@customize-agent/knowledge 4.0.27 → 4.0.28

package/dist/extraction/ocr-providers.d.ts

@@ -49,6 +49,9 @@ export declare class TesseractJsProvider implements OcrProvider {
         filePath?: string;
     }): Promise<OcrResult>;
     getWarnings(): string[];
+    private pushWarning;
+    private recognizeWithTimeout;
+    private resetWorker;
     private readImageDimensions;
     private isTooSmallForOcr;
     private getWorker;

package/dist/extraction/ocr-providers.js

@@ -11,17 +11,12 @@ import { fileURLToPath } from 'node:url';
 import { resolveAndImport, resolvePackage } from './module-resolver.js';
 // ─── 路径工具 ───────────────────────────────────────────────────
 const knowledgeDir = path.dirname(fileURLToPath(import.meta.url));
+const OCR_NOISE_SUPPRESSION_KEY = Symbol.for('customize-agent.ocr-noise-suppression');
 const OCR_NATIVE_NOISE_PATTERNS = [
-    /Image too small to scale/u,
-    /Line cannot be recognized/u,
-    /empty image/iu
+    /^Image too small to scale!!(?:\s*\([^)]*\))?$/u,
+    /^Line cannot be recognized!!$/u,
+    /^empty image$/iu,
 ];
-const OCR_NOISE_SUPPRESSION_KEY = Symbol.for('customize-agent.ocr-noise-suppression');
-function ocrNoiseSuppressionState() {
-    const globalState = globalThis;
-    globalState[OCR_NOISE_SUPPRESSION_KEY] ??= { depth: 0 };
-    return globalState[OCR_NOISE_SUPPRESSION_KEY];
-}
 function tessdataDir() {
     if (process.env.TESSDATA_PREFIX)
         return process.env.TESSDATA_PREFIX;
@@ -30,78 +25,61 @@ function tessdataDir() {
         return pkg;
     return pkg;
 }
-function isNativeOcrNoise(chunk) {
-    const text = Buffer.isBuffer(chunk) ? chunk.toString('utf8') : chunk instanceof Uint8Array ? Buffer.from(chunk).toString('utf8') : typeof chunk === 'string' ? chunk : '';
-    if (!text)
-        return false;
-    return OCR_NATIVE_NOISE_PATTERNS.some(pattern => pattern.test(text));
+function ocrNoiseSuppressionState() {
+    const globalState = globalThis;
+    globalState[OCR_NOISE_SUPPRESSION_KEY] ??= {};
+    return globalState[OCR_NOISE_SUPPRESSION_KEY];
+}
+function textFromChunk(chunk) {
+    return Buffer.isBuffer(chunk) ? chunk.toString('utf8') : chunk instanceof Uint8Array ? Buffer.from(chunk).toString('utf8') : typeof chunk === 'string' ? chunk : '';
+}
+function isOcrNoiseLine(line) {
+    const text = line.trim();
+    return !!text && OCR_NATIVE_NOISE_PATTERNS.some(pattern => pattern.test(text));
 }
-async function suppressNativeOcrNoise(operation) {
+function filterNativeOcrNoiseText(text) {
+    const hasTrailingNewline = /\r?\n$/u.test(text);
+    const lines = text.split(/\r?\n/u);
+    const nonEmptyLines = lines.filter(line => line.trim());
+    if (nonEmptyLines.length > 0 && nonEmptyLines.every(isOcrNoiseLine))
+        return '';
+    const kept = lines.filter(line => !isOcrNoiseLine(line));
+    return kept.join('\n') + (hasTrailingNewline && kept.length > 0 ? '\n' : '');
+}
+function ensureOcrNoiseSuppressed() {
     const state = ocrNoiseSuppressionState();
-    if (state.depth === 0) {
-        state.stdout = process.stdout.write;
-        state.stderr = process.stderr.write;
-        const filter = (original) => function write(chunk, ...args) {
-            if (isNativeOcrNoise(chunk)) {
-                const callback = args.find((arg) => typeof arg === 'function');
-                if (callback)
-                    process.nextTick(callback);
-                return true;
-            }
+    if (state.installed)
+        return;
+    state.installed = true;
+    state.stdout = process.stdout.write;
+    state.stderr = process.stderr.write;
+    state.log = console.log;
+    state.warn = console.warn;
+    state.error = console.error;
+    const filterWrite = (original) => function write(chunk, ...args) {
+        const text = textFromChunk(chunk);
+        if (!text)
             return original.call(this, chunk, ...args);
-        };
-        process.stdout.write = filter(state.stdout);
-        process.stderr.write = filter(state.stderr);
-    }
-    state.depth += 1;
-    try {
-        return await operation();
-    }
-    finally {
-        state.depth -= 1;
-        if (state.depth === 0 && state.stdout && state.stderr) {
-            process.stdout.write = state.stdout;
-            process.stderr.write = state.stderr;
-            state.stdout = undefined;
-            state.stderr = undefined;
+        const filtered = filterNativeOcrNoiseText(text);
+        if (!filtered) {
+            const callback = args.find((arg) => typeof arg === 'function');
+            if (callback)
+                process.nextTick(callback);
+            return true;
         }
-    }
-}
-// 临时屏蔽 Tesseract 底层 WASM 污染日志
-function withSilencedConsole(fn) {
-    const originalLog = console.log;
-    const originalWarn = console.warn;
-    const originalError = console.error;
-    const filter = (orig) => (...args) => {
-        const msg = String(args[0] || '');
-        if (msg.includes('Line cannot be recognized') || msg.includes('Image too small to scale')) {
+        const nextChunk = typeof chunk === 'string' ? filtered : Buffer.from(filtered, 'utf8');
+        return original.call(this, nextChunk, ...args);
+    };
+    const filterConsole = (original) => (...args) => {
+        if (args.length === 1 && typeof args[0] === 'string' && isOcrNoiseLine(args[0]))
             return;
-        }
-        orig(...args);
+        original(...args);
     };
-    console.log = filter(originalLog);
-    console.warn = filter(originalWarn);
-    console.error = filter(originalError);
