kordoc 4.2.0 → 4.2.1

package/dist/{detect-YEZX2XCF.js → detect-FX2ZHKZP.js}

@@ -6,10 +6,11 @@ import {
   isHwp3File,
   isHwpmlFile,
   isHwpxFile,
+  isImageFile,
   isOldHwpFile,
   isPdfFile,
   isZipFile
-} from "./chunk-YOO6ET6M.js";
+} from "./chunk-TCJRGFYY.js";
 export {
   detectFormat,
   detectOle2Format,
@@ -17,8 +18,9 @@ export {
   isHwp3File,
   isHwpmlFile,
   isHwpxFile,
+  isImageFile,
   isOldHwpFile,
   isPdfFile,
   isZipFile
 };
-//# sourceMappingURL=detect-YEZX2XCF.js.map
+//# sourceMappingURL=detect-FX2ZHKZP.js.map

package/dist/image-ocr-DALT7PFQ.js

@@ -0,0 +1,65 @@
+import {
+  detectRulingLines,
+  ensureOcrModels,
+  getOcrEngine,
+  ocrItemsToBlocks,
+  rulingToPdfLines
+} from "./chunk-KRAYUNFM.js";
+import "./chunk-J2RZ444H.js";
+import "./chunk-EZHWPKCH.js";
+import "./chunk-HTCKGSSW.js";
+
+// src/ocr/image-ocr.ts
+var IMAGE_SCALE = 3;
+async function parseImageDocument(buffer, options) {
+  const warnings = [];
+  if (typeof options?.ocr === "function") {
+    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer));
+    if (!text.trim()) {
+      warnings.push({ page: 1, message: "OCR \uACB0\uACFC \uC5C6\uC74C", code: "OCR_FAILED" });
+      return { blocks: [], warnings };
+    }
+    return { blocks: [{ type: "paragraph", text: text.trim(), pageNumber: 1 }], warnings };
+  }
+  const { data, width, height } = await decodeToRgba(buffer);
+  await ensureOcrModels((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)...
+`);
+    }
+  });
+  const engine = await getOcrEngine();
+  const items = await engine.recognizePage(data, width, height);
+  if (items.length === 0) {
+    warnings.push({ page: 1, message: "\uC774\uBBF8\uC9C0\uC5D0\uC11C \uD14D\uC2A4\uD2B8\uB97C \uC778\uC2DD\uD558\uC9C0 \uBABB\uD588\uC2B5\uB2C8\uB2E4", code: "OCR_FAILED" });
+    return { blocks: [], warnings };
+  }
+  const pdfW = width / IMAGE_SCALE;
+  const pdfH = height / IMAGE_SCALE;
+  const ruling = detectRulingLines(data, width, height, IMAGE_SCALE);
+  const extraLines = rulingToPdfLines(ruling, IMAGE_SCALE, pdfH);
+  return { blocks: ocrItemsToBlocks(items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings };
+}
+async function decodeToRgba(buffer) {
+  let sharp;
+  try {
+    const mod = await import("sharp");
+    sharp = typeof mod === "function" ? mod : mod.default ?? mod;
+  } catch (e) {
+    throw new Error(
+      `\uC774\uBBF8\uC9C0 \uD30C\uC2F1\uC5D0\uB294 optional dependency 'sharp' \uAC00 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install sharp\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694. \uC6D0\uC778: ${e.message}`
+    );
+  }
+  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true });
+  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height };
+}
+function detectImageMime(buffer) {
+  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength));
+  if (b[0] === 255 && b[1] === 216) return "image/jpeg";
+  if (b[0] === 82 && b[1] === 73) return "image/webp";
+  return "image/png";
+}
+export {
+  parseImageDocument
+};
+//# sourceMappingURL=image-ocr-DALT7PFQ.js.map

