smart-image-scraper-mcp 2.13.3 → 2.14.0

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "smart-image-scraper-mcp",
-  "version": "2.13.3",
+  "version": "2.14.0",
   "description": "全网智能图片抓取 MCP 服务器 - 支持 Bing/Google 搜索、三种模式(link/download/both)、自定义保存路径、防盗链处理、质量控制、尺寸统一",
   "main": "src/index.js",
   "type": "module",

package/src/providers/bingScraper.js

@@ -74,9 +74,8 @@ export class BingScraper extends BaseScraper {
       const response = await withRetry(
         () => httpClient.get(searchUrl, { 
           timeout: 8000,  // 8秒超时，快速失败
-          signal: AbortSignal.timeout(8000),
         }),
-        { maxRetries: 1, retryCondition: isRetryableError } // 只重试1次
+        { maxRetries: 1, retryCondition: (e) => e.code === 'ECONNRESET' || (e.response?.status >= 500) } // 只重试真正的网络重置或服务器错误，超时不重试
       );
       
       if (response.status !== 200) {

package/src/providers/googleScraper.js

@@ -73,7 +73,6 @@ export class GoogleScraper extends BaseScraper {
       const response = await withRetry(
         () => httpClient.get(searchUrl, {
           timeout: 8000,  // 8秒超时，快速失败
-          signal: AbortSignal.timeout(8000),
           headers: {
             'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36',
             'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',
@@ -81,7 +80,7 @@ export class GoogleScraper extends BaseScraper {
             'Referer': 'https://www.google.com/',
           },
         }),
-        { maxRetries: 1, retryCondition: isRetryableError } // 只重试1次
+        { maxRetries: 1, retryCondition: (e) => e.code === 'ECONNRESET' || (e.response?.status >= 500) } // 只重试真正的网络重置或服务器错误，超时不重试
       );
       
       if (response.status !== 200) {

package/src/services/fileManager.js

@@ -11,13 +11,10 @@ import httpClient from '../infrastructure/httpClient.js';
 import logger from '../infrastructure/logger.js';
 import config from '../config/index.js';
 
-// 全局共享的下载并发限制器（避免每个实例独立限制）
-const globalDownloadLimit = pLimit(config.MAX_DOWNLOAD_CONCURRENCY || 10);
-
 export class FileManager {
   constructor(options = {}) {
     this.saveRoot = options.saveRoot || config.SAVE_ROOT;
-    this.limit = globalDownloadLimit; // 使用全局共享限制器
+    this.limit = pLimit(config.MAX_DOWNLOAD_CONCURRENCY || 3);
   }
 
   /**
@@ -183,7 +180,6 @@ export class FileManager {
         maxContentLength: 50 * 1024 * 1024, // 最大50MB
         maxBodyLength: 50 * 1024 * 1024,
         headers: downloadHeaders,
-        signal: AbortSignal.timeout(12000), // 12秒硬超时，防止连接挂起
       });
 
       if (response.status !== 200) {
@@ -305,13 +301,22 @@ export class FileManager {
    * 批量下载图片
    * @param {string[]} urls - URL列表
    * @param {string} keyword - 搜索关键词
+   * @param {number} maxSuccess - 达到此成功数后跳过剩余下载（提前退出）
    * @returns {Promise<{success: Array, failed: Array}>}
    */
-  async downloadMany(urls, keyword) {
+  async downloadMany(urls, keyword, maxSuccess = Infinity) {
     logger.info(`Downloading ${urls.length} images for "${keyword}" with concurrency ${config.MAX_DOWNLOAD_CONCURRENCY}`);
 
+    let successCount = 0;
     const results = await Promise.all(
-      urls.map(url => this.limit(() => this.downloadOne(url, keyword)))
+      urls.map(url => this.limit(async () => {
+        if (successCount >= maxSuccess) {
+          return { success: false, url, error: 'skipped' };
+        }
+        const result = await this.downloadOne(url, keyword);
+        if (result.success) successCount++;
+        return result;
+      }))
     );
 
     const success = [];

package/src/services/imageProcessor.js

@@ -9,10 +9,12 @@ import path from 'path';
 import pLimit from 'p-limit';
 import logger from '../infrastructure/logger.js';
 
-// sharp 是 CPU 密集型操作，限制并发避免卡死
-const imageProcessLimit = pLimit(2);
 
 export class ImageProcessor {
+  constructor() {
+    // sharp 是 CPU 密集型操作，限制并发避免卡死；每个实例独立防止跨请求堵塞
+    this._limit = pLimit(2);
+  }
   /**
    * 预设尺寸配置
    */
@@ -64,15 +66,16 @@ export class ImageProcessor {
         return { success: false, path: inputPath, error: '文件不存在' };
       }
 