-    try {
-        const result = fn();
-        if (result instanceof Promise) {
-            return result.finally(() => {
-                console.log = originalLog;
-                console.warn = originalWarn;
-                console.error = originalError;
-            });
-        }
-        console.log = originalLog;
-        console.warn = originalWarn;
-        console.error = originalError;
-        return result;
-    }
-    catch (error) {
-        console.log = originalLog;
-        console.warn = originalWarn;
-        console.error = originalError;
-        throw error;
-    }
+    process.stdout.write = filterWrite(state.stdout);
+    process.stderr.write = filterWrite(state.stderr);
+    console.log = filterConsole(state.log);
+    console.warn = filterConsole(state.warn);
+    console.error = filterConsole(state.error);
 }
 export class TesseractJsProvider {
     id = 'tesseract.js';
@@ -154,7 +132,7 @@ export class TesseractJsProvider {
             return { text: '', confidence: 0, regions: [], warnings: [`image too small for OCR: ${width}x${height}`] };
         }
         try {
-            const worker = await withSilencedConsole(() => this.getWorker());
+            ensureOcrNoiseSuppressed();
             let unlock;
             const nextLock = new Promise(resolve => { unlock = resolve; });
             const currentLock = this.workerLock;
@@ -162,7 +140,8 @@ export class TesseractJsProvider {
             await currentLock;
             let result;
             try {
-                result = await withSilencedConsole(() => suppressNativeOcrNoise(() => worker.recognize(pngPath)));
+                const worker = await this.getWorker();
+                result = await this.recognizeWithTimeout(worker, pngPath);
             }
             finally {
                 unlock();
@@ -191,6 +170,41 @@ export class TesseractJsProvider {
     getWarnings() {
         return [...new Set(this.warnings)].slice(-20);
     }
+    pushWarning(message) {
+        const text = message.trim();
+        if (!text || isOcrNoiseLine(text))
+            return;
+        this.warnings.push(text);
+        if (this.warnings.length > 50)
+            this.warnings = this.warnings.slice(-50);
+    }
+    async recognizeWithTimeout(worker, imagePath) {
+        const timeoutMs = 120_000;
+        let timeout;
+        try {
+            return await Promise.race([
+                worker.recognize(imagePath),
+                new Promise((_, reject) => {
+                    timeout = setTimeout(() => reject(new Error(`OCR recognition timed out after ${timeoutMs}ms`)), timeoutMs);
+                }),
+            ]);
+        }
+        catch (error) {
+            await this.resetWorker().catch(() => undefined);
+            throw error;
+        }
+        finally {
+            if (timeout)
+                clearTimeout(timeout);
+        }
+    }
+    async resetWorker() {
+        const worker = this.worker;
+        this.worker = null;
+        this.workerPromise = null;
+        if (worker)
+            await worker.terminate();
+    }
     async readImageDimensions(filePath) {
         try {
             const sharpMod = await resolveAndImport('sharp');
@@ -210,8 +224,12 @@ export class TesseractJsProvider {
     async getWorker() {
         if (this.worker)
             return this.worker;
-        if (!this.workerPromise)
-            this.workerPromise = this.createReusableWorker();
+        if (!this.workerPromise) {
+            this.workerPromise = this.createReusableWorker().catch(error => {
+                this.workerPromise = null;
+                throw error;
+            });
+        }
         this.worker = await this.workerPromise;
         return this.worker;
     }
@@ -225,9 +243,16 @@ export class TesseractJsProvider {
             gzip: false,
             logger: () => undefined,
             errorHandler: (error) => {
-                const message = error instanceof Error ? error.message : String(error);
-                if (message.trim())
-                    this.warnings.push(message.trim());
+                let message = error instanceof Error ? error.message : typeof error === 'string' ? error : '';
+                if (!message) {
+                    try {
+                        message = JSON.stringify(error);
+                    }
+                    catch {
+                        message = String(error);
+                    }
+                }
+                this.pushWarning(message);
             },
         });
         if (typeof worker.setParameters === 'function') {
@@ -236,10 +261,17 @@ export class TesseractJsProvider {
         return worker;
     }
     async dispose() {
-        if (this.worker)
-            await this.worker.terminate();
-        this.worker = null;
-        this.workerPromise = null;
+        let unlock;
+        const nextLock = new Promise(resolve => { unlock = resolve; });
+        const currentLock = this.workerLock;
+        this.workerLock = currentLock.then(() => nextLock).catch(() => nextLock);
+        await currentLock;
+        try {
+            await this.resetWorker();
+        }
+        finally {
+            unlock();
+        }
     }
 }
 // ─── 工厂 ───────────────────────────────────────────────────────

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@customize-agent/knowledge",
-  "version": "4.0.27",
+  "version": "4.0.28",
   "description": "Local knowledge base infrastructure for customize-agent",
   "type": "module",
   "main": "./dist/index.js",