autosnippet 2.8.3 → 2.10.0

package/lib/service/search/SearchEngine.js

@@ -15,6 +15,8 @@ const BM25_B = 0.75;
 
 /**
  * 分词: 中英文混合分词
+ * 英文: camelCase / PascalCase 拆分 + 小写化
+ * 中文: 单字 + 二元组（bigram）— 无需分词词典即可支持子串匹配
  */
 export function tokenize(text) {
   if (!text) return [];
@@ -23,7 +25,35 @@ export function tokenize(text) {
   // 拆全大写前缀：URLSession → URL Session, UITableView → UI Table View
   expanded = expanded.replace(/([A-Z]+)([A-Z][a-z])/g, '$1 $2');
   const normalized = expanded.toLowerCase().replace(/[^\p{L}\p{N}\s_-]/gu, ' ');
-  const tokens = normalized.split(/[\s_-]+/).filter(t => t.length >= 2);
+  const rawTokens = normalized.split(/[\s_-]+/).filter(t => t.length >= 1);
+
+  const tokens = [];
+  // CJK 正则（中日韩统一表意文字 + 扩展区）
+  const cjkRe = /[\u4e00-\u9fff\u3400-\u4dbf\uf900-\ufaff]/;
+
+  for (const raw of rawTokens) {
+    if (cjkRe.test(raw)) {
+      // 中文片段：提取所有 CJK 连续子串，生成单字 + bigram 覆盖
+      const cjkChars = raw.match(/[\u4e00-\u9fff\u3400-\u4dbf\uf900-\ufaff]+/g) || [];
+      for (const seg of cjkChars) {
+        // 单字
+        for (const ch of seg) tokens.push(ch);
+        // bigram
+        for (let i = 0; i < seg.length - 1; i++) tokens.push(seg[i] + seg[i + 1]);
+        // 完整片段（≥3 字时额外保留，提升精确匹配权重）
+        if (seg.length >= 3) tokens.push(seg);
+      }
+      // 非 CJK 部分（英文/数字）也保留
+      const nonCjk = raw.replace(/[\u4e00-\u9fff\u3400-\u4dbf\uf900-\ufaff]+/g, ' ').trim();
+      if (nonCjk) {
+        for (const t of nonCjk.split(/\s+/)) {
+          if (t.length >= 2) tokens.push(t);
+        }
+      }
+    } else if (raw.length >= 2) {
+      tokens.push(raw);
+    }
+  }
   return [...new Set(tokens)];
 }
 
@@ -126,13 +156,33 @@ export class SearchEngine {
     this._cache.clear();
 
     try {
-      // 索引 Recipes（统一模型，包含所有知识类型）
-      const recipes = this.db.prepare(
-        `SELECT id, title, description, language, category, knowledge_type, kind, content_json, status, tags_json, "trigger"
-         FROM recipes WHERE status != 'deprecated'`
-      ).all();
+      let entries = [];
+
+      // V3: 优先从 knowledge_entries 表索引
+      try {
+        entries = this.db.prepare(
+          `SELECT id, title, description, language, category, knowledge_type, kind,
+                  content AS content_json, lifecycle, tags AS tags_json, trigger_key
+           FROM knowledge_entries WHERE lifecycle != 'deprecated'`
+        ).all();
+        // 标准化字段名以便后续处理
+        entries = entries.map(e => ({
+          ...e,
+          trigger: e.trigger_key,
+          status: e.lifecycle,
+        }));
+      } catch {
+        // V3 表不存在，回退到 V2 recipes 表
+        try {
+          entries = this.db.prepare(
+            `SELECT id, title, description, language, category, knowledge_type, kind,
+                    content_json, status, tags_json, "trigger"
+             FROM recipes WHERE status != 'deprecated'`
+          ).all();
+        } catch { /* neither table exists */ }
+      }
 
