;AAAA,QACvD;AACA,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,OAAO,OAAO,MAAM,oCAAW,MAAM;AAAA,CAAI;AAAA,MACzF;AAEA,UAAI,KAAK,UAAU,MAAM,WAAW,GAAG;AACrC,sBAAc,KAAK,QAAQ,QAAQ,OAAO;AAC1C,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,KAAK,MAAM;AAAA,CAAI;AAC7D,mBAAW,QAAQ,KAAK,MAAM,CAAC;AAAA,MACjC,WAAW,KAAK,QAAQ;AACtB,kBAAU,KAAK,QAAQ,EAAE,WAAW,KAAK,CAAC;AAC1C,cAAM,SAAS,KAAK,WAAW,SAAS,UAAU;AAClD,cAAM,UAAU,QAAQ,KAAK,QAAQ,SAAS,QAAQ,YAAY,MAAM,CAAC;AACzE,sBAAc,SAAS,QAAQ,OAAO;AACtC,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,OAAO;AAAA,CAAI;AACzD,mBAAW,OAAO;AAAA,MACpB,OAAO;AACL,gBAAQ,OAAO,MAAM,SAAS,IAAI;AAClC,mBAAW,OAAO;AAAA,MACpB;AAAA,IACF,SAAS,KAAK;AACZ,YAAM,EAAE,cAAc,IAAI,MAAM,OAAO,qBAAY;AACnD,cAAQ,OAAO,MAAM;AAAA,kBAAqB,QAAQ,WAAM,cAAc,GAAG,CAAC;AAAA,CAAI;AAC9E,cAAQ,WAAW;AAAA,IACrB;AAAA,EACF;AACF;AAGA,SAAS,gBAAgB,KAAuB;AAC9C,SAAO,IACJ,OAAO,uBAAuB,2EAAoB,EAClD,OAAO,uBAAuB,0EAAmB,EACjD,OAAO,uBAAuB,mEAA2B,EACzD,OAAO,mBAAmB,wEAAgC,UAAU,EACpE,OAAO,sBAAsB,qEAAmB,EAChD,OAAO,qBAAqB,kHAAkC,EAC9D,OAAO,YAAY,oDAAY,EAC/B,OAAO,gBAAgB,qFAAiE,EACxF,OAAO,kBAAkB,sGAA0C,EACnE,OAAO,wBAAwB,sHAA0D;AAC9F;AAEA,QACG,wBAAwB,EACxB,KAAK,QAAQ,EACb,YAAY,2GAAoD,EAChE,QAAQ,OAAO;AAGlB;AAAA,EACE,QACG,QAAQ,OAAO,EACf,YAAY,mGAA4C,EACxD,SAAS,cAAc,8CAAW;AACvC,EAAE,OAAO,QAAQ;AAGjB;AAAA,EACE,QACG,SAAS,cAAc,2EAAwC;AACpE,EAAE,OAAO,QAAQ;AAEjB,QACG,QAAQ,iBAAiB,EACzB,YAAY,uFAA2B,EACvC,OAAO,uBAAuB,0CAAsB,MAAM,EAC1D,OAAO,uBAAuB,6GAA6B,EAC3D,OAAO,qBAAqB,qHAAqC,EACjE,OAAO,YAAY,kHAA6B,EAChD,OAAO,qBAAqB,uEAA0B,EACtD,OAAO,YAAY,oDAAY,EAC/B,OAAO,OAAO,OAAe,SAAS;AACrC,QAAM,eAAe,CAAC,QAAQ,MAAM;AACpC,MAAI,CAAC,aAAa,SAAS,KAAK,MAAM,GAAG;AACvC,YAAQ,OAAO,MAAM,gEAAwB,KAAK,MAAM;AAAA,CAAmB;AAC3E,YAAQ,KAAK,CAAC;AAAA,EAChB;AAEA,QAAM,WAAW,QAAQ,KAAK;AAC9B,MAAI,CAAC,WAAW,QAAQ,GAAG;AACzB,YAAQ,OAAO,MAAM,6EAA2B,KAAK;AAAA,CAAI;AACzD,YAAQ,KAAK,CAAC;AAAA,EAChB;AAEA,QAAM,OAAO,SAAS,QAAQ,EAAE,QAAQ,YAAY,EAAE;AACtD,QAAM,UAAU,KAAK,SACjB,QAAQ,KAAK,MAAM,IACnB,QAAQ,UAAU,MAAM,GAAG,IAAI,IAAI,KAAK,MAAM,EAAE;AAEpD,MAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAY,SAAS,QAAQ,CAAC,WAAM,SAAS,OAAO,CAAC,MAAM;AAElG,MAAI;AACF,UAAM,WAAW,aAAa,UAAU,OAAO;AAG/C,UAAM,SAAS,KAAK,WAAW,QAAQ,KAAK,QAAQ,IAAI,QAAQ,UAAU,MAAM,OAAO,SAAS;AAChG,UAAM,SAAgD,CAAC;AACvD,QAAI,KAAK,UAAU,WAAW,MAAM,GAAG;AACrC,YAAM,UAAkC;AAAA,QACtC,KAAK;AAAA,QAAa,KAAK;AAAA,QAAc,MAAM;AAAA,QAC3C,KAAK;AAAA,QAAa,KAAK;AAAA,MACzB;AACA,iBAAW,SAAS,YAAY,QAAQ,EAAE,eAAe,KAAK,CAAC,GAAG;AAChE,YAAI,CAAC,MAAM,OAAO,EAAG;AACrB,cAAM,QAAQ,MAAM;AACpB,cAAM,MAAM,QAAQ,KAAK,EAAE,MAAM,CAAC,EAAE,YAAY;AAChD,YAAI,CAAC,QAAQ,GAAG,EAAG;AACnB,cAAM,OAAO,aAAa,QAAQ,QAAQ,KAAK,CAAC;AAChD,eAAO,KAAK,EAAE,UAAU,OAAO,MAAM,IAAI,WAAW,IAAI,GAAG,UAAU,QAAQ,GAAG,EAAE,CAAC;AAAA,MACrF;AACA,UAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,+BAAW,OAAO,MAAM;AAAA,CAAQ;AAAA,IACzE;AAEA,UAAM,WAAqB,CAAC;AAE5B,QAAI;AACJ,QAAI,KAAK,WAAW,QAAQ;AAC1B,UAAI,KAAK,YAAY,CAAC,KAAK,QAAQ;AACjC,gBAAQ,OAAO,MAAM;AAAA;AAAA,CAAiD;AAAA,MACxE;AACA,YAAM,MAAM,eAAe,UAAU,EAAE,UAAU,QAAQ,OAAO,SAAS,SAAS,OAAU,CAAC;AAAA,IAC/F,OAAO;AACL,UAAI;AACJ,UAAI,KAAK,UAAU;AACjB,cAAM,UAAU,aAAa,QAAQ,KAAK,QAAQ,CAAC;AACnD,8BAAsB,QAAQ,OAAO,MAAM,QAAQ,YAAY,QAAQ,aAAa,QAAQ,UAAU;AAAA,MACxG;AACA,YAAM,MAAM,eAAe,UAAU,EAAE,UAAU,QAAQ,OAAO,SAAS,SAAS,QAAW,oBAAoB,CAAC;AAAA,IACpH;AAEA,kBAAc,SAAS,OAAO,KAAK,GAAG,CAAC;AAEvC,QAAI,CAAC,KAAK,QAAQ;AAChB,cAAQ,OAAO,MAAM;AAAA,CAAO;AAC5B,cAAQ,OAAO,MAAM,YAAO,OAAO;AAAA,CAAI;AACvC,UAAI,SAAS,OAAQ,UAAS,QAAQ,OAAK,QAAQ,OAAO,MAAM,KAAK,CAAC;AAAA,CAAI,CAAC;AAAA,IAC7E;AAAA,EACF,SAAS,KAAK;AACZ,UAAM,EAAE,cAAc,IAAI,MAAM,OAAO,qBAAY;AACnD,YAAQ,OAAO,MAAM;AAAA,CAAS;AAC9B,YAAQ,OAAO,MAAM,YAAO,cAAc,GAAG,CAAC;AAAA,CAAI;AAClD,YAAQ,KAAK,CAAC;AAAA,EAChB;AACF,CAAC;AAEH,QACG,QAAQ,aAAa,EACrB,YAAY,4FAAsB,EAClC,OAAO,mBAAmB,4CAAc,EACxC,OAAO,uBAAuB,iEAAe,EAC7C,OAAO,uBAAuB,8CAAW,EACzC,OAAO,mBAAmB,yDAA2B,UAAU,EAC/D,OAAO,YAAY,oDAAY,EAC/B,OAAO,OAAO,KAAa,SAAS;AACnC,QAAM,EAAE,eAAe,IAAI,MAAM,OAAO,qBAAY;AACpD,QAAM,eAAe;AAAA,IACnB;AAAA,IACA,QAAQ,KAAK;AAAA,IACb,SAAS,KAAK;AAAA,IACd,QAAQ,KAAK;AAAA,IACb,OAAO,KAAK;AAAA,IACZ,QAAQ,KAAK;AAAA,EACf,CAAC;AACH,CAAC;AAEH,QAAQ,MAAM;","names":[]}
