@clazic/kordoc 2.4.2 → 2.4.4

package/dist/index.js

@@ -1961,6 +1961,13 @@ var require_cfb = __commonJS({
 });
 
 // src/ocr/auto-detect.ts
+var auto_detect_exports = {};
+__export(auto_detect_exports, {
+  detectAvailableOcr: () => detectAvailableOcr,
+  getAutoFallbackChain: () => getAutoFallbackChain,
+  getTesseractFallbackMessage: () => getTesseractFallbackMessage,
+  validateOcrMode: () => validateOcrMode
+});
 import { execSync } from "child_process";
 function detectAvailableOcr() {
   for (const cli of CLI_PRIORITY) {
@@ -1977,6 +1984,14 @@ function isCliInstalled(name) {
     return false;
   }
 }
+function getAutoFallbackChain() {
+  const chain = [];
+  for (const cli of CLI_PRIORITY) {
+    if (isCliInstalled(cli)) chain.push(cli);
+  }
+  chain.push("tesseract");
+  return chain;
+}
 function validateOcrMode(mode) {
   if (mode === "auto" || mode === "off" || mode === "tesseract") return;
   if (!isCliInstalled(mode)) {
@@ -2425,7 +2440,7 @@ async function resolveOcrProvider(mode, warnings, concurrency, batchSize) {
     return createCliOcrProvider(mode);
   }
   const detected = detectAvailableOcr();
-  if (detected !== "gemini") {
+  if (detected !== "codex") {
     if (detected === "tesseract") {
       warnings?.push({
         message: getTesseractFallbackMessage(),
@@ -2433,7 +2448,7 @@ async function resolveOcrProvider(mode, warnings, concurrency, batchSize) {
       });
     } else {
       warnings?.push({
-        message: `OCR: '${detected}' \uC0AC\uC6A9 \uC911 (gemini CLI\uAC00 \uC5C6\uC5B4 fallback). \uB354 \uB098\uC740 \uD488\uC9C8\uC744 \uC704\uD574 gemini CLI \uC124\uCE58\uB97C \uAD8C\uC7A5\uD569\uB2C8\uB2E4.`,
+        message: `OCR: '${detected}' \uC0AC\uC6A9 \uC911 (codex CLI\uAC00 \uC5C6\uC5B4 fallback). \uB354 \uB098\uC740 \uD488\uC9C8\uC744 \uC704\uD574 codex CLI \uC124\uCE58\uB97C \uAD8C\uC7A5\uD569\uB2C8\uB2E4.`,
         code: "OCR_CLI_FALLBACK"
       });
     }
@@ -2790,7 +2805,7 @@ import JSZip2 from "jszip";
 import { DOMParser } from "@xmldom/xmldom";
 
