autosnippet 3.0.0 → 3.0.2

package/lib/service/recipe/RecipeParser.js

@@ -5,10 +5,11 @@
 
 import fs from 'node:fs';
 import path from 'node:path';
+import { LanguageService } from '../../shared/LanguageService.js';
 
 const FRONTMATTER_RE = /^---\n([\s\S]*?)\n---/;
-const SNIPPET_HEADING_RE = /^##\s+(?:Snippet|Code|代码)/mi;
-const USAGE_HEADING_RE = /^##\s+(?:Usage\s*Guide|用法|使用指南)/mi;
+const _SNIPPET_HEADING_RE = /^##\s+(?:Snippet|Code|代码)/im;
+const USAGE_HEADING_RE = /^##\s+(?:Usage\s*Guide|用法|使用指南)/im;
 const FENCED_CODE_RE = /```(\w*)\n([\s\S]*?)```/;
 
 export class RecipeParser {
@@ -17,7 +18,9 @@ export class RecipeParser {
    * 需包含: frontmatter + 代码块 + Usage Guide
    */
   isCompleteRecipe(text) {
-    if (!text) return false;
+    if (!text) {
+      return false;
+    }
     return FRONTMATTER_RE.test(text) && FENCED_CODE_RE.test(text) && USAGE_HEADING_RE.test(text);
   }
 
@@ -25,7 +28,9 @@ export class RecipeParser {
    * 检查是否为「仅介绍」Recipe（有 frontmatter 但无代码块）
    */
   isIntroOnly(text) {
-    if (!text) return false;
+    if (!text) {
+      return false;
+    }
     return FRONTMATTER_RE.test(text) && !FENCED_CODE_RE.test(text);
   }
 
@@ -35,7 +40,9 @@ export class RecipeParser {
    * @returns {object|null}
    */
   parse(text) {
-    if (!text) return null;
+    if (!text) {
+      return null;
+    }
 
     const frontmatter = this.parseFrontmatter(text);
     const body = text.replace(FRONTMATTER_RE, '').trim();
@@ -54,9 +61,10 @@ export class RecipeParser {
     if (usageMatch) {
       const usageStart = usageMatch.index + usageMatch[0].length;
       const nextHeading = body.slice(usageStart).search(/^##\s+/m);
-      usageGuide = nextHeading > 0
-        ? body.slice(usageStart, usageStart + nextHeading).trim()
-        : body.slice(usageStart).trim();
+      usageGuide =
+        nextHeading > 0
+          ? body.slice(usageStart, usageStart + nextHeading).trim()
+          : body.slice(usageStart).trim();
     }
 
     // 提取标题
@@ -72,8 +80,10 @@ export class RecipeParser {
       description: frontmatter.description || frontmatter.summary || '',
       trigger: frontmatter.trigger || this.#generateTrigger(title),
       category: frontmatter.category || 'general',
-      language: frontmatter.language || (codeBlocks[0]?.language !== 'text' ? codeBlocks[0]?.language : 'swift'),
-      code: codeBlocks.map(b => b.code).join('\n\n'),
+      language:
+        frontmatter.language ||
+        (codeBlocks[0]?.language !== 'text' ? codeBlocks[0]?.language : 'swift'),
+      code: codeBlocks.map((b) => b.code).join('\n\n'),
       codeBlocks,
       usageGuide,
       headers,
@@ -86,9 +96,11 @@ export class RecipeParser {
    * 从文本中解析多段 Recipe（按 `---` 分隔）
    */
   parseAll(text) {
-    if (!text) return [];
-    const segments = text.split(/\n---\n/).filter(s => s.trim().length > 0);
-    return segments.map(s => this.parse(s)).filter(Boolean);
+    if (!text) {
+      return [];
+    }
+    const segments = text.split(/\n---\n/).filter((s) => s.trim().length > 0);
+    return segments.map((s) => this.parse(s)).filter(Boolean);
   }
 