+{"version":3,"sources":["../src/cli.ts"],"sourcesContent":["/** kordoc CLI — 모두 파싱해버리겠다 */\n\nimport { readFileSync, writeFileSync, mkdirSync, statSync, existsSync, readdirSync } from \"fs\"\nimport { basename, resolve, extname } from \"path\"\nimport { cpus } from \"os\"\nimport { Command } from \"commander\"\nimport { parse, detectFormat, markdownToHwpx, markdownToXlsx } from \"./index.js\"\nimport type { ParseOptions, OcrMode } from \"./types.js\"\nimport { VERSION, toArrayBuffer } from \"./utils.js\"\n\nconst program = new Command()\n\n/** 공통 parse 옵션 타입 */\ninterface ParseOpts {\n  output?: string\n  outDir?: string\n  pages?: string\n  format: string\n  headerFooter: boolean\n  imageDir?: string\n  silent?: boolean\n  ocr?: string\n  ocrJobs?: string\n  ocrBatchSize?: string\n}\n\n/** parse 액션 공통 구현 — 루트 커맨드와 `parse` 서브커맨드가 공유 */\nasync function runParse(files: string[], opts: ParseOpts) {\n  const validFormats = [\"markdown\", \"json\"]\n  if (!validFormats.includes(opts.format)) {\n    process.stderr.write(`[kordoc] 지원하지 않는 형식: ${opts.format} (markdown 또는 json)\\n`)\n    process.exit(1)\n  }\n  for (let fi = 0; fi < files.length; fi++) {\n    const filePath = files[fi]\n    const absPath = resolve(filePath)\n    const fileName = basename(absPath)\n    const filePrefix = files.length > 1 ? `[${fi + 1}/${files.length}] ` : \"\"\n\n    try {\n      const fileSize = statSync(absPath).size\n      if (fileSize > 500 * 1024 * 1024) {\n        process.stderr.write(`\\n[kordoc] SKIP: ${fileName} — 파일이 너무 큽니다 (${(fileSize / 1024 / 1024).toFixed(1)}MB)\\n`)\n        process.exitCode = 1\n        continue\n      }\n      const buffer = readFileSync(absPath)\n      const arrayBuffer = toArrayBuffer(buffer)\n      const format = detectFormat(arrayBuffer)\n\n      if (!opts.silent) {\n        process.stderr.write(`[kordoc] ${filePrefix}${fileName} (${format}) ...`)\n      }\n\n      const parseOptions: ParseOptions = {}\n      if (opts.pages) parseOptions.pages = opts.pages as string\n      if (opts.headerFooter === false) parseOptions.removeHeaderFooter = false\n\n      // OCR 모드: CLI 기본값 \"auto\" (라이브러리 API는 undefined 유지)\n      const validOcrModes = [\"auto\", \"gemini\", \"claude\", \"codex\", \"ollama\", \"tesseract\", \"off\"]\n      if (opts.ocr) {\n        if (!validOcrModes.includes(opts.ocr)) {\n          process.stderr.write(`[kordoc] 지원하지 않는 OCR 모드: ${opts.ocr}\\n`)\n          process.stderr.write(`  사용 가능: ${validOcrModes.join(\", \")}\\n`)\n          process.exit(1)\n        }\n        parseOptions.ocrMode = opts.ocr as OcrMode\n      } else {\n        parseOptions.ocrMode = \"auto\"\n      }\n\n      // OCR 병렬 처리 수 (--ocr-jobs)\n      // - tesseract: CPU 코어 수 (워커 스레드 병렬)\n      // - gemini/claude/codex/auto: 4 (배치 병렬 실행)\n      if (opts.ocrJobs) {\n        const n = parseInt(opts.ocrJobs, 10)\n        if (n > 0) parseOptions.ocrConcurrency = n\n      } else if (parseOptions.ocrMode === \"tesseract\") {\n        parseOptions.ocrConcurrency = cpus().length\n      } else {\n        parseOptions.ocrConcurrency = 4\n      }\n\n      // OCR 배치 크기 (--ocr-batch-size)\n      if (opts.ocrBatchSize) {\n        const n = parseInt(opts.ocrBatchSize, 10)\n        if (n > 0) parseOptions.ocrBatchSize = n\n      }\n\n      if (!opts.silent) {\n        parseOptions.onProgress = (current: number, total: number) => {\n          process.stderr.write(`\\r[kordoc] ${filePrefix}${fileName} (${format}) [${current}/${total}]`)\n        }\n      }\n      const result = await parse(arrayBuffer, parseOptions)\n\n      if (!result.success) {\n        process.stderr.write(` FAIL\\n`)\n        process.stderr.write(`  → ${result.error}\\n`)\n        process.exitCode = 1\n        continue\n      }\n\n      if (!opts.silent) process.stderr.write(` OK\\n`)\n\n      // 이미지 기반 PDF OCR 결과 표시\n      if (!opts.silent && result.success && result.isImageBased) {\n        process.stderr.write(`  → 이미지 기반 PDF — OCR 처리됨\\n`)\n      }\n\n      // 경고 표시\n      if (!opts.silent && result.success && result.warnings?.length) {\n        for (const w of result.warnings) {\n          process.stderr.write(`  ⚠ ${w.message}\\n`)\n        }\n      }\n\n      let markdown = result.markdown\n      // --out-dir 시 이미지 참조 경로에 images/ 접두사 추가\n      if (opts.outDir && result.images?.length) {\n        markdown = markdown.replace(/!\\[image\\]\\(image_/g, \"![image](images/image_\")\n      }\n      const output = opts.format === \"json\"\n        ? JSON.stringify(result, null, 2)\n        : markdown\n\n      // 이미지 저장: 출력 MD 파일 기준 폴더 사용 (convert와 일치)\n      const saveImages = (outFilePath: string) => {\n        if (!result.images?.length) return\n        const stem = basename(outFilePath).replace(/\\.[^.]+$/, \"\")\n        const defaultDir = resolve(outFilePath, \"..\", stem + \"_images\")\n        const imgDir = opts.imageDir ? resolve(opts.imageDir) : defaultDir\n        mkdirSync(imgDir, { recursive: true })\n        for (const img of result.images) {\n          writeFileSync(resolve(imgDir, img.filename), img.data)\n        }\n        if (!opts.silent) process.stderr.write(`  → ${result.images.length}개 이미지 → ${imgDir}\\n`)\n      }\n\n      if (opts.output && files.length === 1) {\n        writeFileSync(opts.output, output, \"utf-8\")\n        if (!opts.silent) process.stderr.write(`  → ${opts.output}\\n`)\n        saveImages(resolve(opts.output))\n      } else if (opts.outDir) {\n        mkdirSync(opts.outDir, { recursive: true })\n        const outExt = opts.format === \"json\" ? \".json\" : \".md\"\n        const outPath = resolve(opts.outDir, fileName.replace(/\\.[^.]+$/, outExt))\n        writeFileSync(outPath, output, \"utf-8\")\n        if (!opts.silent) process.stderr.write(`  → ${outPath}\\n`)\n        saveImages(outPath)\n      } else {\n        process.stdout.write(output + \"\\n\")\n        saveImages(absPath)  // stdout 출력 시 입력 파일 기준\n      }\n    } catch (err) {\n      const { sanitizeError } = await import(\"./utils.js\")\n      process.stderr.write(`\\n[kordoc] ERROR: ${fileName} — ${sanitizeError(err)}\\n`)\n      process.exitCode = 1\n    }\n  }\n}\n\n/** 공통 parse 옵션 등록 헬퍼 */\nfunction addParseOptions(cmd: Command): Command {\n  return cmd\n    .option(\"-o, --output <path>\", \"출력 파일 경로 (단일 파일 시)\")\n    .option(\"-d, --out-dir <dir>\", \"출력 디렉토리 (다중 파일 시)\")\n    .option(\"-p, --pages <range>\", \"페이지/섹션 범위 (예: 1-3, 1,3,5)\")\n    .option(\"--format <type>\", \"출력 형식: markdown (기본) 또는 json\", \"markdown\")\n    .option(\"--no-header-footer\", \"PDF 머리글/바닥글 자동 제거\")\n    .option(\"--image-dir <dir>\", \"이미지 저장 폴더 (기본: 입력 파일명_images 폴더)\")\n    .option(\"--silent\", \"진행 메시지 숨기기\")\n    .option(\"--ocr <mode>\", \"OCR 모드: auto(기본), gemini, claude, codex, ollama, tesseract, off\")\n    .option(\"--ocr-jobs <n>\", \"OCR 병렬 처리 수 (기본: CPU 코어 수, tesseract 전용)\")\n    .option(\"--ocr-batch-size <n>\", \"OCR 배치 크기 — CLI당 페이지 수 (기본: gemini/claude 50, codex 100)\")\n}\n\nprogram\n  .enablePositionalOptions()\n  .name(\"kordoc\")\n  .description(\"모두 파싱해버리겠다 — HWP, HWPX, PDF, XLSX, DOCX → Markdown\")\n  .version(VERSION)\n\n// `kordoc parse <files>` 서브커맨드 (권장)\naddParseOptions(\n  program\n    .command(\"parse\")\n    .description(\"파일을 마크다운으로 파싱 (HWP, HWPX, PDF, XLSX, DOCX)\")\n    .argument(\"<files...