@customize-agent/knowledge 4.1.5 → 4.2.0

package/dist/chunking/text-chunker.d.ts

@@ -13,7 +13,6 @@ export interface TextChunk {
 export interface ChunkConfig {
     maxChunkSize: number;
     overlap: number;
-    headerInjection: boolean;
 }
 /** 文本切片器，支持文档、表格、代码等多类型文件的递归式切片 */
 export declare class TextChunker {
@@ -56,7 +55,6 @@ export declare class TextChunker {
     private chunkPartStartOffsets;
     private enforceCandidateLimit;
     private buildTitlePaths;
-    private withHeader;
     private createChunk;
     private kindForCategory;
     private extractSectionTitle;

package/dist/chunking/text-chunker.js

@@ -1,15 +1,15 @@
 import { BgeTokenizer } from './bge-tokenizer.js';
 const DEFAULT_CONFIGS = {
-    document: { maxChunkSize: 800, overlap: 100, headerInjection: true },
-    spreadsheet: { maxChunkSize: 1000, overlap: 120, headerInjection: true },
-    image: { maxChunkSize: 512, overlap: 0, headerInjection: true },
-    cad: { maxChunkSize: 600, overlap: 80, headerInjection: true },
-    code: { maxChunkSize: 1000, overlap: 120, headerInjection: true },
-    data: { maxChunkSize: 600, overlap: 80, headerInjection: true },
-    web: { maxChunkSize: 800, overlap: 100, headerInjection: true },
-    diagram: { maxChunkSize: 512, overlap: 0, headerInjection: true },
-    archive: { maxChunkSize: 500, overlap: 50, headerInjection: false },
-    other: { maxChunkSize: 500, overlap: 50, headerInjection: false },
+    document: { maxChunkSize: 800, overlap: 100 },
+    spreadsheet: { maxChunkSize: 1000, overlap: 120 },
+    image: { maxChunkSize: 512, overlap: 0 },
+    cad: { maxChunkSize: 600, overlap: 80 },
+    code: { maxChunkSize: 1000, overlap: 120 },
+    data: { maxChunkSize: 600, overlap: 80 },
+    web: { maxChunkSize: 800, overlap: 100 },
+    diagram: { maxChunkSize: 512, overlap: 0 },
+    archive: { maxChunkSize: 500, overlap: 50 },
+    other: { maxChunkSize: 500, overlap: 50 },
 };
 const RECURSIVE_SEPARATORS = [
     /\n(?=#{1,6}\s)/u,
@@ -44,7 +44,7 @@ export class TextChunker {
         if (source.length === 0)
             return [];
         const config = DEFAULT_CONFIGS[file.category];
-        const normalized = this.withHeader(source, file, config);
+        const normalized = source;
         const rawCandidates = this.createCandidates(normalized, file, config);
         const candidates = this.enforceCandidateLimit(rawCandidates, config);
         return candidates.map((candidate, index) => this.createChunk(index, candidate, file, metadata));
@@ -493,11 +493,6 @@ export class TextChunker {
             return stack.map(item => item.title).join(' > ') || section.title;
         });
     }
-    withHeader(text, file, config) {
-        if (!config.headerInjection)
-            return text;
-        return `资料类型: ${file.category}/${file.format}\n\n${text}`;
-    }
     createChunk(index, candidate, file, metadata) {
         const text = candidate.text.trim();
         return {

package/dist/cleaning/text-cleaner.d.ts

@@ -44,6 +44,8 @@ export interface TextCleanStats {
     cadTitleBlockLines: number;
     /** 图纸：CAD 属性行数（图层/颜色/线型，K3） */
     cadAttributeLines: number;
+    /** 图纸：CAD 图元属性枚举行数（管道符表格形态的实体罗列，K3） */
+    cadEntityPropertyLines: number;
     /** 清单：纯报价表格段行数（费汇总/暂估价/规费/税金，K3） */
     billPricingLines: number;
     /** 清单：扉页签章段行数（K3） */