 // src/utils.ts
-var VERSION = true ? "2.4.2" : "0.0.0-dev";
+var VERSION = true ? "2.4.4" : "0.0.0-dev";
 function toArrayBuffer(buf) {
   if (buf.byteOffset === 0 && buf.byteLength === buf.buffer.byteLength) {
     return buf.buffer;
@@ -6318,66 +6333,93 @@ async function parsePdfDocument(buffer, options) {
         warnings.push({ page: i, message: `\uD398\uC774\uC9C0 ${i} \uD30C\uC2F1 \uC2E4\uD328: ${pageErr instanceof Error ? pageErr.message : "\uC54C \uC218 \uC5C6\uB294 \uC624\uB958"}`, code: "PARTIAL_PARSE" });
       }
     };
-    const sampleCount = Math.min(5, targetPageNums.length);
-    for (let si = 0; si < sampleCount; si++) {
+    const SAMPLE_SIZE = Math.min(10, targetPageNums.length);
+    const sampledIndices = /* @__PURE__ */ new Set();
+    if (targetPageNums.length <= SAMPLE_SIZE) {
+      for (let i = 0; i < targetPageNums.length; i++) sampledIndices.add(i);
+    } else {
+      for (let i = 0; i < SAMPLE_SIZE; i++) {
+        const idx = Math.round(i * (targetPageNums.length - 1) / (SAMPLE_SIZE - 1));
+        sampledIndices.add(idx);
+      }
+    }
+    for (const si of sampledIndices) {
       await parseSinglePage(targetPageNums[si]);
     }
-    const sampleParsed = parsedPages || sampleCount;
+    const sampleParsed = parsedPages || sampledIndices.size;
     const isImageBased = totalChars / Math.max(sampleParsed, 1) < 10;
     if (!isImageBased) {
-      for (let si = sampleCount; si < targetPageNums.length; si++) {
-        await parseSinglePage(targetPageNums[si]);
+      for (let si = 0; si < targetPageNums.length; si++) {
+        if (!sampledIndices.has(si)) {
+          await parseSinglePage(targetPageNums[si]);
+        }
       }
     }
     const parsedPageCount = parsedPages || (pageFilter ? pageFilter.size : effectivePageCount);
     if (isImageBased) {
-      let ocrProvider = options?.ocr ?? null;
       const ocrMode = options?.ocrMode ?? "auto";
-      if (!ocrProvider && ocrMode !== "off") {
-        try {
-          const { resolveOcrProvider: resolveOcrProvider2 } = await Promise.resolve().then(() => (init_resolve(), resolve_exports));
-          const concurrency = options?.ocrConcurrency ?? 1;
-          const batchSize = options?.ocrBatchSize;
-          ocrProvider = await resolveOcrProvider2(ocrMode, warnings, concurrency, batchSize);
-        } catch (resolveErr) {
-          if (ocrMode !== "auto") {
-            throw Object.assign(
-              new KordocError(resolveErr instanceof Error ? resolveErr.message : "OCR \uD504\uB85C\uBC14\uC774\uB354 \uCD08\uAE30\uD654 \uC2E4\uD328"),
-              { isImageBased: true }
-            );
-          }
-        }
+      const concurrency = options?.ocrConcurrency ?? 1;
+      const batchSize = options?.ocrBatchSize;
+      if (ocrMode === "off") {
+        throw Object.assign(new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
       }
-      if (ocrProvider) {
-        let ocrBlocks = [];
+      const { resolveOcrProvider: resolveOcrProvider2 } = await Promise.resolve().then(() => (init_resolve(), resolve_exports));
+      const { ocrPages: ocrPages2 } = await Promise.resolve().then(() => (init_provider(), provider_exports));
+      const tryProvider = async (provider) => {
         try {
-          const { ocrPages: ocrPages2 } = await Promise.resolve().then(() => (init_provider(), provider_exports));
-          const concurrency = options?.ocrConcurrency ?? 1;
-          ocrBlocks = await ocrPages2(doc, ocrProvider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
+          return await ocrPages2(doc, provider, pageFilter, effectivePageCount, warnings, concurrency, options?.onProgress);
         } catch {
+          return [];
         } finally {
-          const terminable = ocrProvider;
+          const terminable = provider;
           if (typeof terminable.terminate === "function") {
             await terminable.terminate().catch(() => {
             });
           }
         }
-        if (ocrBlocks.length > 0) {
-          const ocrMarkdown = blocksToMarkdown(ocrBlocks);
-          return {
-            markdown: ocrMarkdown,
-            blocks: ocrBlocks,
-            metadata,
-            warnings: warnings.length > 0 ? warnings : void 0,
-            isImageBased: true
-          };
+      };
+      let ocrBlocks = [];
+      if (options?.ocr) {
+        ocrBlocks = await tryProvider(options.ocr);
+      } else if (ocrMode === "auto") {
+        const { getAutoFallbackChain: getAutoFallbackChain2 } = await Promise.resolve().then(() => (init_auto_detect(), auto_detect_exports));
+        for (const mode of getAutoFallbackChain2()) {
+          try {
+            const provider = await resolveOcrProvider2(mode, warnings, concurrency, batchSize);
+            const blocks2 = await tryProvider(provider);
+            if (blocks2.length > 0) {
+              ocrBlocks = blocks2;
+              break;
+            }
+            warnings.push({ message: `OCR: '${mode}' \uACB0\uACFC \uC5C6\uC74C, \uB2E4\uC74C \uC5D4\uC9C4\uC73C\uB85C \uC2DC\uB3C4`, code: "OCR_CLI_FALLBACK" });
+          } catch {
+          }
+        }
+      } else {
+        try {
+          const provider = await resolveOcrProvider2(ocrMode, warnings, concurrency, batchSize);
+          ocrBlocks = await tryProvider(provider);
+        } catch (resolveErr) {
+          throw Object.assign(
+            new KordocError(resolveErr instanceof Error ? resolveErr.message : "OCR \uD504\uB85C\uBC14\uC774\uB354 \uCD08\uAE30\uD654 \uC2E4\uD328"),
+            { isImageBased: true }
+          );
         }
       }
-      if (ocrMode === "off") {
-        throw Object.assign(new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`), { isImageBased: true });
+      if (ocrBlocks.length > 0) {
+        const ocrMarkdown = blocksToMarkdown(ocrBlocks);
+        return {
+          markdown: ocrMarkdown,
+          blocks: ocrBlocks,
+          metadata,
+          warnings: warnings.length > 0 ? warnings : void 0,
+          isImageBased: true
+        };
       }
-      const errMsg = ocrMode ? `\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF \u2014 OCR \uC2E4\uD328 (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)` : `\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`;
-      throw Object.assign(new KordocError(errMsg), { isImageBased: true });
+      throw Object.assign(
+        new KordocError(`\uC774\uBBF8\uC9C0 \uAE30\uBC18 PDF \u2014 OCR \uC2E4\uD328 (${pageCount}\uD398\uC774\uC9C0, ${totalChars}\uC790)`),
+        { isImageBased: true }
+      );
     }
     if (options?.removeHeaderFooter !== false && parsedPageCount >= 3) {
       const removed = removeHeaderFooterBlocks(blocks, pageHeights, warnings);