>\", \"변환할 파일 경로\")\n).action(runParse)\n\n// `kordoc <files>` 루트 커맨드 (하위 호환)\naddParseOptions(\n  program\n    .argument(\"<files...>\", \"변환할 파일 경로 (HWP, HWPX, PDF, XLSX, DOCX)\")\n).action(runParse)\n\nprogram\n  .command(\"convert <input>\")\n  .description(\"마크다운 파일을 HWPX 또는 XLSX로 변환\")\n  .option(\"-f, --format <type>\", \"출력 포맷: hwpx | xlsx\", \"hwpx\")\n  .option(\"-o, --output <path>\", \"출력 파일 경로 (기본: 입력명 + 포맷 확장자)\")\n  .option(\"--image-dir <dir>\", \"이미지 폴더 경로 (기본: 입력 MD 파일명_images 폴더)\")\n  .option(\"--images\", \"이미지 포함 (기본: 생략, 레이아웃 문제 방지)\")\n  .option(\"--template <path>\", \"HWPX 템플릿 파일 경로 (hwpx 전용)\")\n  .option(\"--silent\", \"진행 메시지 숨기기\")\n  .action(async (input: string, opts) => {\n    const validFormats = [\"hwpx\", \"xlsx\"]\n    if (!validFormats.includes(opts.format)) {\n      process.stderr.write(`[kordoc] 지원하지 않는 포맷: ${opts.format} (hwpx 또는 xlsx)\\n`)\n      process.exit(1)\n    }\n\n    const absInput = resolve(input)\n    if (!existsSync(absInput)) {\n      process.stderr.write(`[kordoc] 파일을 찾을 수 없습니다: ${input}\\n`)\n      process.exit(1)\n    }\n\n    const stem = basename(absInput).replace(/\\.[^.]+$/, \"\")\n    const outPath = opts.output\n      ? resolve(opts.output)\n      : resolve(absInput, \"..\", `${stem}.${opts.format}`)\n\n    if (!opts.silent) process.stderr.write(`[kordoc] ${basename(absInput)} → ${basename(outPath)} ...`)\n\n    try {\n      const markdown = readFileSync(absInput, \"utf-8\")\n\n      // 이미지 폴더에서 이미지 로드 (--images 플래그 필요)\n      const imgDir = opts.imageDir ? resolve(opts.imageDir) : resolve(absInput, \"..\", stem + \"_images\")\n      const images: import(\"./types.js\").ExtractedImage[] = []\n      if (opts.images && 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        if (!opts.silent) process.stderr.write(`  → 이미지 ${images.length}개 로드\\n`)\n      }\n\n      const warnings: string[] = []\n\n      let buf: ArrayBuffer\n      if (opts.format === \"xlsx\") {\n        if (opts.template && !opts.silent) {\n          process.stderr.write(`\\n[kordoc] 경고: --template은 hwpx 전용입니다. 무시됩니다.\\n`)\n        }\n        buf = await markdownToXlsx(markdown, { warnings, images: images.length ? images : undefined })\n      } else {\n        let templateArrayBuffer: ArrayBuffer | undefined\n        if (opts.template) {\n          const tmplBuf = readFileSync(resolve(opts.template))\n          templateArrayBuffer = tmplBuf.buffer.slice(tmplBuf.byteOffset, tmplBuf.byteOffset + tmplBuf.byteLength)\n        }\n        buf = await markdownToHwpx(markdown, { warnings, images: images.length ? images : undefined, templateArrayBuffer })\n      }\n\n      writeFileSync(outPath, Buffer.from(buf))\n\n      if (!opts.silent) {\n        process.stderr.write(` OK\\n`)\n        process.stderr.write(`  → ${outPath}\\n`)\n        if (warnings.length) warnings.forEach(w => process.stderr.write(`  ${w}\\n`))\n      }\n    } catch (err) {\n      const { sanitizeError } = await import(\"./utils.js\")\n      process.stderr.write(` FAIL\\n`)\n      process.stderr.write(`  → ${sanitizeError(err)}\\n`)\n      process.exit(1)\n    }\n  })\n\nprogram\n  .command(\"watch <dir>\")\n  .description(\"디렉토리 감시 — 새 문서 자동 변환\")\n  .option(\"--webhook <url>\", \"결과 전송 웹훅 URL\")\n  .option(\"-d, --out-dir <dir>\", \"변환 결과 출력 디렉토리\")\n  .option(\"-p, --pages <range>\", \"페이지/섹션 범위\")\n  .option(\"--format <type>\", \"출력 형식: markdown 또는 json\", \"markdown\")\n  .option(\"--silent\", \"진행 메시지 숨기기\")\n  .action(async (dir: string, opts) => {\n    const { watchDirectory } = await import(\"./watch.js\")\n    await watchDirectory({\n      dir,\n      outDir: opts.outDir,\n      webhook: opts.webhook,\n      format: opts.format,\n      pages: opts.pages,\n      silent: opts.silent,\n    })\n  })\n\nprogram.parse()\n"],"mappings":";;;;;;;;;;;;;;;AAEA,SAAS,cAAc,eAAe,WAAW,UAAU,YAAY,mBAAmB;AAC1F,SAAS,UAAU,SAAS,eAAe;AAC3C,SAAS,YAAY;AACrB,SAAS,eAAe;AAKxB,IAAM,UAAU,IAAI,QAAQ;AAiB5B,eAAe,SAAS,OAAiB,MAAiB;AACxD,QAAM,eAAe,CAAC,YAAY,MAAM;AACxC,MAAI,CAAC,aAAa,SAAS,KAAK,MAAM,GAAG;AACvC,YAAQ,OAAO,MAAM,gEAAwB,KAAK,MAAM;AAAA,CAAuB;AAC/E,YAAQ,KAAK,CAAC;AAAA,EAChB;AACA,WAAS,KAAK,GAAG,KAAK,MAAM,QAAQ,MAAM;AACxC,UAAM,WAAW,MAAM,EAAE;AACzB,UAAM,UAAU,QAAQ,QAAQ;AAChC,UAAM,WAAW,SAAS,OAAO;AACjC,UAAM,aAAa,MAAM,SAAS,IAAI,IAAI,KAAK,CAAC,IAAI,MAAM,MAAM,OAAO;AAEvE,QAAI;AACF,YAAM,WAAW,SAAS,OAAO,EAAE;AACnC,UAAI,WAAW,MAAM,OAAO,MAAM;AAChC,gBAAQ,OAAO,MAAM;AAAA,iBAAoB,QAAQ,gEAAmB,WAAW,OAAO,MAAM,QAAQ,CAAC,CAAC;AAAA,CAAO;AAC7G,gBAAQ,WAAW;AACnB;AAAA,MACF;AACA,YAAM,SAAS,aAAa,OAAO;AACnC,YAAM,cAAc,cAAc,MAAM;AACxC,YAAM,SAAS,aAAa,WAAW;AAEvC,UAAI,CAAC,KAAK,QAAQ;AAChB,gBAAQ,OAAO,MAAM,YAAY,UAAU,GAAG,QAAQ,KAAK,MAAM,OAAO;AAAA,MAC1E;AAEA,YAAM,eAA6B,CAAC;AACpC,UAAI,KAAK,MAAO,cAAa,QAAQ,KAAK;AAC1C,UAAI,KAAK,iBAAiB,MAAO,cAAa,qBAAqB;AAGnE,YAAM,gBAAgB,CAAC,QAAQ,UAAU,UAAU,SAAS,UAAU,aAAa,KAAK;AACxF,UAAI,KAAK,KAAK;AACZ,YAAI,CAAC,cAAc,SAAS,KAAK,GAAG,GAAG;AACrC,kBAAQ,OAAO,MAAM,oEAA4B,KAAK,GAAG;AAAA,CAAI;AAC7D,kBAAQ,OAAO,MAAM,gCAAY,cAAc,KAAK,IAAI,CAAC;AAAA,CAAI;AAC7D,kBAAQ,KAAK,CAAC;AAAA,QAChB;AACA,qBAAa,UAAU,KAAK;AAAA,MAC9B,OAAO;AACL,qBAAa,UAAU;AAAA,MACzB;AAKA,UAAI,KAAK,SAAS;AAChB,cAAM,IAAI,SAAS,KAAK,SAAS,EAAE;AACnC,YAAI,IAAI,EAAG,cAAa,iBAAiB;AAAA,MAC3C,WAAW,aAAa,YAAY,aAAa;AAC/C,qBAAa,iBAAiB,KAAK,EAAE;AAAA,MACvC,OAAO;AACL,qBAAa,iBAAiB;AAAA,MAChC;AAGA,UAAI,KAAK,cAAc;AACrB,cAAM,IAAI,SAAS,KAAK,cAAc,EAAE;AACxC,YAAI,IAAI,EAAG,cAAa,eAAe;AAAA,MACzC;AAEA,UAAI,CAAC,KAAK,QAAQ;AAChB,qBAAa,aAAa,CAAC,SAAiB,UAAkB;AAC5D,kBAAQ,OAAO,MAAM,cAAc,UAAU,GAAG,QAAQ,KAAK,MAAM,MAAM,OAAO,IAAI,KAAK,GAAG;AAAA,QAC9F;AAAA,MACF;AACA,YAAM,SAAS,MAAM,MAAM,aAAa,YAAY;AAEpD,UAAI,CAAC,OAAO,SAAS;AACnB,gBAAQ,OAAO,MAAM;AAAA,CAAS;AAC9B,gBAAQ,OAAO,MAAM,YAAO,OAAO,KAAK;AAAA,CAAI;AAC5C,gBAAQ,WAAW;AACnB;AAAA,MACF;AAEA,UAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM;AAAA,CAAO;AAG9C,UAAI,CAAC,KAAK,UAAU,OAAO,WAAW,OAAO,cAAc;AACzD,gBAAQ,OAAO,MAAM;AAAA,CAA4B;AAAA,MACnD;AAGA,UAAI,CAAC,KAAK,UAAU,OAAO,WAAW,OAAO,UAAU,QAAQ;AAC7D,mBAAW,KAAK,OAAO,UAAU;AAC/B,kBAAQ,OAAO,MAAM,YAAO,EAAE,OAAO;AAAA,CAAI;AAAA,QAC3C;AAAA,MACF;AAEA,UAAI,WAAW,OAAO;AAEtB,UAAI,KAAK,UAAU,OAAO,QAAQ,QAAQ;AACxC,mBAAW,SAAS,QAAQ,uBAAuB,wBAAwB;AAAA,MAC7E;AACA,YAAM,SAAS,KAAK,WAAW,SAC3B,KAAK,UAAU,QAAQ,MAAM,CAAC,IAC9B;AAGJ,YAAM,aAAa,CAAC,gBAAwB;AAC1C,YAAI,CAAC,OAAO,QAAQ,OAAQ;AAC5B,cAAM,OAAO,SAAS,WAAW,EAAE,QAAQ,YAAY,EAAE;AACzD,cAAM,aAAa,QAAQ,aAAa,MAAM,OAAO,SAAS;AAC9D,cAAM,SAAS,KAAK,WAAW,QAAQ,KAAK,QAAQ,IAAI;AACxD,kBAAU,QAAQ,EAAE,WAAW,KAAK,CAAC;AACrC,mBAAW,OAAO,OAAO,QAAQ;AAC/B,wBAAc,QAAQ,QAAQ,IAAI,QAAQ,GAAG,IAAI,IAAI;AAAA,QACvD;AACA,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,OAAO,OAAO,MAAM,oCAAW,MAAM;AAAA,CAAI;AAAA,MACzF;AAEA,UAAI,KAAK,UAAU,MAAM,WAAW,GAAG;AACrC,sBAAc,KAAK,QAAQ,QAAQ,OAAO;AAC1C,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,KAAK,MAAM;AAAA,CAAI;AAC7D,mBAAW,QAAQ,KAAK,MAAM,CAAC;AAAA,MACjC,WAAW,KAAK,QAAQ;AACtB,kBAAU,KAAK,QAAQ,EAAE,WAAW,KAAK,CAAC;AAC1C,cAAM,SAAS,KAAK,WAAW,SAAS,UAAU;AAClD,cAAM,UAAU,QAAQ,KAAK,QAAQ,SAAS,QAAQ,YAAY,MAAM,CAAC;AACzE,sBAAc,SAAS,