package/dist/image-ocr-DALT7PFQ.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/ocr/image-ocr.ts"],"sourcesContent":["/**\n * 이미지 직접 입력 OCR — PNG/JPEG/WebP 를 PDF 래핑 없이 바로 파싱한다.\n *\n * 이미지는 텍스트층이 없으므로 OCR 이 항상 필요하다 — `--ocr` 플래그 없이도\n * 내장 엔진(PP-OCRv5 korean)을 기본 적용하고, OcrProvider 가 주어지면 위임한다.\n * 좌표는 216dpi(스케일 3) 가정으로 PDF pt 공간에 환산해 기존 블록 파이프라인\n * (괘선 감지 → 선 기반 표 → 클러스터 → xy-cut)을 그대로 태운다.\n *\n * 디코딩은 optional dependency `sharp` — 미설치면 MISSING_DEPENDENCY 안내.\n */\n\nimport type { IRBlock, ParseOptions, ParseWarning } from \"../types.js\"\nimport { getOcrEngine } from \"./engine.js\"\nimport { ensureOcrModels } from \"./models.js\"\nimport { detectRulingLines, rulingToPdfLines } from \"./ruling-lines.js\"\nimport { ocrItemsToBlocks } from \"./pdf-ocr.js\"\n\n/** 좌표 환산 스케일 (px/pt) — PDF OCR 렌더(216dpi)와 동일 기준 */\nconst IMAGE_SCALE = 3\n\nexport interface ImageOcrResult {\n  blocks: IRBlock[]\n  warnings: ParseWarning[]\n}\n\n/** 이미지 버퍼 → OCR → IRBlock[] (표 괘선 감지 포함) */\nexport async function parseImageDocument(\n  buffer: ArrayBuffer,\n  options?: ParseOptions,\n): Promise<ImageOcrResult> {\n  const warnings: ParseWarning[] = []\n\n  // 사용자 OcrProvider — 원본 바이트 그대로 위임 (종전 PDF 경로와 같은 계약)\n  if (typeof options?.ocr === \"function\") {\n    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer))\n    if (!text.trim()) {\n      warnings.push({ page: 1, message: \"OCR 결과 없음\", code: \"OCR_FAILED\" })\n      return { blocks: [], warnings }\n    }\n    return { blocks: [{ type: \"paragraph\", text: text.trim(), pageNumber: 1 }], warnings }\n  }\n\n  const { data, width, height } = await decodeToRgba(buffer)\n\n  await ensureOcrModels(p => {\n    if (p.phase === \"download\" && p.downloaded === 0) {\n      process.stderr.write(`[kordoc-ocr] ${p.spec.name} 다운로드 중 (~${p.spec.sizeMb}MB)...\\n`)\n    }\n  })\n  const engine = await getOcrEngine()\n  const items = await engine.recognizePage(data, width, height)\n  if (items.length === 0) {\n    warnings.push({ page: 1, message: \"이미지에서 텍스트를 인식하지 못했습니다\", code: \"OCR_FAILED\" })\n    return { blocks: [], warnings }\n  }\n\n  const pdfW = width / IMAGE_SCALE\n  const pdfH = height / IMAGE_SCALE\n  const ruling = detectRulingLines(data, width, height, IMAGE_SCALE)\n  const extraLines = rulingToPdfLines(ruling, IMAGE_SCALE, pdfH)\n  return { blocks: ocrItemsToBlocks(items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings }\n}\n\n/** sharp 로 RGBA 디코딩 — 미설치는 MISSING_DEPENDENCY 로 분류되도록 안내 메시지 throw */\nasync function decodeToRgba(\n  buffer: ArrayBuffer,\n): Promise<{ data: Uint8Array; width: number; height: number }> {\n  type SharpFactory = (input: Buffer) => {\n    ensureAlpha(): { raw(): { toBuffer(opts: { resolveWithObject: true }): Promise<{ data: Buffer; info: { width: number; height: number } }> } }\n  }\n  let sharp: SharpFactory\n  try {\n    const mod = (await import(\"sharp\")) as unknown as { default?: SharpFactory } & SharpFactory\n    sharp = typeof mod === \"function\" ? mod : (mod.default ?? (mod as unknown as SharpFactory))\n  } catch (e) {\n    throw new Error(\n      \"이미지 파싱에는 optional dependency 'sharp' 가 필요합니다. \" +\n        `\\`npm install sharp\\` 후 다시 실행하세요. 원인: ${(e as Error).message}`,\n    )\n  }\n  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true })\n  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height }\n}\n\n/** OcrProvider 계약용 mime 판별 */\nfunction detectImageMime(buffer: ArrayBuffer): string {\n  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength))\n  if (b[0] === 0xff && b[1] === 0xd8) return \"image/jpeg\"\n  if (b[0] === 0x52 && b[1] === 0x49) return \"image/webp\"\n  return \"image/png\"\n}\n"],"mappings":";;;;;;;;;;;;AAkBA,IAAM,cAAc;AAQpB,eAAsB,mBACpB,QACA,SACyB;AACzB,QAAM,WAA2B,CAAC;AAGlC,MAAI,OAAO,SAAS,QAAQ,YAAY;AACtC,UAAM,OAAO,MAAM,QAAQ,IAAI,IAAI,WAAW,MAAM,GAAG,GAAG,gBAAgB,MAAM,CAAC;AACjF,QAAI,CAAC,KAAK,KAAK,GAAG;AAChB,eAAS,KAAK,EAAE,MAAM,GAAG,SAAS,iCAAa,MAAM,aAAa,CAAC;AACnE,aAAO,EAAE,QAAQ,CAAC,GAAG,SAAS;AAAA,IAChC;AACA,WAAO,EAAE,QAAQ,CAAC,EAAE,MAAM,aAAa,MAAM,KAAK,KAAK,GAAG,YAAY,EAAE,CAAC,GAAG,SAAS;AAAA,EACvF;AAEA,QAAM,EAAE,MAAM,OAAO,OAAO,IAAI,MAAM,aAAa,MAAM;AAEzD,QAAM,gBAAgB,OAAK;AACzB,QAAI,EAAE,UAAU,cAAc,EAAE,eAAe,GAAG;AAChD,cAAQ,OAAO,MAAM,gBAAgB,EAAE,KAAK,IAAI,sCAAa,EAAE,KAAK,MAAM;AAAA,CAAU;AAAA,IACtF;AAAA,EACF,CAAC;AACD,QAAM,SAAS,MAAM,aAAa;AAClC,QAAM,QAAQ,MAAM,OAAO,cAAc,MAAM,OAAO,MAAM;AAC5D,MAAI,MAAM,WAAW,GAAG;AACtB,aAAS,KAAK,EAAE,MAAM,GAAG,SAAS,mHAAyB,MAAM,aAAa,CAAC;AAC/E,WAAO,EAAE,QAAQ,CAAC,GAAG,SAAS;AAAA,EAChC;AAEA,QAAM,OAAO,QAAQ;AACrB,QAAM,OAAO,SAAS;AACtB,QAAM,SAAS,kBAAkB,MAAM,OAAO,QAAQ,WAAW;AACjE,QAAM,aAAa,iBAAiB,QAAQ,aAAa,IAAI;AAC7D,SAAO,EAAE,QAAQ,iBAAiB,OAAO,GAAG,MAAM,MAAM,aAAa,UAAU,GAAG,SAAS;AAC7F;AAGA,eAAe,aACb,QAC8D;AAI9D,MAAI;AACJ,MAAI;AACF,UAAM,MAAO,MAAM,OAAO,OAAO;AACjC,YAAQ,OAAO,QAAQ,aAAa,MAAO,IAAI,WAAY;AAAA,EAC7D,SAAS,GAAG;AACV,UAAM,IAAI;AAAA,MACR,0MAC4C,EAAY,OAAO;AAAA,IACjE;AAAA,EACF;AACA,QAAM,EAAE,MAAM,KAAK,IAAI,MAAM,MAAM,OAAO,KAAK,MAAM,CAAC,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,mBAAmB,KAAK,CAAC;AAChH,SAAO,EAAE,MAAM,IAAI,WAAW,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU,GAAG,OAAO,KAAK,OAAO,QAAQ,KAAK,OAAO;AACvH;AAGA,SAAS,gBAAgB,QAA6B;AACpD,QAAM,IAAI,IAAI,WAAW,QAAQ,GAAG,KAAK,IAAI,IAAI,OAAO,UAAU,CAAC;AACnE,MAAI,EAAE,CAAC,MAAM,OAAQ,EAAE,CAAC,MAAM,IAAM,QAAO;AAC3C,MAAI,EAAE,CAAC,MAAM,MAAQ,EAAE,CAAC,MAAM,GAAM,QAAO;AAC3C,SAAO;AACT;","names":[]}

package/dist/image-ocr-N5Z32R4U.cjs

@@ -0,0 +1,66 @@
+"use strict";Object.defineProperty(exports, "__esModule", {value: true}); function _interopRequireWildcard(obj) { if (obj && obj.__esModule) { return obj; } else { var newObj = {}; if (obj != null) { for (var key in obj) { if (Object.prototype.hasOwnProperty.call(obj, key)) { newObj[key] = obj[key]; } } } newObj.default = obj; return newObj; } } function _nullishCoalesce(lhs, rhsFn) { if (lhs != null) { return lhs; } else { return rhsFn(); } } function _optionalChain(ops) { let lastAccessLHS = undefined; let value = ops[0]; let i = 1; while (i < ops.length) { const op = ops[i]; const fn = ops[i + 1]; i += 2; if ((op === 'optionalAccess' || op === 'optionalCall') && value == null) { return undefined; } if (op === 'access' || op === 'optionalAccess') { lastAccessLHS = value; value = fn(value); } else if (op === 'call' || op === 'optionalCall') { value = fn((...args) => value.call(lastAccessLHS, ...args)); lastAccessLHS = undefined; } } return value; }
+
+
+
+
+
+var _chunk35U3MVA2cjs = require('./chunk-35U3MVA2.cjs');
+require('./chunk-H7OWEYH2.cjs');
+require('./chunk-R22KDUKW.cjs');
+require('./chunk-RBUBZ2TA.cjs');
+require('./chunk-GS7T56RP.cjs');
+
+// src/ocr/image-ocr.ts
+var IMAGE_SCALE = 3;
+async function parseImageDocument(buffer, options) {
+  const warnings = [];
+  if (typeof _optionalChain([options, 'optionalAccess', _ => _.ocr]) === "function") {
+    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer));
+    if (!text.trim()) {
+      warnings.push({ page: 1, message: "OCR \uACB0\uACFC \uC5C6\uC74C", code: "OCR_FAILED" });
+      return { blocks: [], warnings };
+    }
+    return { blocks: [{ type: "paragraph", text: text.trim(), pageNumber: 1 }], warnings };
+  }
+  const { data, width, height } = await decodeToRgba(buffer);
+  await _chunk35U3MVA2cjs.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)...
