@customize-agent/knowledge 4.0.13 → 4.0.15

package/dist/core/knowledge-base-manager.d.ts

@@ -1,5 +1,4 @@
 import { type EmbeddingProvider } from '../embedding/embedding-provider.js';
-import type { ExternalExtractorRegistry } from '../extraction/external-extractor.js';
 import type { LLMSearchProvider } from '../llm/llm-search-provider.js';
 import { type FederatedResult, type FederatedSearchItem, type RetrievalWeights, type SearchFilters } from '../search/federation-search.js';
 import type { DiffResult, IndexStateRecord, KBScope, KnowledgeBaseStats, ProjectConfig } from '../types.js';
@@ -23,7 +22,6 @@ export interface KnowledgeBaseManagerOptions {
     storageRoot?: string;
     embeddingProvider?: EmbeddingProvider;
     vectorStores?: Map<string, VectorStoreInterface>;
-    externalExtractors?: ExternalExtractorRegistry;
     onProgress?: (progress: KnowledgeIndexProgress) => void;
     /** 可选的 LLM Provider，用于查询扩展和语义重排序 */
     llmProvider?: LLMSearchProvider;

package/dist/core/knowledge-base-manager.js

@@ -43,7 +43,7 @@ export class KnowledgeBaseManager {
         this.projectId = options.projectId;
         this.embeddingProvider = options.embeddingProvider ?? createEmbeddingProviderFromEnvironment();
         this.vectorStores = options.vectorStores ?? new Map();
-        this.extractor = new ContentExtractor(options.externalExtractors);
+        this.extractor = new ContentExtractor();
         this.llmProvider = options.llmProvider;
         this.onProgress = options.onProgress;
         const storageRoot = options.storageRoot ?? path.join(os.homedir(), USER_DATA_DIR);

package/dist/extraction/content-extractor.d.ts

@@ -1,4 +1,3 @@
-import { ExternalExtractorRegistry } from './external-extractor.js';
 import type { ClassifiedFile } from '../types.js';
 /** 文件内容提取结果 */
 export interface ExtractionResult {
@@ -9,10 +8,7 @@ export interface ExtractionResult {
 }
 /** 文件内容提取器，支持文档、表格、图片、CAD 等多种文件格式的内容抽取 */
 export declare class ContentExtractor {
-    private readonly externalExtractors;
-    constructor(externalExtractors?: ExternalExtractorRegistry);
     extract(file: ClassifiedFile): Promise<ExtractionResult>;
-    private tryExternalExtractor;
     private extractTextClipping;
     private swapUtf16Bytes;
     private extractReadableFragments;
@@ -52,14 +48,17 @@ export declare class ContentExtractor {
     private extractOfficeZip;
     private extractRasterImage;
     private tryPaddleOcrLayout;
-    private parseOcrJson;
-    private formatOcrRegions;
-    private classifyOcrRegion;
     private formatBoundingBox;
     private validateRasterImage;
     private extractPdf;
-    private shouldAugmentPdfWithOcr;
-    private extractScannedPdfOcr;
+    private extractPdfHybridPages;
+    /** PyMuPDF 渲染（300 DPI 原生提取，质量远高于 pdfjs-dist） */
+    private tryRenderWithPyMuPDF;
+    /** pdfjs-dist + canvas 渲染（降级方案） */
+    private tryRenderWithPdfJs;
+    private scoreOcrText;
+    /** 加载图片像素数据（依赖 sharp） */
+    private loadImagePixels;
     private extractPdfText;
     private normalizedTextLength;
     private toPdfTextItem;