package/dist/cleaning/text-cleaner.js

@@ -3,7 +3,7 @@ function zeroStats() {
         headerFooterLines: 0, pageNumberLines: 0, tocRegionLines: 0, blankLines: 0,
         tenderFormatLines: 0, tenderGenericLines: 0, clarificationLines: 0, cadNoiseLines: 0,
         contractGeneralClauseLines: 0, announcementProcedureLines: 0, businessReviewLines: 0,
-        cadTitleBlockLines: 0, cadAttributeLines: 0, billPricingLines: 0, billTitlePageLines: 0,
+        cadTitleBlockLines: 0, cadAttributeLines: 0, cadEntityPropertyLines: 0, billPricingLines: 0, billTitlePageLines: 0,
     };
 }
 /** 文档类型判定：文件名 + 文首内容双重特征（招标文件 / 补疑答疑 / 工程量清单 / 其他） */
@@ -145,6 +145,23 @@ function isCadAttributeLine(trimmed) {
         return false;
     return true;
 }
+/**
+ * 图纸：CAD 图元属性枚举行（「| 图层: X | 块: Y | 实体类型: … | 坐标: (…)」
+ * 管道符表格形态或「└── 标注文本: …」图元注释行）——专业转换器/历史版本产物形态，
+ * 图元属性包装与施组编制无关，「坐标/关联对象/状态」类纯图元定位信息会稀释检索语义。
+ * 行首键名 + 冒号 + 短值多重特征判定；「图纸节点: 文件名」锚定行不删（文件溯源有用）。
+ */
+function isCadEntityPropertyLine(trimmed) {
+    if (trimmed.length > 50)
+        return false;
+    if (/^[|└\s]*(?:图层|块|实体类型|坐标|关联对象|状态)\s*[:：]/u.test(trimmed))
+        return true;
+    if (/^└──\s*(?:标注文本\s*[:：])?/u.test(trimmed))
+        return true;
+    if (/^图纸节点\s*[:：].{0,100}(?:图层|块|实体类型)\s*[:：]/u.test(trimmed))
+        return true;
+    return false;
+}
 /** 合同通用条款章节标题（GF 系列示范文本的标准通用条款，不含项目数据） */
 const CONTRACT_GENERAL_TITLE_RE = /通用合同条款|合同通用条款/u;
 /** 专用合同条款标题：项目数据所在章节，必须保留，同时作为通用条款章节的边界 */
@@ -398,6 +415,12 @@ export function cleanExtractedText(input) {
             index += 1;
             continue;
         }
+        // 10. 图纸：CAD 图元属性枚举行（管道符表格形态实体罗列）
+        if (isCad && isCadEntityPropertyLine(trimmed)) {
+            drop(line, s => { s.cadEntityPropertyLines += 1; });
+            index += 1;
+            continue;
+        }
         kept.push(line);
         index += 1;
     }

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

@@ -23,8 +23,6 @@ export declare class ContentExtractor {
     private extractDxf;
     private buildCadSemanticNodes;
     private extractDxfTextAnnotations;
-    private findNearestCadAnnotation;
-    private inferCadEntityType;
     private tryConvertDwgWithBundledWasm;
     private tryConvertDwgToDxf;
     private extractCadMesh;

