kordoc 4.17.6 → 4.18.0

package/README.md

@@ -342,11 +342,16 @@ for (const p of r.pageQuality ?? []) if (p.needsOcr) console.log(`p${p.page} 검
 
 ## 💻 CLI
 
+[병렬 변환 성능 측정과 제한 사항](docs/parallel-batch.md)
+
+`--jobs N`은 상주 프로세스 N개로 파일 단위 병렬 변환을 실행합니다(기본 1). `--jobs > 1`의 다중 입력은 `-d`가 필요하며, 확장자를 제외한 파일 이름이 서로 달라야 합니다. 일부 파일이 실패해도 나머지는 계속 처리하고 종료 코드는 실패로 설정합니다. 완료 로그와 실패 JSON은 완료 순서로 나옵니다. 프로세스 수만큼 메모리 사용이 늘고, 작은 파일은 오히려 느릴 수 있습니다. OCR은 자체 네이티브 스레드도 사용합니다. 단일 문서의 페이지 병렬화나 JavaScript/MCP API 변경은 포함하지 않습니다.
+
 ```bash
 # 변환
 npx kordoc 사업계획서.hwpx                           # 터미널 출력
 npx kordoc 보고서.hwp -o 보고서.md                   # 파일 저장 (그림은 images/보고서/)
 npx kordoc *.pdf -d ./변환결과/                      # 일괄 변환
+npx kordoc *.pdf --jobs 4 -d ./변환결과/             # 병렬 변환 (상주 프로세스 4개)
 npx kordoc 검토서.hwpx --format json                # JSON (blocks + pages + metadata)
 npx kordoc 검토서.pdf --format chunks               # RAG 구조 청크 JSON (breadcrumb + 표 독립 청크)
 npx kordoc 보고서.hwpx --pages 1-3                   # 쪽 범위

package/dist/{-Y6U3UTVZ.js → -DHSHP3AB.js}

@@ -50,17 +50,17 @@ import {
   unknownFontWarnings,
   usesGaejosikMunche,
   validateHwpx
-} from "./chunk-ZS5KKM4L.js";
+} from "./chunk-OLOMR452.js";
 import {
   blocksToPages
-} from "./chunk-7F7CUU22.js";
+} from "./chunk-VVJCQKXV.js";
 import {
   applySplices,
   buildParagraphSplices,
   buildRangeSplices,
   placeSealHwpx,
   scanSectionXml
-} from "./chunk-PGHIQ6E5.js";
+} from "./chunk-Q6SKAFWN.js";
 import {
   DEFAULT_REDACT_RULES,
   redactMarkdown,
@@ -68,7 +68,7 @@ import {
 } from "./chunk-3WSNLH4B.js";
 import {
   blocksToChunks
-} from "./chunk-B3RQHQCC.js";
+} from "./chunk-WDJJ77HP.js";
 import {
   SPACE_EM_FIXED,
   SPACE_EM_FONT,
@@ -93,20 +93,20 @@ import {
   renderSceneToHtml,
   simulateWrap,
   simulateWrapKeepWord
-} from "./chunk-7R7NFGHY.js";
+} from "./chunk-NKCKN4WW.js";
 import "./chunk-7FCNOKOR.js";
-import "./chunk-2WUE6JIT.js";
+import "./chunk-FPU7WPQM.js";
 import "./chunk-VQFA55IC.js";
-import "./chunk-6A4Z6VYA.js";
+import "./chunk-LLD4EY2F.js";
 import {
   blocksToMarkdown,
   flattenLayoutTables,
   hasStructuredCellContent
-} from "./chunk-5VHV6W7W.js";
-import "./chunk-MOM67IND.js";
+} from "./chunk-DSF64VZN.js";
+import "./chunk-WSYTO525.js";
 import {
   VERSION
-} from "./chunk-T5O6HEHG.js";
+} from "./chunk-32FQLCUJ.js";
 export {
   BUILTIN_TEMPLATES,
   DEFAULT_REDACT_RULES,
@@ -196,4 +196,4 @@ export {
   usesGaejosikMunche,
   validateHwpx
 };
-//# sourceMappingURL=-Y6U3UTVZ.js.map
+//# sourceMappingURL=-DHSHP3AB.js.map

package/dist/{chunk-LLFXHX7K.cjs → chunk-2QBZMCRT.cjs}

@@ -6,10 +6,10 @@
 
 
 
-var _chunkCZOPNDTOcjs = require('./chunk-CZOPNDTO.cjs');
+var _chunkELTPM4HMcjs = require('./chunk-ELTPM4HM.cjs');
 
 
-var _chunkFVESLY6Vcjs = require('./chunk-FVESLY6V.cjs');
+var _chunkCKMGVW6Scjs = require('./chunk-CKMGVW6S.cjs');
 
 // src/ocr/ruling-lines.ts
 var INK_LUMA_MAX = 205;
@@ -1410,7 +1410,7 @@ var OcrEngine = class _OcrEngine {
     ]);
     const sharpAny = sharpModRaw;
     const sharpMod = typeof sharpAny === "function" ? sharpAny : _nullishCoalesce(sharpAny.default, () => ( sharpAny));
-    const dir = _chunkCZOPNDTOcjs.getOcrModelsDir.call(void 0, );
+    const dir = _chunkELTPM4HMcjs.getOcrModelsDir.call(void 0, );
     const sessionOpts = {
       graphOptimizationLevel: "all",
       executionProviders: ["cpu"],
@@ -1418,11 +1418,11 @@ var OcrEngine = class _OcrEngine {
       // paddle2onnx 변환 잔여물 W 로그 폭주 억제
     };
     const [det, rec, dictYml] = await Promise.all([
-      ortMod.InferenceSession.create(_path.join.call(void 0, dir, _chunkCZOPNDTOcjs.OCR_DET_MODEL.filename), sessionOpts),
-      ortMod.InferenceSession.create(_path.join.call(void 0, dir, _chunkCZOPNDTOcjs.OCR_REC_MODEL.filename), sessionOpts),