QAAQ,OAAO;AACtC,YAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAO,OAAO;AAAA,CAAI;AACzD,mBAAW,OAAO;AAAA,MACpB,OAAO;AACL,gBAAQ,OAAO,MAAM,SAAS,IAAI;AAClC,mBAAW,OAAO;AAAA,MACpB;AAAA,IACF,SAAS,KAAK;AACZ,YAAM,EAAE,cAAc,IAAI,MAAM,OAAO,qBAAY;AACnD,cAAQ,OAAO,MAAM;AAAA,kBAAqB,QAAQ,WAAM,cAAc,GAAG,CAAC;AAAA,CAAI;AAC9E,cAAQ,WAAW;AAAA,IACrB;AAAA,EACF;AACF;AAGA,SAAS,gBAAgB,KAAuB;AAC9C,SAAO,IACJ,OAAO,uBAAuB,2EAAoB,EAClD,OAAO,uBAAuB,0EAAmB,EACjD,OAAO,uBAAuB,mEAA2B,EACzD,OAAO,mBAAmB,wEAAgC,UAAU,EACpE,OAAO,sBAAsB,qEAAmB,EAChD,OAAO,qBAAqB,kHAAkC,EAC9D,OAAO,YAAY,oDAAY,EAC/B,OAAO,gBAAgB,qFAAiE,EACxF,OAAO,kBAAkB,sGAA0C,EACnE,OAAO,wBAAwB,sHAA0D;AAC9F;AAEA,QACG,wBAAwB,EACxB,KAAK,QAAQ,EACb,YAAY,2GAAoD,EAChE,QAAQ,OAAO;AAGlB;AAAA,EACE,QACG,QAAQ,OAAO,EACf,YAAY,mGAA4C,EACxD,SAAS,cAAc,8CAAW;AACvC,EAAE,OAAO,QAAQ;AAGjB;AAAA,EACE,QACG,SAAS,cAAc,2EAAwC;AACpE,EAAE,OAAO,QAAQ;AAEjB,QACG,QAAQ,iBAAiB,EACzB,YAAY,uFAA2B,EACvC,OAAO,uBAAuB,0CAAsB,MAAM,EAC1D,OAAO,uBAAuB,6GAA6B,EAC3D,OAAO,qBAAqB,qHAAqC,EACjE,OAAO,YAAY,kHAA6B,EAChD,OAAO,qBAAqB,uEAA0B,EACtD,OAAO,YAAY,oDAAY,EAC/B,OAAO,OAAO,OAAe,SAAS;AACrC,QAAM,eAAe,CAAC,QAAQ,MAAM;AACpC,MAAI,CAAC,aAAa,SAAS,KAAK,MAAM,GAAG;AACvC,YAAQ,OAAO,MAAM,gEAAwB,KAAK,MAAM;AAAA,CAAmB;AAC3E,YAAQ,KAAK,CAAC;AAAA,EAChB;AAEA,QAAM,WAAW,QAAQ,KAAK;AAC9B,MAAI,CAAC,WAAW,QAAQ,GAAG;AACzB,YAAQ,OAAO,MAAM,6EAA2B,KAAK;AAAA,CAAI;AACzD,YAAQ,KAAK,CAAC;AAAA,EAChB;AAEA,QAAM,OAAO,SAAS,QAAQ,EAAE,QAAQ,YAAY,EAAE;AACtD,QAAM,UAAU,KAAK,SACjB,QAAQ,KAAK,MAAM,IACnB,QAAQ,UAAU,MAAM,GAAG,IAAI,IAAI,KAAK,MAAM,EAAE;AAEpD,MAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,YAAY,SAAS,QAAQ,CAAC,WAAM,SAAS,OAAO,CAAC,MAAM;AAElG,MAAI;AACF,UAAM,WAAW,aAAa,UAAU,OAAO;AAG/C,UAAM,SAAS,KAAK,WAAW,QAAQ,KAAK,QAAQ,IAAI,QAAQ,UAAU,MAAM,OAAO,SAAS;AAChG,UAAM,SAAgD,CAAC;AACvD,QAAI,KAAK,UAAU,WAAW,MAAM,GAAG;AACrC,YAAM,UAAkC;AAAA,QACtC,KAAK;AAAA,QAAa,KAAK;AAAA,QAAc,MAAM;AAAA,QAC3C,KAAK;AAAA,QAAa,KAAK;AAAA,MACzB;AACA,iBAAW,SAAS,YAAY,QAAQ,EAAE,eAAe,KAAK,CAAC,GAAG;AAChE,YAAI,CAAC,MAAM,OAAO,EAAG;AACrB,cAAM,QAAQ,MAAM;AACpB,cAAM,MAAM,QAAQ,KAAK,EAAE,MAAM,CAAC,EAAE,YAAY;AAChD,YAAI,CAAC,QAAQ,GAAG,EAAG;AACnB,cAAM,OAAO,aAAa,QAAQ,QAAQ,KAAK,CAAC;AAChD,eAAO,KAAK,EAAE,UAAU,OAAO,MAAM,IAAI,WAAW,IAAI,GAAG,UAAU,QAAQ,GAAG,EAAE,CAAC;AAAA,MACrF;AACA,UAAI,CAAC,KAAK,OAAQ,SAAQ,OAAO,MAAM,+BAAW,OAAO,MAAM;AAAA,CAAQ;AAAA,IACzE;AAEA,UAAM,WAAqB,CAAC;AAE5B,QAAI;AACJ,QAAI,KAAK,WAAW,QAAQ;AAC1B,UAAI,KAAK,YAAY,CAAC,KAAK,QAAQ;AACjC,gBAAQ,OAAO,MAAM;AAAA;AAAA,CAAiD;AAAA,MACxE;AACA,YAAM,MAAM,eAAe,UAAU,EAAE,UAAU,QAAQ,OAAO,SAAS,SAAS,OAAU,CAAC;AAAA,IAC/F,OAAO;AACL,UAAI;AACJ,UAAI,KAAK,UAAU;AACjB,cAAM,UAAU,aAAa,QAAQ,KAAK,QAAQ,CAAC;AACnD,8BAAsB,QAAQ,OAAO,MAAM,QAAQ,YAAY,QAAQ,aAAa,QAAQ,UAAU;AAAA,MACxG;AACA,YAAM,MAAM,eAAe,UAAU,EAAE,UAAU,QAAQ,OAAO,SAAS,SAAS,QAAW,oBAAoB,CAAC;AAAA,IACpH;AAEA,kBAAc,SAAS,OAAO,KAAK,GAAG,CAAC;AAEvC,QAAI,CAAC,KAAK,QAAQ;AAChB,cAAQ,OAAO,MAAM;AAAA,CAAO;AAC5B,cAAQ,OAAO,MAAM,YAAO,OAAO;AAAA,CAAI;AACvC,UAAI,SAAS,OAAQ,UAAS,QAAQ,OAAK,QAAQ,OAAO,MAAM,KAAK,CAAC;AAAA,CAAI,CAAC;AAAA,IAC7E;AAAA,EACF,SAAS,KAAK;AACZ,UAAM,EAAE,cAAc,IAAI,MAAM,OAAO,qBAAY;AACnD,YAAQ,OAAO,MAAM;AAAA,CAAS;AAC9B,YAAQ,OAAO,MAAM,YAAO,cAAc,GAAG,CAAC;AAAA,CAAI;AAClD,YAAQ,KAAK,CAAC;AAAA,EAChB;AACF,CAAC;AAEH,QACG,QAAQ,aAAa,EACrB,YAAY,4FAAsB,EAClC,OAAO,mBAAmB,4CAAc,EACxC,OAAO,uBAAuB,iEAAe,EAC7C,OAAO,uBAAuB,8CAAW,EACzC,OAAO,mBAAmB,yDAA2B,UAAU,EAC/D,OAAO,YAAY,oDAAY,EAC/B,OAAO,OAAO,KAAa,SAAS;AACnC,QAAM,EAAE,eAAe,IAAI,MAAM,OAAO,qBAAY;AACpD,QAAM,eAAe;AAAA,IACnB;AAAA,IACA,QAAQ,KAAK;AAAA,IACb,SAAS,KAAK;AAAA,IACd,QAAQ,KAAK;AAAA,IACb,OAAO,KAAK;AAAA,IACZ,QAAQ,KAAK;AAAA,EACf,CAAC;AACH,CAAC;AAEH,QAAQ,MAAM;","names":[]}

package/dist/index.cjs

@@ -2242,9 +2242,9 @@ function createBatchCliProvider(mode, batchSize) {
         }
         let output;