-      for (const r of recipes) {
+      for (const r of entries) {
         let contentText = '';
         try {
           const content = JSON.parse(r.content_json || '{}');
@@ -143,12 +193,12 @@ export class SearchEngine {
         try { tagText = JSON.parse(r.tags_json || '[]').join(' '); } catch { /* ignore */ }
         const text = [r.title, r.description, r.trigger, r.language, r.category, r.knowledge_type, tagText, contentText]
           .filter(Boolean).join(' ');
-        this.scorer.addDocument(r.id, text, { type: 'recipe', title: r.title, trigger: r.trigger || '', status: r.status, knowledgeType: r.knowledge_type, kind: r.kind || 'pattern', language: r.language || '', category: r.category || '' });
+        this.scorer.addDocument(r.id, text, { type: 'knowledge', title: r.title, trigger: r.trigger || '', status: r.status, knowledgeType: r.knowledge_type, kind: r.kind || 'pattern', language: r.language || '', category: r.category || '' });
       }
 
       this._indexed = true;
       this.logger.info('Search index built', {
-        recipes: recipes.length,
+        entries: entries.length,
         total: this.scorer.totalDocs,
       });
     } catch (err) {
@@ -229,14 +279,26 @@ export class SearchEngine {
     const escaped = query.replace(/[%_\\]/g, ch => `\\${ch}`);
     const pattern = `%${escaped}%`;
 
-    if (type === 'all' || type === 'recipe' || type === 'rule' || type === 'solution') {
+    if (type === 'all' || type === 'recipe' || type === 'knowledge' || type === 'rule' || type === 'solution') {
       try {
-        const rows = this.db.prepare(
-          `SELECT id, title, description, language, category, knowledge_type, kind, status, content_json, dimensions_json, "trigger", 'recipe' as type
-           FROM recipes
-           WHERE status != 'deprecated' AND (title LIKE ? ESCAPE '\\' OR description LIKE ? ESCAPE '\\' OR "trigger" LIKE ? ESCAPE '\\' OR content_json LIKE ? ESCAPE '\\')
-           LIMIT ?`
-        ).all(pattern, pattern, pattern, pattern, limit);
+        let rows = [];
+        // V3: 优先 knowledge_entries
+        try {
+          rows = this.db.prepare(
+            `SELECT id, title, description, language, category, knowledge_type, kind, lifecycle as status, content AS content_json, trigger_key as "trigger", headers, module_name, 'knowledge' as type
+             FROM knowledge_entries
+             WHERE lifecycle != 'deprecated' AND (title LIKE ? ESCAPE '\\' OR description LIKE ? ESCAPE '\\' OR trigger_key LIKE ? ESCAPE '\\' OR content LIKE ? ESCAPE '\\')
+             LIMIT ?`
+          ).all(pattern, pattern, pattern, pattern, limit);
+        } catch {
+          // V3 表不存在，回退到 V2
+          rows = this.db.prepare(
+            `SELECT id, title, description, language, category, knowledge_type, kind, status, content_json, dimensions_json, "trigger", 'recipe' as type
+             FROM recipes
+             WHERE status != 'deprecated' AND (title LIKE ? ESCAPE '\\' OR description LIKE ? ESCAPE '\\' OR "trigger" LIKE ? ESCAPE '\\' OR content_json LIKE ? ESCAPE '\\')
+             LIMIT ?`
+          ).all(pattern, pattern, pattern, pattern, limit);
+        }
         // 基础相关性排序：trigger 精确 > 标题匹配 > 描述匹配 > 内容匹配
         const lowerQ = query.toLowerCase();
         results.push(...rows.map(r => {
@@ -349,9 +411,17 @@ export class SearchEngine {
     try {
       const ids = items.map(it => it.id);
       const placeholders = ids.map(() => '?').join(',');
-      const rows = this.db.prepare(
-        `SELECT id, content_json, dimensions_json, description, "trigger" FROM recipes WHERE id IN (${placeholders})`
-      ).all(...ids);
+      let rows = [];
+      // V3: 优先 knowledge_entries
+      try {
+        rows = this.db.prepare(
+          `SELECT id, content AS content_json, description, trigger_key as "trigger", headers, module_name FROM knowledge_entries WHERE id IN (${placeholders})`
+        ).all(...ids);
+      } catch {
+        rows = this.db.prepare(
+          `SELECT id, content_json, dimensions_json, description, "trigger" FROM recipes WHERE id IN (${placeholders})`
+        ).all(...ids);
+      }
       const rowMap = new Map(rows.map(r => [r.id, r]));
       for (const item of items) {
         const row = rowMap.get(item.id);
@@ -360,6 +430,9 @@ export class SearchEngine {
           item.dimensions_json = row.dimensions_json || null;
           item.description = item.description || row.description || '';
           item.trigger = item.trigger || row.trigger || '';
+          // V3: headers & module_name 是独立列
+          if (row.headers) item.headers = row.headers;
+          if (row.module_name) item.module_name = row.module_name;
         }
       }
     } catch { /* DB may not be available */ }

package/lib/service/skills/SignalCollector.js

@@ -128,8 +128,9 @@ export class SignalCollector {
       `[SignalCollector] started — mode=${this.#mode}, initialInterval=${this.#intervalMs}ms, AI-driven`
     );
 