+`);
+    }
+  });
+  const engine = await _chunk35U3MVA2cjs.getOcrEngine.call(void 0, );
+  const items = await engine.recognizePage(data, width, height);
+  if (items.length === 0) {
+    warnings.push({ page: 1, message: "\uC774\uBBF8\uC9C0\uC5D0\uC11C \uD14D\uC2A4\uD2B8\uB97C \uC778\uC2DD\uD558\uC9C0 \uBABB\uD588\uC2B5\uB2C8\uB2E4", code: "OCR_FAILED" });
+    return { blocks: [], warnings };
+  }
+  const pdfW = width / IMAGE_SCALE;
+  const pdfH = height / IMAGE_SCALE;
+  const ruling = _chunk35U3MVA2cjs.detectRulingLines.call(void 0, data, width, height, IMAGE_SCALE);
+  const extraLines = _chunk35U3MVA2cjs.rulingToPdfLines.call(void 0, ruling, IMAGE_SCALE, pdfH);
+  return { blocks: _chunk35U3MVA2cjs.ocrItemsToBlocks.call(void 0, items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings };
+}
+async function decodeToRgba(buffer) {
+  let sharp;
+  try {
+    const mod = await Promise.resolve().then(() => _interopRequireWildcard(require("sharp")));
+    sharp = typeof mod === "function" ? mod : _nullishCoalesce(mod.default, () => ( mod));
+  } catch (e) {
+    throw new Error(
+      `\uC774\uBBF8\uC9C0 \uD30C\uC2F1\uC5D0\uB294 optional dependency 'sharp' \uAC00 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install sharp\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694. \uC6D0\uC778: ${e.message}`
+    );
+  }
+  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true });
+  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height };
+}
+function detectImageMime(buffer) {
+  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength));
+  if (b[0] === 255 && b[1] === 216) return "image/jpeg";
+  if (b[0] === 82 && b[1] === 73) return "image/webp";
+  return "image/png";
+}
+
+
+exports.parseImageDocument = parseImageDocument;
+//# sourceMappingURL=image-ocr-N5Z32R4U.cjs.map

package/dist/image-ocr-N5Z32R4U.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["/Users/chris_gomdori/workspace/kordoc/dist/image-ocr-N5Z32R4U.cjs","../src/ocr/image-ocr.ts"],"names":[],"mappings":"AAAA;AACE;AACA;AACA;AACA;AACA;AACF,wDAA6B;AAC7B,gCAA6B;AAC7B,gCAA6B;AAC7B,gCAA6B;AAC7B,gCAA6B;AAC7B;AACA;ACMA,IAAM,YAAA,EAAc,CAAA;AAQpB,MAAA,SAAsB,kBAAA,CACpB,MAAA,EACA,OAAA,EACyB;AACzB,EAAA,MAAM,SAAA,EAA2B,CAAC,CAAA;AAGlC,EAAA,GAAA,CAAI,uBAAO,OAAA,2BAAS,MAAA,IAAQ,UAAA,EAAY;AACtC,IAAA,MAAM,KAAA,EAAO,MAAM,OAAA,CAAQ,GAAA,CAAI,IAAI,UAAA,CAAW,MAAM,CAAA,EAAG,CAAA,EAAG,eAAA,CAAgB,MAAM,CAAC,CAAA;AACjF,IAAA,GAAA,CAAI,CAAC,IAAA,CAAK,IAAA,CAAK,CAAA,EAAG;AAChB,MAAA,QAAA,CAAS,IAAA,CAAK,EAAE,IAAA,EAAM,CAAA,EAAG,OAAA,EAAS,+BAAA,EAAa,IAAA,EAAM,aAAa,CAAC,CAAA;AACnE,MAAA,OAAO,EAAE,MAAA,EAAQ,CAAC,CAAA,EAAG,SAAS,CAAA;AAAA,IAChC;AACA,IAAA,OAAO,EAAE,MAAA,EAAQ,CAAC,EAAE,IAAA,EAAM,WAAA,EAAa,IAAA,EAAM,IAAA,CAAK,IAAA,CAAK,CAAA,EAAG,UAAA,EAAY,EAAE,CAAC,CAAA,EAAG,SAAS,CAAA;AAAA,EACvF;AAEA,EAAA,MAAM,EAAE,IAAA,EAAM,KAAA,EAAO,OAAO,EAAA,EAAI,MAAM,YAAA,CAAa,MAAM,CAAA;AAEzD,EAAA,MAAM,+CAAA,CAAgB,CAAA,EAAA,GAAK;AACzB,IAAA,GAAA,CAAI,CAAA,CAAE,MAAA,IAAU,WAAA,GAAc,CAAA,CAAE,WAAA,IAAe,CAAA,EAAG;AAChD,MAAA,OAAA,CAAQ,MAAA,CAAO,KAAA,CAAM,CAAA,aAAA,EAAgB,CAAA,CAAE,IAAA,CAAK,IAAI,CAAA,mCAAA,EAAa,CAAA,CAAE,IAAA,CAAK,MAAM,CAAA;AAAA,CAAU,CAAA;AAAA,IACtF;AAAA,EACF,CAAC,CAAA;AACD,EAAA,MAAM,OAAA,EAAS,MAAM,4CAAA,CAAa;AAClC,EAAA,MAAM,MAAA,EAAQ,MAAM,MAAA,CAAO,aAAA,CAAc,IAAA,EAAM,KAAA,EAAO,MAAM,CAAA;AAC5D,EAAA,GAAA,CAAI,KAAA,CAAM,OAAA,IAAW,CAAA,EAAG;AACtB,IAAA,QAAA,CAAS,IAAA,CAAK,EAAE,IAAA,EAAM,CAAA,EAAG,OAAA,EAAS,iHAAA,EAAyB,IAAA,EAAM,aAAa,CAAC,CAAA;AAC/E,IAAA,OAAO,EAAE,MAAA,EAAQ,CAAC,CAAA,EAAG,SAAS,CAAA;AAAA,EAChC;AAEA,EAAA,MAAM,KAAA,EAAO,MAAA,EAAQ,WAAA;AACrB,EAAA,MAAM,KAAA,EAAO,OAAA,EAAS,WAAA;AACtB,EAAA,MAAM,OAAA,EAAS,iDAAA,IAAkB,EAAM,KAAA,EAAO,MAAA,EAAQ,WAAW,CAAA;AACjE,EAAA,MAAM,WAAA,EAAa,gDAAA,MAAiB,EAAQ,WAAA,EAAa,IAAI,CAAA;AAC7D,EAAA,OAAO,EAAE,MAAA,EAAQ,gDAAA,KAAiB,EAAO,CAAA,EAAG,IAAA,EAAM,IAAA,EAAM,WAAA,EAAa,UAAU,CAAA,EAAG,SAAS,CAAA;AAC7F;AAGA,MAAA,SAAe,YAAA,CACb,MAAA,EAC8D;AAI9D,EAAA,IAAI,KAAA;AACJ,EAAA,IAAI;AACF,IAAA,MAAM,IAAA,EAAO,MAAM,4DAAA,CAAO,OAAO,GAAA;AACjC,IAAA,MAAA,EAAQ,OAAO,IAAA,IAAQ,WAAA,EAAa,IAAA,mBAAO,GAAA,CAAI,OAAA,UAAY,KAAA;AAAA,EAC7D,EAAA,MAAA,CAAS,CAAA,EAAG;AACV,IAAA,MAAM,IAAI,KAAA;AAAA,MACR,CAAA,uMAAA,EAC4C,CAAA,CAAY,OAAO,CAAA;AAAA,IAAA;AACjE,EAAA;AAEF,EAAA;AACA,EAAA;AACF;AAGA;AACE,EAAA;AACA,EAAA;AACA,EAAA;AACA,EAAA;AACF;AD5BA;AACA;AACA","file":"/Users/chris_gomdori/workspace/kordoc/dist/image-ocr-N5Z32R4U.cjs","sourcesContent":[null,"/**\n * 이미지 직접 입력 OCR — PNG/JPEG/WebP 를 PDF 래핑 없이 바로 파싱한다.\n *\n * 이미지는 텍스트층이 없으므로 OCR 이 항상 필요하다 — `--ocr` 플래그 없이도\n * 내장 엔진(PP-OCRv5 korean)을 기본 적용하고, OcrProvider 가 주어지면 위임한다.\n * 좌표는 216dpi(스케일 3) 가정으로 PDF pt 공간에 환산해 기존 블록 파이프라인\n * (괘선 감지 → 선 기반 표 → 클러스터 → xy-cut)을 그대로 태운다.\n *\n * 디코딩은 optional dependency `sharp` — 미설치면 MISSING_DEPENDENCY 안내.\n */\n\nimport type { IRBlock, ParseOptions, ParseWarning } from \"../types.js\"\nimport { getOcrEngine } from \"./engine.js\"\nimport { ensureOcrModels } from \"./models.js\"\nimport { detectRulingLines, rulingToPdfLines } from \"./ruling-lines.js\"\nimport { ocrItemsToBlocks } from \"./pdf-ocr.js\"\n\n/** 좌표 환산 스케일 (px/pt) — PDF OCR 렌더(216dpi)와 동일 기준 */\nconst IMAGE_SCALE = 3\n\nexport interface ImageOcrResult {\n  blocks: IRBlock[]\n  warnings: ParseWarning[]\n}\n\n/** 이미지 버퍼 → OCR → IRBlock[] (표 괘선 감지 포함) */\nexport async function parseImageDocument(\n  buffer: ArrayBuffer,\n  options?: ParseOptions,\n): Promise<ImageOcrResult> {\n  const warnings: ParseWarning[] = []\n\n  // 사용자 OcrProvider — 원본 바이트 그대로 위임 (종전 PDF 경로와 같은 계약)\n  if (typeof options?.ocr === \"function\") {\n    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer))\n    if (!text.trim()) {\n      warnings.push({ page: 1, message: \"OCR 결과 없음\", code: \"OCR_FAILED\" })\n      return { blocks: [], warnings }\n    }\n    return { blocks: [{ type: \"paragraph\", text: text.trim(), pageNumber: 1 }], warnings }\n  }\n\n  const { data, width, height } = await decodeToRgba(buffer)\n\n  await ensureOcrModels(p => {\n    if (p.phase === \"download\" && p.downloaded === 0) {\n      process.stderr.write(`[kordoc-ocr] ${p.spec.name} 다운로드 중 (~${p.spec.sizeMb}MB)...