   /**
@@ -96,7 +108,9 @@ export class RecipeParser {
    */
   parseFrontmatter(text) {
     const match = text.match(FRONTMATTER_RE);
-    if (!match) return {};
+    if (!match) {
+      return {};
+    }
 
     const fm = {};
     for (const line of match[1].split('\n')) {
@@ -105,11 +119,19 @@ export class RecipeParser {
         const key = line.slice(0, colonIdx).trim();
         let value = line.slice(colonIdx + 1).trim();
         if (value.startsWith('[') && value.endsWith(']')) {
-          value = value.slice(1, -1).split(',').map(s => s.trim().replace(/^['"]|['"]$/g, ''));
-        } else if (value === 'true') value = true;
-        else if (value === 'false') value = false;
-        else if (/^\d+$/.test(value)) value = parseInt(value, 10);
-        else value = value.replace(/^['"]|['"]$/g, '');
+          value = value
+            .slice(1, -1)
+            .split(',')
+            .map((s) => s.trim().replace(/^['"]|['"]$/g, ''));
+        } else if (value === 'true') {
+          value = true;
+        } else if (value === 'false') {
+          value = false;
+        } else if (/^\d+$/.test(value)) {
+          value = parseInt(value, 10);
+        } else {
+          value = value.replace(/^['"]|['"]$/g, '');
+        }
         fm[key] = value;
       }
     }
@@ -143,8 +165,7 @@ export class RecipeParser {
 
     const content = fs.readFileSync(fullPath, 'utf8');
     const ext = path.extname(fullPath).toLowerCase();
-    const langMap = { '.swift': 'swift', '.m': 'objc', '.h': 'objc', '.mm': 'objc', '.ts': 'typescript', '.js': 'javascript', '.py': 'python' };
-    const language = langMap[ext] || 'swift';
+    const language = LanguageService.langFromExt(ext);
 
     // 尝试解析为完整 Recipe Markdown
     if (this.isCompleteRecipe(content)) {
@@ -163,20 +184,22 @@ export class RecipeParser {
     // 回退: 将整个文件内容作为代码片段
     const title = path.basename(fullPath, ext);
     return {
-      items: [{
-        title,
-        summary: '',
-        description: '',
-        trigger: this.#generateTrigger(title),
-        category: 'Utility',
-        language,
-        code: content,
-        codeBlocks: [{ language, code: content }],
-        usageGuide: '',
-        headers: this.#extractHeaders(content),
-        includeHeaders: false,
-        frontmatter: {},
-      }],
+      items: [
+        {
+          title,
+          summary: '',
+          description: '',
+          trigger: this.#generateTrigger(title),
+          category: 'Utility',
+          language,
+          code: content,
+          codeBlocks: [{ language, code: content }],
+          usageGuide: '',
+          headers: this.#extractHeaders(content),
+          includeHeaders: false,
+          frontmatter: {},
+        },
+      ],
       isMarked: false,
     };
   }
@@ -197,12 +220,16 @@ export class RecipeParser {
     // 尝试完整 Recipe 解析
     if (this.isCompleteRecipe(text)) {
       const parsed = this.parse(text);
-      if (parsed) return parsed;
+      if (parsed) {
+        return parsed;
+      }
     }
 
     // 尝试批量解析
     const all = this.parseAll(text);
-    if (all.length > 0) return all;
+    if (all.length > 0) {
+      return all;
+    }
 
     throw new Error('文本不是有效的 Recipe Markdown 格式');
   }
@@ -219,13 +246,15 @@ export class RecipeParser {
       throw new Error('文本内容为空');
     }
 
-    const language = opts.language || 'swift';
+    const language = opts.language || 'unknown';
 
     // 先尝试标准解析
     try {
       const result = await this.parseFromText(text, opts);
       return result;
-    } catch { /* 继续兜底逻辑 */ }
+    } catch {
+      /* 继续兜底逻辑 */
+    }
 
     // 提取代码块
     const codeBlocks = [];
@@ -235,12 +264,10 @@ export class RecipeParser {
       codeBlocks.push({ language: match[1] || language, code: match[2].trim() });
     }
 
-    const code = codeBlocks.length > 0
-      ? codeBlocks.map(b => b.code).join('\n\n')
-      : text.trim();
+    const code = codeBlocks.length > 0 ? codeBlocks.map((b) => b.code).join('\n\n') : text.trim();
 