package/dist/extraction/content-extractor.js

@@ -120,7 +120,7 @@ export class ContentExtractor {
             .slice(0, 50);
         const text = unique.join('\n');
         return {
-            text: text ? [this.metadataOnlyText(file), text].join('\n') : this.metadataOnlyText(file),
+            text: text ?? this.metadataOnlyText(file),
             metadata: {
                 extractionMode: 'builtin_text_clipping',
                 vectorizable: true,
@@ -348,14 +348,13 @@ export class ContentExtractor {
             }
             metadata.contentCoverage = 'dxf_semantic_layer_block_annotations';
             metadata.characterDataCount = characterDataCount;
-            const semanticNodes = this.buildCadSemanticNodes(file, uniqueLayers, uniqueBlocks, uniqueEntityTypes, textEntities);
+            const semanticNodes = this.buildCadSemanticNodes(file, textEntities);
             return {
                 text: [
-                    this.metadataOnlyText(file),
                     `CAD DXF 图层: ${uniqueLayers.join(', ')}`,
                     `CAD DXF 块/符号: ${uniqueBlocks.join(', ')}`,
                     `CAD DXF 实体类型: ${uniqueEntityTypes.join(', ')}`,
-                    'CAD 语义图纸节点:',
+                    'CAD 语义标注文本:',
                     ...semanticNodes,
                 ].join('\n'),
                 metadata,
@@ -378,7 +377,6 @@ export class ContentExtractor {
             metadata.contentCoverage = 'step_products_materials_entities_names';
             return {
                 text: [
-                    this.metadataOnlyText(file),
                     `STEP 产品/零件:\n${products.join('\n')}`,
                     `STEP 材料: ${materials.join(', ')}`,
                     `STEP 实体类型: ${uniqueStepEntities.join(', ')}`,
@@ -399,7 +397,7 @@ export class ContentExtractor {
             metadata.entityTypes = uniqueIgesTypes.slice(0, 80);
             metadata.contentCoverage = 'iges_entity_names_types';
             return {
-                text: [this.metadataOnlyText(file), `IGES 实体类型: ${uniqueIgesTypes.join(', ')}`, `IGES 实体/名称:\n${names.join('\n')}`].join('\n'),
+                text: [`IGES 实体类型: ${uniqueIgesTypes.join(', ')}`, `IGES 实体/名称:\n${names.join('\n')}`].join('\n'),
                 metadata,
                 warnings,
             };
@@ -462,7 +460,7 @@ export class ContentExtractor {
         else
             warnings.push(`${file.format} 内置 DWG→DXF 转换未成功，已使用低置信度可读标注/标题块兜底抽取并过滤疑似乱码 ${filteredCount} 条；该结果仅作为兜底证据，不应等同于完整图层、块、标注和尺寸语义解析`);
         return {
-            text: hasCharacterData ? [this.metadataOnlyText(file), `CAD 图纸可读标注/标题块/属性:\n${readable.join('\n')}`].join('\n') : this.metadataOnlyText(file),
+            text: hasCharacterData ? `CAD 图纸可读标注/标题块/属性:\n${readable.join('\n')}` : this.metadataOnlyText(file),
             metadata,
             warnings,
         };
@@ -503,35 +501,29 @@ export class ContentExtractor {
         }
         metadata.contentCoverage = 'dxf_semantic_layer_block_annotations';
         metadata.characterDataCount = characterDataCount;
-        const semanticNodes = this.buildCadSemanticNodes(file, uniqueLayers, uniqueBlocks, uniqueEntityTypes, textEntities);
+        const semanticNodes = this.buildCadSemanticNodes(file, textEntities);
         return {
             text: [
-                this.metadataOnlyText(file),
                 `CAD DXF 图层: ${uniqueLayers.join(', ')}`,
                 `CAD DXF 块/符号: ${uniqueBlocks.join(', ')}`,
                 `CAD DXF 实体类型: ${uniqueEntityTypes.join(', ')}`,
-                'CAD 语义图纸节点:',
+                'CAD 语义标注文本:',
                 ...semanticNodes,
