@clazic/kordoc 2.2.6 → 2.2.8

package/dist/index.d.cts

@@ -110,6 +110,13 @@ interface ParseOptions {
      * - undefined: 라이브러리 API 기존 동작 유지 (자동 탐색 안 함)
      */
     ocrMode?: OcrMode;
+    /**
+     * OCR 병렬 처리 수.
+     * - tesseract: 기본값은 CPU 코어 수 (병렬 처리로 속도 향상)
+     * - CLI 제공 프로바이더(gemini/claude/codex): 기본 1 (rate limit 보호)
+     * - 1: 순차 처리 (기존 동작)
+     */
+    ocrConcurrency?: number;
     /** 진행률 콜백 — current: 현재 페이지/섹션, total: 전체 수 */
     onProgress?: (current: number, total: number) => void;
     /** PDF 머리글/바닥글 자동 제거 */

package/dist/index.d.ts

@@ -110,6 +110,13 @@ interface ParseOptions {
      * - undefined: 라이브러리 API 기존 동작 유지 (자동 탐색 안 함)
      */
     ocrMode?: OcrMode;
+    /**
+     * OCR 병렬 처리 수.
+     * - tesseract: 기본값은 CPU 코어 수 (병렬 처리로 속도 향상)
+     * - CLI 제공 프로바이더(gemini/claude/codex): 기본 1 (rate limit 보호)
+     * - 1: 순차 처리 (기존 동작)
+     */
+    ocrConcurrency?: number;
     /** 진행률 콜백 — current: 현재 페이지/섹션, total: 전체 수 */
     onProgress?: (current: number, total: number) => void;
     /** PDF 머리글/바닥글 자동 제거 */