         if (mode === "codex") {
-          output = callBatchCodexCli(tempFiles);
+          output = await callBatchCodexCli(tempFiles);
         } else {
-          output = callBatchCli(mode, tempFiles);
+          output = await callBatchCli(mode, tempFiles);
         }
         const cleaned = stripCodeFence2(output.trim());
         const parts = cleaned.split(/<!--\s*PAGE_BREAK\s*-->/).map((p) => p.trim()).filter((p) => p.length > 0);
@@ -2266,7 +2266,48 @@ function createBatchCliProvider(mode, batchSize) {
     }
   };
 }
-function callBatchCli(mode, imagePaths) {
+function spawnAsync(cmd, args, opts) {
+  return new Promise((resolve, reject) => {
+    const child = (0, import_child_process3.spawn)(cmd, args, {
+      cwd: opts.cwd,
+      env: process.env,
+      stdio: ["pipe", "pipe", "pipe"]
+    });
+    let stdout = "";
+    let stderr = "";
+    let killed = false;
+    child.stdout.setEncoding("utf-8");
+    child.stderr.setEncoding("utf-8");
+    child.stdout.on("data", (d) => {
+      stdout += d;
+    });
+    child.stderr.on("data", (d) => {
+      stderr += d;
+    });
+    const timer = setTimeout(() => {
+      killed = true;
+      child.kill("SIGTERM");
+    }, opts.timeoutMs);
+    if (opts.stdin !== void 0) {
+      child.stdin.end(opts.stdin);
+    } else {
+      child.stdin.end();
+    }
+    child.on("close", (code) => {
+      clearTimeout(timer);
+      if (killed) {
+        reject(new Error(`\uD0C0\uC784\uC544\uC6C3 (${Math.round(opts.timeoutMs / 1e3)}\uCD08)`));
+      } else {
+        resolve({ stdout, stderr, exitCode: code ?? 1 });
+      }
+    });
+    child.on("error", (err) => {
+      clearTimeout(timer);
+      reject(err);
+    });
+  });
+}
+async function callBatchCli(mode, imagePaths) {
   const fileRefs = imagePaths.map((p) => `@${p}`).join("\n");
   const prompt = `${BATCH_OCR_PROMPT}
 
@@ -2282,24 +2323,18 @@ ${fileRefs}`;
     if (model) args.push("--model", model);
   }
   const timeoutMs = 6e4 + imagePaths.length * 2e4;
-  const result = (0, import_child_process3.spawnSync)(mode, args, {
-    encoding: "utf-8",
-    timeout: timeoutMs,
-    maxBuffer: 50 * 1024 * 1024,
-    // 50MB (large batch output)
+  const result = await spawnAsync(mode, args, {
+    timeoutMs,
     ...mode === "claude" ? { cwd: (0, import_os2.tmpdir)() } : {}
   });
-  if (result.error) {
-    throw new Error(`${mode} \uBC30\uCE58 OCR \uC2E4\uD328: ${result.error.message}`);
-  }
-  if (result.status !== 0) {
-    const errMsg = result.stderr?.trim() || `exit code ${result.status}`;
+  if (result.exitCode !== 0) {
+    const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`;
     throw new Error(`${mode} \uBC30\uCE58 OCR \uC2E4\uD328: ${errMsg}`);
   }
   return result.stdout || "";
 }
-function callBatchCodexCli(imagePaths) {
-  const outPath = (0, import_path2.join)((0, import_os2.tmpdir)(), `kordoc-codex-batch-${Date.now()}.txt`);
+async function callBatchCodexCli(imagePaths) {
+  const outPath = (0, import_path2.join)((0, import_os2.tmpdir)(), `kordoc-codex-batch-${Date.now()}-${Math.random().toString(36).slice(2)}.txt`);
   try {
     const args = ["exec", BATCH_OCR_PROMPT];
     for (const p of imagePaths) {
@@ -2309,17 +2344,12 @@ function callBatchCodexCli(imagePaths) {
     const model = process.env.KORDOC_CODEX_MODEL;
     if (model) args.push("--model", model);
     const timeoutMs = 6e4 + imagePaths.length * 2e4;
-    const result = (0, import_child_process3.spawnSync)("codex", args, {
-      encoding: "utf-8",
-      timeout: timeoutMs,
-      maxBuffer: 50 * 1024 * 1024,
-      input: ""
+    const result = await spawnAsync("codex", args, {