-    // 首次延迟 15 秒后执行（等启动流程稳定）
-    this.#timer = setTimeout(() => this.#tick(), 15_000);
+    // 首次按正常间隔执行（不立即触发，避免启动时消耗 AI token）
+    // 如果有事件推送（EventAggregator batch），会提前触发
+    this.#timer = setTimeout(() => this.#tick(), this.#intervalMs);
   }
 
   stop() {
@@ -335,11 +336,16 @@ export class SignalCollector {
   #collectRecipeSignals() {
     try {
       if (!this.#db) return [];
+      // V3: knowledge_entries 统一表，统计字段在 stats/quality JSON 中
       const rows = this.#db.prepare(
         `SELECT id, title, knowledge_type, category, language,
-                adoption_count, application_count, success_count,
-                quality_overall, updated_at
-         FROM recipes ORDER BY updated_at DESC LIMIT 30`
+                json_extract(stats, '$.adoptions') as adoption_count,
+                json_extract(stats, '$.applications') as application_count,
+                json_extract(quality, '$.overall') as quality_overall,
+                updated_at
+         FROM knowledge_entries
+         WHERE lifecycle = 'active'
+         ORDER BY updated_at DESC LIMIT 30`
       ).all();
       return rows;
     } catch { return []; }
@@ -348,11 +354,11 @@ export class SignalCollector {
   #collectCandidateSignals() {
     try {
       if (!this.#db) return [];
+      // V3: candidates 已合并到 knowledge_entries，lifecycle='pending' 即为候选
       const rows = this.#db.prepare(
-        `SELECT id, source, status, language, category,
-                json_extract(metadata_json, '$.title') as title,
-                created_at
-         FROM candidates WHERE status = 'pending'
+        `SELECT id, source, lifecycle as status, language, category,
+                title, created_at
+         FROM knowledge_entries WHERE lifecycle = 'pending'
          ORDER BY created_at DESC LIMIT 30`
       ).all();
       return rows;

package/lib/service/skills/SkillAdvisor.js

@@ -202,8 +202,8 @@ export class SkillAdvisor {
       // 按 category 分布
       const categories = this.#db.prepare(`
         SELECT category, COUNT(*) as cnt
-        FROM recipes
-        WHERE category IS NOT NULL AND category != ''
+        FROM knowledge_entries
+        WHERE lifecycle = 'active' AND category IS NOT NULL AND category != ''
         GROUP BY category
         ORDER BY cnt DESC
       `).all();
@@ -211,8 +211,8 @@ export class SkillAdvisor {
       // 按 language 分布
       const languages = this.#db.prepare(`
         SELECT language, COUNT(*) as cnt
-        FROM recipes
-        WHERE language IS NOT NULL AND language != ''
+        FROM knowledge_entries
+        WHERE lifecycle = 'active' AND language IS NOT NULL AND language != ''
         GROUP BY language
         ORDER BY cnt DESC
       `).all();
@@ -231,14 +231,15 @@ export class SkillAdvisor {
         });
       }
 