-      _promises.readFile.call(void 0, _path.join.call(void 0, dir, _chunkCZOPNDTOcjs.OCR_REC_DICT.filename), "utf-8")
+      ortMod.InferenceSession.create(_path.join.call(void 0, dir, _chunkELTPM4HMcjs.OCR_DET_MODEL.filename), sessionOpts),
+      ortMod.InferenceSession.create(_path.join.call(void 0, dir, _chunkELTPM4HMcjs.OCR_REC_MODEL.filename), sessionOpts),
+      _promises.readFile.call(void 0, _path.join.call(void 0, dir, _chunkELTPM4HMcjs.OCR_REC_DICT.filename), "utf-8")
     ]);
-    const dict = _chunkCZOPNDTOcjs.parseCharacterDict.call(void 0, dictYml);
+    const dict = _chunkELTPM4HMcjs.parseCharacterDict.call(void 0, dictYml);
     if (dict.length === 0) throw new Error("OCR \uC0AC\uC804 \uD30C\uC2F1 \uC2E4\uD328 \u2014 \uBAA8\uB378 \uCE90\uC2DC\uB97C \uC0AD\uC81C \uD6C4 \uC7AC\uB2E4\uC6B4\uB85C\uB4DC\uD558\uC138\uC694");
     return new _OcrEngine({ det, rec, dict, ort: ortMod, sharp: sharpMod });
   }
@@ -1750,7 +1750,7 @@ async function tryImport(name, loader) {
     return await loader();
   } catch (e) {
     throw new Error(
-      `\uB0B4\uC7A5 OCR \uC744 \uC0AC\uC6A9\uD558\uB824\uBA74 optional dependency '${name}' \uC774 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install ${name}\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694.${_chunkFVESLY6Vcjs.OPTIONAL_DEP_INSTALL_HINT} \uC6D0\uC778: ${e.message}`
+      `\uB0B4\uC7A5 OCR \uC744 \uC0AC\uC6A9\uD558\uB824\uBA74 optional dependency '${name}' \uC774 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install ${name}\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694.${_chunkCKMGVW6Scjs.OPTIONAL_DEP_INSTALL_HINT} \uC6D0\uC778: ${e.message}`
     );
   }
 }
@@ -1864,7 +1864,7 @@ async function runPdfOcr(buffer, targets, mode, warnings, onProgress, detectTabl
     () => Promise.resolve().then(() => _interopRequireWildcard(require("@hyzyla/pdfium")))
   );
   if (mode === "builtin") {
-    await _chunkCZOPNDTOcjs.ensureOcrModels.call(void 0, (p) => {
+    await _chunkELTPM4HMcjs.ensureOcrModels.call(void 0, (p) => {
       if (p.phase === "download" && p.downloaded === 0) {
         process.stderr.write(`[kordoc-ocr] ${p.spec.name} \uB2E4\uC6B4\uB85C\uB4DC \uC911 (~${p.spec.sizeMb}MB)...
 `);
@@ -2012,14 +2012,14 @@ function ocrItemsToBlocks(items, pageNumber, pdfW, pdfH, scale, extraLines, dete
       isHidden: false
     };
   });
-  return _chunkCZOPNDTOcjs.extractPageBlocksWithLines.call(void 0, norm, pageNumber, opList, pdfW, pdfH, extraLines, detectTables);
+  return _chunkELTPM4HMcjs.extractPageBlocksWithLines.call(void 0, norm, pageNumber, opList, pdfW, pdfH, extraLines, detectTables);
 }
 async function tryImport2(name, loader) {
   try {
     return await loader();
   } catch (e) {
     throw new Error(
-      `OCR \uC744 \uC0AC\uC6A9\uD558\uB824\uBA74 optional dependency '${name}' \uC774 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install ${name}\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694.${_chunkFVESLY6Vcjs.OPTIONAL_DEP_INSTALL_HINT} \uC6D0\uC778: ${e.message}`
+      `OCR \uC744 \uC0AC\uC6A9\uD558\uB824\uBA74 optional dependency '${name}' \uC774 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install ${name}\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694.${_chunkCKMGVW6Scjs.OPTIONAL_DEP_INSTALL_HINT} \uC6D0\uC778: ${e.message}`
     );
   }
 }
@@ -2058,4 +2058,4 @@ function bgraToRgba(bgra) {
 
 
 exports.getOcrEngine = getOcrEngine; exports.detectRulingLines = detectRulingLines; exports.rulingToPdfLines = rulingToPdfLines; exports.deskewPage = deskewPage; exports.MAX_OCR_PIXELS = MAX_OCR_PIXELS; exports.runPdfOcr = runPdfOcr; exports.ocrItemsToBlocks = ocrItemsToBlocks;
-//# sourceMappingURL=chunk-LLFXHX7K.cjs.map
+//# sourceMappingURL=chunk-2QBZMCRT.cjs.map