package/dist/index.js

@@ -2051,8 +2051,10 @@ function callCli(mode, imagePath) {
   const args = buildCliArgs(mode, imagePath);
   const result = spawnSync(mode, args, {
     encoding: "utf-8",
-    timeout: 6e4,
-    maxBuffer: 10 * 1024 * 1024
+    timeout: 18e4,
+    maxBuffer: 10 * 1024 * 1024,
+    // claude: /tmp에서 실행하여 프로젝트 CLAUDE.md의 규칙 간섭 방지
+    ...mode === "claude" ? { cwd: tmpdir() } : {}
   });
   if (result.error) {
     throw new Error(`${mode} CLI \uC2E4\uD589 \uC2E4\uD328: ${result.error.message}`);
@@ -2158,6 +2160,7 @@ var init_cli_provider = __esm({
 // src/ocr/tesseract-provider.ts
 var tesseract_provider_exports = {};
 __export(tesseract_provider_exports, {
+  createTesseractPoolProvider: () => createTesseractPoolProvider,
   createTesseractProvider: () => createTesseractProvider
 });
 import { createWorker } from "tesseract.js";
@@ -2176,6 +2179,37 @@ async function createTesseractProvider() {
   };
   return provider;
 }
+async function createTesseractPoolProvider(concurrency) {
+  const workers = await Promise.all(
+    Array.from({ length: concurrency }, () => createWorker("kor+eng"))
+  );
+  const idle = [...workers];
+  const waitQueue = [];
+  function acquire() {
+    if (idle.length > 0) return Promise.resolve(idle.pop());
+    return new Promise((resolve) => waitQueue.push(resolve));
+  }
+  function release(w) {
+    if (waitQueue.length > 0) {
+      waitQueue.shift()(w);
+    } else {
+      idle.push(w);
+    }
+  }
+  const provider = async (pageImage, _pageNumber, _mimeType) => {
+    const w = await acquire();
+    try {
+      const { data } = await w.recognize(pageImage);
+      return data.text;
+    } finally {
+      release(w);
+    }
+  };
+  provider.terminate = async () => {
+    await Promise.all(workers.map((w) => w.terminate()));
+  };
+  return provider;
+}
 var init_tesseract_provider = __esm({
   "src/ocr/tesseract-provider.ts"() {
     "use strict";
@@ -2187,14 +2221,17 @@ var resolve_exports = {};
 __export(resolve_exports, {
   resolveOcrProvider: () => resolveOcrProvider
 });
-async function resolveOcrProvider(mode, warnings) {
+async function resolveOcrProvider(mode, warnings, concurrency) {
   if (mode === "off") {
     throw new Error("OCR\uC774 \uBE44\uD65C\uC131\uD654\uB418\uC5B4 \uC788\uC2B5\uB2C8\uB2E4 (--ocr off).");
   }
   if (mode !== "auto") {
     validateOcrMode(mode);
     if (mode === "tesseract") {
-      const { createTesseractProvider: createTesseractProvider2 } = await Promise.resolve().then(() => (init_tesseract_provider(), tesseract_provider_exports));
+      const { createTesseractProvider: createTesseractProvider2, createTesseractPoolProvider: createTesseractPoolProvider2 } = await Promise.resolve().then(() => (init_tesseract_provider(), tesseract_provider_exports));
+      if (concurrency && concurrency > 1) {
+        return createTesseractPoolProvider2(concurrency);
+      }
       return createTesseractProvider2();
     }
     return createCliOcrProvider(mode);
@@ -2214,7 +2251,10 @@ async function resolveOcrProvider(mode, warnings) {
     }
   }
   if (detected === "tesseract") {
-    const { createTesseractProvider: createTesseractProvider2 } = await Promise.resolve().then(() => (init_tesseract_provider(), tesseract_provider_exports));
+    const { createTesseractProvider: createTesseractProvider2, createTesseractPoolProvider: createTesseractPoolProvider2 } = await Promise.resolve().then(() => (init_tesseract_provider(), tesseract_provider_exports));
+    if (concurrency && concurrency > 1) {
+      return createTesseractPoolProvider2(concurrency);
+    }
     return createTesseractProvider2();
   }
   return createCliOcrProvider(detected);
@@ -2352,32 +2392,77 @@ var provider_exports = {};
 __export(provider_exports, {
   ocrPages: () => ocrPages
 });
-async function ocrPages(doc, provider, pageFilter, effectivePageCount, warnings) {
+async function runWithConcurrency(tasks, limit) {
+  const results = new Array(tasks.length);
+  let nextIndex = 0;
+  async function worker() {
+    while (nextIndex < tasks.length) {
+      const idx = nextIndex++;
+      results[idx] = await tasks[idx]();
+    }
+  }
+  await Promise.all(Array.from({ length: Math.min(limit, tasks.length) }, () => worker()));
+  return results;
+}
+function ocrResultToBlocks(result, pageNum) {
+  const pageBlocks = [];
+  if (typeof result === "string") {
+    if (result.trim()) {
+      pageBlocks.push({ type: "paragraph", text: result.trim(), pageNumber: pageNum });
+    }
+  } else if (result && typeof result === "object" && "markdown" in result) {
+    const structured = result;
+    if (structured.markdown.trim()) {
+      const converted = markdownToBlocks(structured.markdown, pageNum);
+      for (const b of converted) pageBlocks.push(b);
+    }
+  }
+  return pageBlocks;
+}
+async function ocrPages(doc, provider, pageFilter, effectivePageCount, warnings, concurrency = 1) {
   const blocks = [];
+  if (concurrency <= 1) {
+    for (let i = 1; i <= effectivePageCount; i++) {
+      if (pageFilter && !pageFilter.has(i)) continue;
+      const page = await doc.getPage(i);
+      try {
+        const imageData = await renderPageToPng(page);
+        const result = await provider(imageData, i, "image/png");
+        for (const b of ocrResultToBlocks(result, i)) blocks.push(b);
+      } catch (err) {
+        warnings?.push({
+          page: i,
+          message: `\uD398\uC774\uC9C0 ${i} OCR \uC2E4\uD328: ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
+          code: "OCR_PAGE_FAILED"
+        });
+      }
+    }
+    return blocks;
+  }
+  const pageNumbers = [];
   for (let i = 1; i <= effectivePageCount; i++) {
     if (pageFilter && !pageFilter.has(i)) continue;
-    const page = await doc.getPage(i);
+    pageNumbers.push(i);
+  }
+  const tasks = pageNumbers.map((pageNum) => async () => {
     try {
+      const page = await doc.getPage(pageNum);
       const imageData = await renderPageToPng(page);
-      const result = await provider(imageData, i, "image/png");
-      if (typeof result === "string") {
-        if (result.trim()) {
-          blocks.push({ type: "paragraph", text: result.trim(), pageNumber: i });
-        }
-      } else if (result && typeof result === "object" && "markdown" in result) {
-        const structured = result;
-        if (structured.markdown.trim()) {
-          const pageBlocks = markdownToBlocks(structured.markdown, i);
-          for (const b of pageBlocks) blocks.push(b);
-        }
-      }
+      const result = await provider(imageData, pageNum, "image/png");
+      return { pageNum, pageBlocks: ocrResultToBlocks(result, pageNum) };
     } catch (err) {
       warnings?.push({
-        page: i,
-        message: `\uD398\uC774\uC9C0 ${i} OCR \uC2E4\uD328: ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
+        page: pageNum,
+        message: `\uD398\uC774\uC9C0 ${pageNum} OCR \uC2E4\uD328: ${err instanceof Error ? err.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`,
         code: "OCR_PAGE_FAILED"
       });
+      return null;
     }
+  });
+  const taskResults = await runWithConcurrency(tasks, concurrency);
+  for (const item of taskResults) {
+    if (!item) continue;
+    for (const b of item.pageBlocks) blocks.push(b);
   }
   return blocks;
 }
@@ -2446,7 +2531,7 @@ import JSZip2 from "jszip";
 import { DOMParser } from "@xmldom/xmldom";
 
 // src/utils.ts
-var VERSION = true ? "2.2.6" : "0.0.0-dev";
+var VERSION = true ? "2.2.8" : "0.0.0-dev";
 function toArrayBuffer(buf) {
   if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
     return buf.buffer;
@@ -5963,7 +6048,8 @@ async function parsePdfDocument(buffer, options) {
       if (!ocrProvider && ocrMode && ocrMode !== "off") {
         try {
           const { resolveOcrProvider: resolveOcrProvider2 } = await Promise.resolve().then(() => (init_resolve(), resolve_exports));
-          ocrProvider = await resolveOcrProvider2(ocrMode, warnings);
+          const concurrency = options?.ocrConcurrency ?? 1;
+          ocrProvider = await resolveOcrProvider2(ocrMode, warnings, concurrency);
         } catch (resolveErr) {
           if (ocrMode !== "auto") {
             throw Object.assign(
@@ -5976,7 +6062,8 @@ async function parsePdfDocument(buffer, options) {
       if (ocrProvider) {
         try {
           const { ocrPages: ocrPages2 } = await Promise.resolve().then(() => (init_provider(), provider_exports));
-          const ocrBlocks = await ocrPages2(doc, ocrProvider, pageFilter, effectivePageCount, warnings);
+          const concurrency = options?.ocrConcurrency ?? 1;
+          const ocrBlocks = await ocrPages2(doc, ocrProvider, pageFilter, effectivePageCount, warnings, concurrency);
           if (ocrBlocks.length > 0) {
             const ocrMarkdown = blocksToMarkdown(ocrBlocks);
             return {