+      timeoutMs,
+      stdin: ""
     });
-    if (result.error) {
-      throw new Error(`codex \uBC30\uCE58 OCR \uC2E4\uD328: ${result.error.message}`);
-    }
-    if (result.status !== 0) {
-      const errMsg = result.stderr?.trim() || `exit code ${result.status}`;
+    if (result.exitCode !== 0) {
+      const errMsg = result.stderr?.trim() || `exit code ${result.exitCode}`;
       throw new Error(`codex \uBC30\uCE58 OCR \uC2E4\uD328: ${errMsg}`);
     }
     try {
@@ -2581,7 +2611,7 @@ function isBatchProvider(p) {
 async function ocrPages(doc, provider, pageFilter, effectivePageCount, warnings, concurrency = 1, onProgress) {
   const blocks = [];
   if (isBatchProvider(provider)) {
-    return ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warnings, onProgress);
+    return ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warnings, concurrency, onProgress);
   }
   if (concurrency <= 1) {
     for (let i = 1; i <= effectivePageCount; i++) {
@@ -2628,8 +2658,7 @@ async function ocrPages(doc, provider, pageFilter, effectivePageCount, warnings,
   }
   return blocks;
 }
-async function ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warnings, onProgress) {
-  const blocks = [];
+async function ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warnings, concurrency = 1, onProgress) {
   const pageNumbers = [];
   for (let i = 1; i <= effectivePageCount; i++) {
     if (pageFilter && !pageFilter.has(i)) continue;
@@ -2646,16 +2675,16 @@ async function ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warn
     batches.push(pageImages.slice(i, i + provider.batchSize));
   }
   let processed = 0;
-  for (const batch of batches) {
+  const batchTasks = batches.map((batch, batchIdx) => async () => {
+    const pageBlocks = [];
     try {
       const results = await provider.processBatch(batch);
       for (const { pageNum } of batch) {
         const result = results.get(pageNum);
-        if (result) {
-          for (const b of ocrResultToBlocks(result, pageNum)) blocks.push(b);
-        }
-        processed++;
-        onProgress?.(processed, pageNumbers.length);
+        pageBlocks.push({
+          pageNum,
+          blocks: result ? ocrResultToBlocks(result, pageNum) : []
+        });
       }
     } catch (err) {
       const range = `${batch[0].pageNum}-${batch[batch.length - 1].pageNum}`;
@@ -2663,8 +2692,20 @@ async function ocrPagesBatch(doc, provider, pageFilter, effectivePageCount, warn
         message: `\uBC30\uCE58 OCR \uC2E4\uD328 (\uD398\uC774\uC9C0 ${range}): ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
         code: "OCR_PAGE_FAILED"
       });
-      processed += batch.length;
-      onProgress?.(processed, pageNumbers.length);
+      for (const { pageNum } of batch) {
+        pageBlocks.push({ pageNum, blocks: [] });
+      }
+    }
+    processed += batch.length;
+    onProgress?.(processed, pageNumbers.length);
+    return { batchIdx, pageBlocks };
+  });
+  const effectiveConcurrency = Math.max(1, concurrency);
+  const batchResults = await runWithConcurrency(batchTasks, effectiveConcurrency);
+  const blocks = [];
+  for (const result of batchResults) {
+    for (const { blocks: pageBlks } of result.pageBlocks) {
+      for (const b of pageBlks) blocks.push(b);
     }
   }
   return blocks;
@@ -2757,7 +2798,7 @@ var import_jszip2 = __toESM(require("jszip"), 1);
 var import_xmldom = require("@xmldom/xmldom");
 
 // src/utils.ts
-var VERSION = true ? "2.3.0" : "0.0.0-dev";
+var VERSION = true ? "2.3.1" : "0.0.0-dev";
 function toArrayBuffer(buf) {
   if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
     return buf.buffer;