-      // 读取原始图片信息
-      const metadata = await sharp(inputPath).metadata();
+      // 用 Buffer 读取，避免 Windows 文件句柄锁导致后续 move 失败
+      const inputBuffer = await fs.readFile(inputPath);
+      const metadata = await sharp(inputBuffer).metadata();
       
       // 确定输出路径
       const finalOutputPath = outputPath || inputPath;
       const tempPath = inputPath + '.tmp';
 
       // 处理图片
-      await sharp(inputPath)
+      await sharp(inputBuffer)
         .resize(width, height, {
           fit: fit,
           position: position,
@@ -131,7 +134,7 @@ export class ImageProcessor {
    */
   async processMany(files, options = {}) {
     const results = await Promise.all(
-      files.map(file => imageProcessLimit(() => this.processOne(file.path, options)))
+      files.map(file => this._limit(() => this.processOne(file.path, options)))
     );
 
     const success = [];

package/src/services/linkValidator.js

@@ -8,9 +8,6 @@ import httpClient from '../infrastructure/httpClient.js';
 import logger from '../infrastructure/logger.js';
 import config from '../config/index.js';
 
-// 使用配置中的并发数，避免硬编码与配置不一致
-const globalValidateLimit = pLimit(config.MAX_VALIDATE_CONCURRENCY);
-
 // 已知有严格防盗链保护的域名列表（仅包含确实无法在浏览器直接打开的）
 // 这些域名的图片在浏览器地址栏直接打开会返回 403、替换图或空白
 const HOTLINK_PROTECTED_DOMAINS = [
@@ -27,7 +24,7 @@ const HOTLINK_PROTECTED_DOMAINS = [
 
 export class LinkValidator {
   constructor() {
-    this.limit = globalValidateLimit;
+    this.limit = pLimit(config.MAX_VALIDATE_CONCURRENCY);
   }
 
   /**

package/src/services/orchestrator.js

@@ -320,8 +320,8 @@ export class Orchestrator {
       // 检查是否已中止
       if (signal?.aborted) throw new Error('操作已取消');
 
-      // 根据 quality 模式处理（限制验证数量避免超时）
-      const maxValidate = Math.min(rawUrls.length, count * 2 + 5);
+      // 根据 quality 模式处理（只多取少量备用，避免下载过多）
+      const maxValidate = Math.min(rawUrls.length, count + 3);
       let urlsToDownload = rawUrls.slice(0, maxValidate);
       if (prioritizeQuality) {
         const sortByQuality = qualityMode === 'high';
@@ -347,8 +347,8 @@ export class Orchestrator {
         logger.info(`Quality filtered: ${urlsToDownload.length} valid URLs`);
       }
 
-      // 下载图片（已按质量排序，高质量优先）
-      const { success, failed } = await this.fileManager.downloadMany(urlsToDownload, keyword);
+      // 下载图片（已按质量排序，高质量优先），达到目标数量后跳过剩余
+      const { success, failed } = await this.fileManager.downloadMany(urlsToDownload, keyword, count);
       
       // 截取需要的数量
       let resultDownloads = success.slice(0, count);
@@ -490,11 +490,11 @@ export class Orchestrator {
       
       if (fastMode) {
         // fast 模式：不验证
-        allValidUrls = rawUrls.slice(0, count * 2 + 5);
+        allValidUrls = rawUrls.slice(0, count + 3);
         qualityModeLabel = '快速模式（跳过验证）';
       } else {
         // balanced/high 模式：统一验证所有原始 URL
-        const maxValidate = Math.min(rawUrls.length, count * 2 + 5);
+        const maxValidate = Math.min(rawUrls.length, count + 3);
         const urlsToValidate = rawUrls.slice(0, maxValidate);
         const { valid } = await this.linkValidator.validateMany(urlsToValidate, {
           fetchQuality: prioritizeQuality,
@@ -528,11 +528,11 @@ export class Orchestrator {
       }
 
       // 下载列表：使用所有验证通过的 URL（不过滤防盗链，下载时加 Referer 绕过）
-      downloadUrls = allValidUrls.slice(0, count * 2);
+      downloadUrls = allValidUrls.slice(0, count + 3);
 
       if (signal?.aborted) throw new Error('操作已取消');
 
-      const { success, failed } = await this.fileManager.downloadMany(downloadUrls, keyword);
+      const { success, failed } = await this.fileManager.downloadMany(downloadUrls, keyword, count);
       
       let resultDownloads = success.slice(0, count);