-      // 高使用量 Recipe 统计（adoption_count + application_count >= 5）
+      // 高使用量 Recipe 统计（V3: stats JSON 中的 adoptions + applications）
       let hotRecipes = [];
       try {
         hotRecipes = this.#db.prepare(`
           SELECT title, category,
-                 (adoption_count + application_count) as total_usage
-          FROM recipes
-          WHERE (adoption_count + application_count) >= 5
+                 (COALESCE(json_extract(stats, '$.adoptions'), 0) + COALESCE(json_extract(stats, '$.applications'), 0)) as total_usage
+          FROM knowledge_entries
+          WHERE lifecycle = 'active'
+            AND (COALESCE(json_extract(stats, '$.adoptions'), 0) + COALESCE(json_extract(stats, '$.applications'), 0)) >= 5
           ORDER BY total_usage DESC
           LIMIT 10
         `).all();
@@ -262,12 +263,13 @@ export class SkillAdvisor {
     if (!this.#db) return { summary: 'DB 不可用', suggestions };
 
     try {
+      // V3: candidates 已合并到 knowledge_entries，ifecycle='pending' 即为候选
       const stats = this.#db.prepare(`
         SELECT
           COUNT(*) as total,
-          SUM(CASE WHEN status='pending' THEN 1 ELSE 0 END) as pending,
-          SUM(CASE WHEN status='rejected' THEN 1 ELSE 0 END) as rejected
-        FROM candidates
+          SUM(CASE WHEN lifecycle='pending' THEN 1 ELSE 0 END) as pending,
+          SUM(CASE WHEN lifecycle='deprecated' THEN 1 ELSE 0 END) as rejected
+        FROM knowledge_entries
       `).get();
 
       // 大量被拒绝 → 提示候选质量 Skill

package/lib/service/snippet/SnippetFactory.js

@@ -45,16 +45,16 @@ const SNIPPET_TEMPLATE = `<?xml version="1.0" encoding="UTF-8"?>
 
 export class SnippetFactory {
   /**
-   * @param {object} [recipeRepository] — RecipeRepositoryImpl (可选，用于列表查询)
+   * @param {object} [knowledgeRepository] — KnowledgeRepositoryImpl（可选，用于列表查询）
    */
-  constructor(recipeRepository) {
-    this._recipeRepo = recipeRepository || null;
+  constructor(knowledgeRepository) {
+    this._recipeRepo = knowledgeRepository || null;
   }
 
   /**
-   * 运行时注入 recipeRepository（用于延迟绑定场景）
+   * 运行时注入 knowledgeRepository（用于延迟绑定场景）
    */
-  setRecipeRepository(repo) {
+  setKnowledgeRepository(repo) {
     this._recipeRepo = repo;
   }
 

package/lib/service/spm/SpmService.js

@@ -7,8 +7,9 @@ import Logger from '../../infrastructure/logging/Logger.js';
 import { PackageSwiftParser } from './PackageSwiftParser.js';
 import { DependencyGraph } from './DependencyGraph.js';
 import { PolicyEngine } from './PolicyEngine.js';
+import { GraphCache } from '../../infrastructure/cache/GraphCache.js';
 import { dirname, relative, sep, resolve as pathResolve } from 'node:path';
-import { readFileSync, writeFileSync } from 'node:fs';
+import { readFileSync, writeFileSync, existsSync } from 'node:fs';
 
 export class SpmService {
   #parser;
@@ -35,6 +36,12 @@ export class SpmService {
    */
   #packageDepGraph;
 
+  /** @type {GraphCache} 磁盘缓存层 */
+  #graphCache;
+
+  /** @type {Map<string, object[]>} target 文件列表缓存 */
+  #targetFilesCache;
+
   constructor(projectRoot, options = {}) {
     this.#projectRoot = projectRoot;
     this.#aiFactory = options.aiFactory || null;
@@ -49,60 +56,92 @@ export class SpmService {
     this.#logger = Logger.getInstance();
     this.#targetPackageMap = new Map();
     this.#packageDepGraph = new Map();
+    this.#graphCache = new GraphCache(projectRoot);
+    this.#targetFilesCache = new Map();
   }
 
   /**
    * 加载并解析 Package.swift，构建依赖图
    * 支持多 Package 项目（如 BiliDemo 有多个子 Package）
+   * 优先从磁盘缓存加载（Package.swift contentHash 匹配即命中）
    */
   async load() {
     this.#targetPackageMap.clear();
     this.#packageDepGraph.clear();
+    this.#targetFilesCache.clear();
 