\\n`)\n    }\n  })\n  const engine = await getOcrEngine()\n  const items = await engine.recognizePage(data, width, height)\n  if (items.length === 0) {\n    warnings.push({ page: 1, message: \"이미지에서 텍스트를 인식하지 못했습니다\", code: \"OCR_FAILED\" })\n    return { blocks: [], warnings }\n  }\n\n  const pdfW = width / IMAGE_SCALE\n  const pdfH = height / IMAGE_SCALE\n  const ruling = detectRulingLines(data, width, height, IMAGE_SCALE)\n  const extraLines = rulingToPdfLines(ruling, IMAGE_SCALE, pdfH)\n  return { blocks: ocrItemsToBlocks(items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings }\n}\n\n/** sharp 로 RGBA 디코딩 — 미설치는 MISSING_DEPENDENCY 로 분류되도록 안내 메시지 throw */\nasync function decodeToRgba(\n  buffer: ArrayBuffer,\n): Promise<{ data: Uint8Array; width: number; height: number }> {\n  type SharpFactory = (input: Buffer) => {\n    ensureAlpha(): { raw(): { toBuffer(opts: { resolveWithObject: true }): Promise<{ data: Buffer; info: { width: number; height: number } }> } }\n  }\n  let sharp: SharpFactory\n  try {\n    const mod = (await import(\"sharp\")) as unknown as { default?: SharpFactory } & SharpFactory\n    sharp = typeof mod === \"function\" ? mod : (mod.default ?? (mod as unknown as SharpFactory))\n  } catch (e) {\n    throw new Error(\n      \"이미지 파싱에는 optional dependency 'sharp' 가 필요합니다. \" +\n        `\\`npm install sharp\\` 후 다시 실행하세요. 원인: ${(e as Error).message}`,\n    )\n  }\n  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true })\n  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height }\n}\n\n/** OcrProvider 계약용 mime 판별 */\nfunction detectImageMime(buffer: ArrayBuffer): string {\n  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength))\n  if (b[0] === 0xff && b[1] === 0xd8) return \"image/jpeg\"\n  if (b[0] === 0x52 && b[1] === 0x49) return \"image/webp\"\n  return \"image/png\"\n}\n"]}

package/dist/image-ocr-SLUEM3FA.js

@@ -0,0 +1,69 @@
+#!/usr/bin/env node
+import {
+  detectRulingLines,
+  getOcrEngine,
+  ocrItemsToBlocks,
+  rulingToPdfLines
+} from "./chunk-F5ECB53I.js";
+import {
+  ensureOcrModels
+} from "./chunk-GEPINOL6.js";
+import "./chunk-NNDVFNMJ.js";
+import "./chunk-L5JWKIOT.js";
+import "./chunk-54WVWWT4.js";
+import "./chunk-Z6LPEWIX.js";
+
+// src/ocr/image-ocr.ts
+var IMAGE_SCALE = 3;
+async function parseImageDocument(buffer, options) {
+  const warnings = [];
+  if (typeof options?.ocr === "function") {
+    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer));
+    if (!text.trim()) {
+      warnings.push({ page: 1, message: "OCR \uACB0\uACFC \uC5C6\uC74C", code: "OCR_FAILED" });
+      return { blocks: [], warnings };
+    }
+    return { blocks: [{ type: "paragraph", text: text.trim(), pageNumber: 1 }], warnings };
+  }
+  const { data, width, height } = await decodeToRgba(buffer);
+  await ensureOcrModels((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)...