             ].join('\n'),
             metadata,
             warnings,
         };
     }
-    buildCadSemanticNodes(file, layers, blocks, entityTypes, texts) {
+    buildCadSemanticNodes(file, texts) {
+        // 图纸语义节点 = 文件名锚定 + 纯标注文本列表。逐实体枚举的图元属性包装
+        // （图层/块/实体类型/坐标/关联对象/状态）是重复结构模板噪音，会稀释检索语义
+        // （如「混凝土强度 C35」被「实体类型:|坐标:(…)」图元罗列干扰）；图层/块/实体
+        // 类型汇总已在 CAD DXF 汇总行体现，此处只保留图纸真实文字（图名/说明/尺寸标注值）
         const fileName = path.basename(file.relativePath);
-        const defaultLayer = layers[0] ?? '未命名图层';
-        const defaultBlock = blocks[0] ?? '全局模型空间';
-        const defaultEntity = entityTypes.find(type => /DIMENSION|TEXT|MTEXT|LEADER/u.test(type)) ?? entityTypes[0] ?? 'UNKNOWN';
-        const annotations = texts.length > 0 ? texts : [{ text: '未提取到文字标注', layer: defaultLayer, block: defaultBlock, entityType: defaultEntity }];
-        return annotations.map((annotation, index) => {
-            const nearest = this.findNearestCadAnnotation(annotation, annotations);
-            const layer = annotation.layer ?? layers[index % Math.max(1, layers.length)] ?? defaultLayer;
-            const block = annotation.block ?? blocks[index % Math.max(1, blocks.length)] ?? defaultBlock;
-            const entity = annotation.entityType ?? this.inferCadEntityType(annotation.text, defaultEntity);
-            const status = /关键|critical|尺寸|dim|mm|cm|m\b|°|φ|Φ|R\d/iu.test(annotation.text) ? '关键尺寸/约束候选' : '普通标注';
-            const position = annotation.x != null && annotation.y != null ? ` | 坐标: (${annotation.x.toFixed(2)}, ${annotation.y.toFixed(2)})` : '';
-            return [`图纸节点: ${fileName} | 图层: ${layer} | 块: ${block} | 实体类型: ${entity}${position}`, `└── 标注文本: ${annotation.text} | 关联对象: ${nearest ? `邻近标注 ${nearest.text}` : '空间邻近候选'} | 状态: ${status}`].join('\n');
-        });
+        const annotations = texts.filter(item => item.text && item.text.trim().length > 0);
+        if (annotations.length === 0)
+            return [`图纸节点: ${fileName} | 未提取到文字标注`];
+        return [`图纸节点: ${fileName}`, ...annotations.map(item => item.text)];
     }
     extractDxfTextAnnotations(raw) {
         const entities = raw.split(/(?:^|\r?\n)\s*0\s*\r?\n/u).filter(section => /^(?:TEXT|MTEXT|DIMENSION|LEADER)/u.test(section.trim()));
@@ -553,23 +545,6 @@ export class ContentExtractor {
                 }].map(item => ({ ...item, x: Number.isFinite(item.x) ? item.x : undefined, y: Number.isFinite(item.y) ? item.y : undefined }));
         });
     }
-    findNearestCadAnnotation(target, annotations) {
-        if (target.x == null || target.y == null)
-            return undefined;
-        return annotations
-            .filter(item => item !== target && item.x != null && item.y != null)
-            .map(item => ({ item, distance: Math.hypot((item.x ?? 0) - target.x, (item.y ?? 0) - target.y) }))
-            .sort((a, b) => a.distance - b.distance)[0]?.item;
-    }
-    inferCadEntityType(text, fallback) {
-        if (/\b\d+(?:\.\d+)?\s*(?:mm|cm|m)\b|φ|Φ|R\d/iu.test(text))
-            return '线性尺寸/半径尺寸';
-        if (/°|angle|角度/iu.test(text))
-            return '角度尺寸';
-        if (/note|说明|备注/iu.test(text))
-            return '文字说明';
-        return fallback;
-    }
     async tryConvertDwgWithBundledWasm(filePath) {
         try {
             const mod = await resolveAndImport('dwgdxf');

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@customize-agent/knowledge",
-  "version": "4.1.5",
+  "version": "4.2.0",
   "description": "Local knowledge base infrastructure for customize-agent",
   "type": "module",
   "main": "./dist/index.js",