-    // 优先尝试根目录的 Package.swift
+    // ── 收集所有 Package.swift 路径 + 联合 hash ──
     const packagePath = this.#parser.findPackageSwift(this.#projectRoot);
+    const allPaths = packagePath ? [packagePath] : this.#parser.findAllPackageSwifts(this.#projectRoot);
+    if (allPaths.length === 0) {
+      this.#logger.warn('[SpmService] Package.swift 未找到');
+      return null;
+    }
+
+    const combinedHash = allPaths
+      .map(p => this.#graphCache.computeFileHash(p))
+      .join(':');
+
+    // ── 尝试命中缓存 ──
+    const cached = this.#graphCache.load('spm-graph');
+    if (cached && cached.contentHash === combinedHash) {
+      this.#restoreFromCache(cached.data);
+      this.#logger.info(`[SpmService] ⚡ 缓存命中 (${this.#graph.getNodes().length} targets, hash=${combinedHash.substring(0, 8)})`);
+      return cached.data.parsedResult;
+    }
+
+    // ── 缓存未命中，走完整解析 ──
+    const startTime = Date.now();
+    let parsedResult;
+
     if (packagePath) {
+      // 单包模式
       const parsed = this.#parser.parse(packagePath);
       this.#graph.buildFromParsed(parsed);
-      // 构建 target→package 映射
       for (const t of parsed.targets || []) {
         this.#targetPackageMap.set(t.name, { packageName: parsed.name, packagePath });
       }
-      // 构建包级依赖图（解析 .package(path: "...") 引用）
       this.#buildPackageDepGraph([{ path: packagePath, parsed }]);
       this.#logger.info(`[SpmService] 加载完成: ${parsed.name} (${parsed.targets.length} targets)`);
-      return parsed;
+      parsedResult = parsed;
+    } else {
+      // 多包模式
+      parsedResult = this.#loadMultiPackage(allPaths);
     }
 
-    // 没有根 Package.swift，扫描子目录中的所有 Package.swift
-    const allPaths = this.#parser.findAllPackageSwifts(this.#projectRoot);
-    if (allPaths.length === 0) {
-      this.#logger.warn('[SpmService] Package.swift 未找到');
-      return null;
+    // ── 写入缓存 ──
+    if (parsedResult) {
+      this.#saveToCache(combinedHash, parsedResult);
+      this.#logger.info(`[SpmService] 缓存已写入 (${Date.now() - startTime}ms 解析)`);
     }
 
+    return parsedResult;
+  }
+
+  /**
+   * 多包加载（从 load() 拆出）
+   * @param {string[]} allPaths Package.swift 路径数组
+   * @returns {object|null}
+   */
+  #loadMultiPackage(allPaths) {
     this.#logger.info(`[SpmService] 发现 ${allPaths.length} 个 Package.swift，逐一解析...`);
     let mergedTargets = [];
     let lastName = 'multi-package';
     const allParsed = [];
 
-    // 先清空图，然后逐个添加节点和边（不用 buildFromParsed 避免重复 clear）
     this.#graph.clear();
     for (const pkgPath of allPaths) {
       try {
         const parsed = this.#parser.parse(pkgPath);
         if (parsed) {
           allParsed.push({ path: pkgPath, parsed });
-          // 逐个添加 target 到图中
           for (const t of parsed.targets || []) {
             this.#graph.addNode(t.name);
             for (const dep of t.dependencies || []) {
               this.#graph.addEdge(t.name, dep);
             }
-            // 构建 target→package 映射
             this.#targetPackageMap.set(t.name, { packageName: parsed.name, packagePath: pkgPath });
           }
-          // 合并 target 列表（带 packageName 标记）
           for (const t of parsed.targets) {
             mergedTargets.push({
               ...t,
@@ -117,7 +156,6 @@ export class SpmService {
       }
     }
 
-    // 构建包级依赖图
     this.#buildPackageDepGraph(allParsed);
 
     this.#logger.info(`[SpmService] 多包加载完成: ${mergedTargets.length} targets from ${allPaths.length} packages`);
@@ -128,6 +166,51 @@ export class SpmService {
     };
   }
 