+`);
+    }
+  });
+  const engine = await getOcrEngine();
+  const items = await engine.recognizePage(data, width, height);
+  if (items.length === 0) {
+    warnings.push({ page: 1, message: "\uC774\uBBF8\uC9C0\uC5D0\uC11C \uD14D\uC2A4\uD2B8\uB97C \uC778\uC2DD\uD558\uC9C0 \uBABB\uD588\uC2B5\uB2C8\uB2E4", code: "OCR_FAILED" });
+    return { blocks: [], warnings };
+  }
+  const pdfW = width / IMAGE_SCALE;
+  const pdfH = height / IMAGE_SCALE;
+  const ruling = detectRulingLines(data, width, height, IMAGE_SCALE);
+  const extraLines = rulingToPdfLines(ruling, IMAGE_SCALE, pdfH);
+  return { blocks: ocrItemsToBlocks(items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings };
+}
+async function decodeToRgba(buffer) {
+  let sharp;
+  try {
+    const mod = await import("sharp");
+    sharp = typeof mod === "function" ? mod : mod.default ?? mod;
+  } catch (e) {
+    throw new Error(
+      `\uC774\uBBF8\uC9C0 \uD30C\uC2F1\uC5D0\uB294 optional dependency 'sharp' \uAC00 \uD544\uC694\uD569\uB2C8\uB2E4. \`npm install sharp\` \uD6C4 \uB2E4\uC2DC \uC2E4\uD589\uD558\uC138\uC694. \uC6D0\uC778: ${e.message}`
+    );
+  }
+  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true });
+  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height };
+}
+function detectImageMime(buffer) {
+  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength));
+  if (b[0] === 255 && b[1] === 216) return "image/jpeg";
+  if (b[0] === 82 && b[1] === 73) return "image/webp";
+  return "image/png";
+}
+export {
+  parseImageDocument
+};
+//# sourceMappingURL=image-ocr-SLUEM3FA.js.map

package/dist/image-ocr-SLUEM3FA.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../src/ocr/image-ocr.ts"],"sourcesContent":["/**\n * 이미지 직접 입력 OCR — PNG/JPEG/WebP 를 PDF 래핑 없이 바로 파싱한다.\n *\n * 이미지는 텍스트층이 없으므로 OCR 이 항상 필요하다 — `--ocr` 플래그 없이도\n * 내장 엔진(PP-OCRv5 korean)을 기본 적용하고, OcrProvider 가 주어지면 위임한다.\n * 좌표는 216dpi(스케일 3) 가정으로 PDF pt 공간에 환산해 기존 블록 파이프라인\n * (괘선 감지 → 선 기반 표 → 클러스터 → xy-cut)을 그대로 태운다.\n *\n * 디코딩은 optional dependency `sharp` — 미설치면 MISSING_DEPENDENCY 안내.\n */\n\nimport type { IRBlock, ParseOptions, ParseWarning } from \"../types.js\"\nimport { getOcrEngine } from \"./engine.js\"\nimport { ensureOcrModels } from \"./models.js\"\nimport { detectRulingLines, rulingToPdfLines } from \"./ruling-lines.js\"\nimport { ocrItemsToBlocks } from \"./pdf-ocr.js\"\n\n/** 좌표 환산 스케일 (px/pt) — PDF OCR 렌더(216dpi)와 동일 기준 */\nconst IMAGE_SCALE = 3\n\nexport interface ImageOcrResult {\n  blocks: IRBlock[]\n  warnings: ParseWarning[]\n}\n\n/** 이미지 버퍼 → OCR → IRBlock[] (표 괘선 감지 포함) */\nexport async function parseImageDocument(\n  buffer: ArrayBuffer,\n  options?: ParseOptions,\n): Promise<ImageOcrResult> {\n  const warnings: ParseWarning[] = []\n\n  // 사용자 OcrProvider — 원본 바이트 그대로 위임 (종전 PDF 경로와 같은 계약)\n  if (typeof options?.ocr === \"function\") {\n    const text = await options.ocr(new Uint8Array(buffer), 1, detectImageMime(buffer))\n    if (!text.trim()) {\n      warnings.push({ page: 1, message: \"OCR 결과 없음\", code: \"OCR_FAILED\" })\n      return { blocks: [], warnings }\n    }\n    return { blocks: [{ type: \"paragraph\", text: text.trim(), pageNumber: 1 }], warnings }\n  }\n\n  const { data, width, height } = await decodeToRgba(buffer)\n\n  await ensureOcrModels(p => {\n    if (p.phase === \"download\" && p.downloaded === 0) {\n      process.stderr.write(`[kordoc-ocr] ${p.spec.name} 다운로드 중 (~${p.spec.sizeMb}MB)...\\n`)\n    }\n  })\n  const engine = await getOcrEngine()\n  const items = await engine.recognizePage(data, width, height)\n  if (items.length === 0) {\n    warnings.push({ page: 1, message: \"이미지에서 텍스트를 인식하지 못했습니다\", code: \"OCR_FAILED\" })\n    return { blocks: [], warnings }\n  }\n\n  const pdfW = width / IMAGE_SCALE\n  const pdfH = height / IMAGE_SCALE\n  const ruling = detectRulingLines(data, width, height, IMAGE_SCALE)\n  const extraLines = rulingToPdfLines(ruling, IMAGE_SCALE, pdfH)\n  return { blocks: ocrItemsToBlocks(items, 1, pdfW, pdfH, IMAGE_SCALE, extraLines), warnings }\n}\n\n/** sharp 로 RGBA 디코딩 — 미설치는 MISSING_DEPENDENCY 로 분류되도록 안내 메시지 throw */\nasync function decodeToRgba(\n  buffer: ArrayBuffer,\n): Promise<{ data: Uint8Array; width: number; height: number }> {\n  type SharpFactory = (input: Buffer) => {\n    ensureAlpha(): { raw(): { toBuffer(opts: { resolveWithObject: true }): Promise<{ data: Buffer; info: { width: number; height: number } }> } }\n  }\n  let sharp: SharpFactory\n  try {\n    const mod = (await import(\"sharp\")) as unknown as { default?: SharpFactory } & SharpFactory\n    sharp = typeof mod === \"function\" ? mod : (mod.default ?? (mod as unknown as SharpFactory))\n  } catch (e) {\n    throw new Error(\n      \"이미지 파싱에는 optional dependency 'sharp' 가 필요합니다. \" +\n        `\\`npm install sharp\\` 후 다시 실행하세요. 원인: ${(e as Error).message}`,\n    )\n  }\n  const { data, info } = await sharp(Buffer.from(buffer)).ensureAlpha().raw().toBuffer({ resolveWithObject: true })\n  return { data: new Uint8Array(data.buffer, data.byteOffset, data.byteLength), width: info.width, height: info.height }\n}\n\n/** OcrProvider 계약용 mime 판별 */\nfunction detectImageMime(buffer: ArrayBuffer): string {\n  const b = new Uint8Array(buffer, 0, Math.min(12, buffer.byteLength))\n  if (b[0] === 0xff && b[1] === 0xd8) return \"image/jpeg\"\n  if (b[0] === 0x52 && b[1] === 0x49) return \"image/webp\"\n  return \"image/png\"\n}\n"],"mappings":";;;;;;;;;;;;;;;;AAkBA,IAAM,cAAc;AAQpB,eAAsB,mBACpB,QACA,SACyB;AACzB,QAAM,WAA2B,CAAC;AAGlC,MAAI,OAAO,SAAS,QAAQ,YAAY;AACtC,UAAM,OAAO,MAAM,QAAQ,IAAI,IAAI,WAAW,MAAM,GAAG,GAAG,gBAAgB,MAAM,CAAC;AACjF,QAAI,CAAC,KAAK,KAAK,GAAG;AAChB,eAAS,KAAK,EAAE,MAAM,GAAG,SAAS,iCAAa,MAAM,aAAa,CAAC;AACnE,aAAO,EAAE,QAAQ,CAAC,GAAG,SAAS;AAAA,IAChC;AACA,WAAO,EAAE,QAAQ,CAAC,EAAE,MAAM,aAAa,MAAM,KAAK,KAAK,GAAG,YAAY,EAAE,CAAC,GAAG,SAAS;AAAA,EACvF;AAEA,QAAM,EAAE,MAAM,OAAO,OAAO,IAAI,MAAM,aAAa,MAAM;AAEzD,QAAM,gBAAgB,OAAK;AACzB,QAAI,EAAE,UAAU,cAAc,EAAE,eAAe,GAAG;AAChD,cAAQ,OAAO,MAAM,gBAAgB,EAAE,KAAK,IAAI,sCAAa,EAAE,KAAK,MAAM;AAAA,CAAU;AAAA,IACtF;AAAA,EACF,CAAC;AACD,QAAM,SAAS,MAAM,aAAa;AAClC,QAAM,QAAQ,MAAM,OAAO,cAAc,MAAM,OAAO,MAAM;AAC5D,MAAI,MAAM,WAAW,GAAG;AACtB,aAAS,KAAK,EAAE,MAAM,GAAG,SAAS,mHAAyB,MAAM,aAAa,CAAC;AAC/E,WAAO,EAAE,QAAQ,CAAC,GAAG,SAAS;AAAA,EAChC;AAEA,QAAM,OAAO,QAAQ;AACrB,QAAM,OAAO,SAAS;AACtB,QAAM,SAAS,kBAAkB,MAAM,OAAO,QAAQ,WAAW;AACjE,QAAM,aAAa,iBAAiB,QAAQ,aAAa,IAAI;AAC7D,SAAO,EAAE,QAAQ,iBAAiB,OAAO,GAAG,MAAM,MAAM,aAAa,UAAU,GAAG,SAAS;AAC7F;AAGA,eAAe,aACb,QAC8D;AAI9D,MAAI;AACJ,MAAI;AACF,UAAM,MAAO,MAAM,OAAO,OAAO;AACjC,YAAQ,OAAO,QAAQ,aAAa,MAAO,IAAI,WAAY;AAAA,EAC7D,SAAS,GAAG;AACV,UAAM,IAAI;AAAA,MACR,0MAC4C,EAAY,OAAO;AAAA,IACjE;AAAA,EACF;AACA,QAAM,EAAE,MAAM,KAAK,IAAI,MAAM,MAAM,OAAO,KAAK,MAAM,CAAC,EAAE,YAAY,EAAE,IAAI,EAAE,SAAS,EAAE,mBAAmB,KAAK,CAAC;AAChH,SAAO,EAAE,MAAM,IAAI,WAAW,KAAK,QAAQ,KAAK,YAAY,KAAK,UAAU,GAAG,OAAO,KAAK,OAAO,QAAQ,KAAK,OAAO;AACvH;AAGA,SAAS,gBAAgB,QAA6B;AACpD,QAAM,IAAI,IAAI,WAAW,QAAQ,GAAG,KAAK,IAAI,IAAI,OAAO,UAAU,CAAC;AACnE,MAAI,EAAE,CAAC,MAAM,OAAQ,EAAE,CAAC,MAAM,IAAM,QAAO;AAC3C,MAAI,EAAE,CAAC,MAAM,MAAQ,EAAE,CAAC,MAAM,GAAM,QAAO;AAC3C,SAAO;AACT;","names":[]}