+  /**
+   * 将当前内存状态序列化到缓存
+   */
+  #saveToCache(contentHash, parsedResult) {
+    const graphJSON = this.#graph.toJSON();
+    const targetPackageEntries = [...this.#targetPackageMap.entries()];
+    const packageDepEntries = [...this.#packageDepGraph.entries()].map(
+      ([k, v]) => [k, [...v]]
+    );
+
+    this.#graphCache.save('spm-graph', {
+      parsedResult,
+      graphNodes: graphJSON.nodes,
+      graphEdges: graphJSON.edges,
+      targetPackageMap: targetPackageEntries,
+      packageDepGraph: packageDepEntries,
+    }, { contentHash });
+  }
+
+  /**
+   * 从缓存数据恢复内存状态
+   */
+  #restoreFromCache(data) {
+    // 恢复 DependencyGraph
+    this.#graph.clear();
+    for (const node of data.graphNodes || []) {
+      this.#graph.addNode(node);
+    }
+    for (const edge of data.graphEdges || []) {
+      this.#graph.addEdge(edge.from, edge.to);
+    }
+
+    // 恢复 targetPackageMap
+    this.#targetPackageMap.clear();
+    for (const [name, info] of data.targetPackageMap || []) {
+      this.#targetPackageMap.set(name, info);
+    }
+
+    // 恢复 packageDepGraph
+    this.#packageDepGraph.clear();
+    for (const [pkgPath, deps] of data.packageDepGraph || []) {
+      this.#packageDepGraph.set(pkgPath, new Set(deps));
+    }
+  }
+
   // ─────────────── 包级依赖图构建 ───────────────
 
   /**
@@ -585,14 +668,29 @@ export class SpmService {
    * 获取 Target 源文件列表（路由: POST /spm/target-files）
    * 支持多 Package 项目：先在 target 所属 package 目录查找
    * 支持非 SPM 项目：虚拟 directory target 直接扫描对应目录
+   * 使用内存缓存避免重复 fs walk
    */
   async getTargetFiles(target) {
-    const { existsSync, readdirSync, statSync } = await import('fs');
-    const { join, dirname, relative, extname } = await import('path');
-
     const targetName = typeof target === 'string' ? target : target?.name;
     if (!targetName) return [];
 
+    // 内存缓存命中
+    if (this.#targetFilesCache.has(targetName)) {
+      return this.#targetFilesCache.get(targetName);
+    }
+
+    const result = await this.#walkTargetFiles(target, targetName);
+    this.#targetFilesCache.set(targetName, result);
+    return result;
+  }
+
+  /**
+   * 实际的文件遍历逻辑（拆出以供缓存用）
+   */
+  async #walkTargetFiles(target, targetName) {
+    const { existsSync, readdirSync, statSync } = await import('fs');
+    const { join, dirname, relative, extname } = await import('path');
+
     // 判断是否为虚拟目录 target（非 SPM fallback）
     const isDirectoryTarget = typeof target === 'object' && target?.type === 'directory';
 
@@ -1103,8 +1201,11 @@ export class SpmService {
 
   /**
    * 刷新依赖映射（路由: POST /commands/spm-map）
+   * 会主动使缓存失效并重新解析
    */
   async updateDependencyMap(options = {}) {
+    this.#graphCache.invalidate('spm-graph');
+    this.#targetFilesCache.clear();
     this.#graph.clear();
     const parsed = await this.load();
     if (!parsed) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "autosnippet",
-  "version": "2.8.3",
+  "version": "2.10.0",
   "description": "AutoSnippet - 连接开发者、AI 与项目知识库的工具",
   "type": "module",
   "main": "lib/bootstrap.js",

package/scripts/install-cursor-skill.js

@@ -22,7 +22,6 @@ const require = createRequire(import.meta.url);
 import fs from 'node:fs';
 import path from 'node:path';
 import * as defaults from '../lib/infrastructure/config/Defaults.js';
-import { RecipeStatsTracker } from '../lib/service/recipe/RecipeStatsTracker.js';
 
 const autoSnippetRoot = path.join(__dirname, '..');
 const skillsSource = path.join(autoSnippetRoot, 'skills');
@@ -156,11 +155,6 @@ function buildProjectRecipesContext(projectRoot) {
   if (!fs.existsSync(recipesDir)) return null;
   const mdFiles = collectMdFiles(recipesDir, recipesDir).sort();
   if (mdFiles.length === 0) return null;
-  let stats = { byFile: {} };
-  try {
-    const tracker = new RecipeStatsTracker(projectRoot);
-    stats = tracker.getStats();
-  } catch (_) {}
 
   const lines = [
     '# Project Recipes Index\n\n',