     // 简单标题推断
-    const titleLine = text.split('\n').find(l => l.trim().startsWith('#'));
+    const titleLine = text.split('\n').find((l) => l.trim().startsWith('#'));
     const title = titleLine ? titleLine.replace(/^#+\s*/, '').trim() : 'Untitled Snippet';
 
     return {
@@ -270,7 +297,13 @@ export class RecipeParser {
   }
 
   #generateTrigger(title) {
-    if (!title) return '';
-    return title.toLowerCase().replace(/[^a-z0-9]+/g, '_').replace(/(^_|_$)/g, '').slice(0, 30);
+    if (!title) {
+      return '';
+    }
+    return title
+      .toLowerCase()
+      .replace(/[^a-z0-9]+/g, '_')
+      .replace(/(^_|_$)/g, '')
+      .slice(0, 30);
   }
 }

package/lib/service/search/CoarseRanker.js

@@ -8,11 +8,11 @@ export class CoarseRanker {
 
   constructor(options = {}) {
     this.#weights = {
-      bm25:       options.bm25Weight       ?? 0.30,
-      semantic:   options.semanticWeight   ?? 0.30,
-      quality:    options.qualityWeight    ?? 0.20,
-      freshness:  options.freshnessWeight  ?? 0.10,
-      popularity: options.popularityWeight ?? 0.10,
+      bm25: options.bm25Weight ?? 0.3,
+      semantic: options.semanticWeight ?? 0.3,
+      quality: options.qualityWeight ?? 0.2,
+      freshness: options.freshnessWeight ?? 0.1,
+      popularity: options.popularityWeight ?? 0.1,
     };
   }
 
@@ -22,24 +22,32 @@ export class CoarseRanker {
    * @returns {Array} — sorted with coarseScore
    */
   rank(candidates) {
-    if (!candidates || candidates.length === 0) return [];
+    if (!candidates || candidates.length === 0) {
+      return [];
+    }
 
-    return candidates.map(c => {
-      const bm25      = this.#normalize(c.bm25Score || c.score || 0);
-      const semantic   = this.#normalize(c.semanticScore || 0);
-      const quality    = this.#computeQuality(c);
-      const freshness  = this.#computeFreshness(c);
-      const popularity = this.#computePopularity(c);
+    return candidates
+      .map((c) => {
+        const bm25 = this.#normalize(c.bm25Score || c.score || 0);
+        const semantic = this.#normalize(c.semanticScore || 0);
+        const quality = this.#computeQuality(c);
+        const freshness = this.#computeFreshness(c);
+        const popularity = this.#computePopularity(c);
 
-      const coarseScore =
-        bm25       * this.#weights.bm25 +
-        semantic   * this.#weights.semantic +
-        quality    * this.#weights.quality +
-        freshness  * this.#weights.freshness +
-        popularity * this.#weights.popularity;
+        const coarseScore =
+          bm25 * this.#weights.bm25 +
+          semantic * this.#weights.semantic +
+          quality * this.#weights.quality +
+          freshness * this.#weights.freshness +
+          popularity * this.#weights.popularity;
 
-      return { ...c, coarseScore, coarseSignals: { bm25, semantic, quality, freshness, popularity } };
-    }).sort((a, b) => b.coarseScore - a.coarseScore);
+        return {
+          ...c,
+          coarseScore,
+          coarseSignals: { bm25, semantic, quality, freshness, popularity },
+        };
+      })
+      .sort((a, b) => b.coarseScore - a.coarseScore);
   }
 
   /**
@@ -55,13 +63,13 @@ export class CoarseRanker {
     const hasTitle = !!candidate.title;
     const hasCode = !!(candidate.code || candidate.content);
     const hasDesc = !!(candidate.description || candidate.summary);
-    score += (hasTitle ? 0.15 : 0) + (hasCode ? 0.15 : 0) + (hasDesc ? 0.10 : 0);
+    score += (hasTitle ? 0.15 : 0) + (hasCode ? 0.15 : 0) + (hasDesc ? 0.1 : 0);
 
     // 结构质量 (30%)
     const hasCat = !!candidate.category;
     const hasLang = !!candidate.language;
     const hasTags = Array.isArray(candidate.tags) && candidate.tags.length > 0;
-    score += (hasCat ? 0.10 : 0) + (hasLang ? 0.10 : 0) + (hasTags ? 0.10 : 0);
+    score += (hasCat ? 0.1 : 0) + (hasLang ? 0.1 : 0) + (hasTags ? 0.1 : 0);
 
     // 代码可读性 (30%)
     const code = candidate.code || candidate.content || '';
@@ -75,14 +83,21 @@ export class CoarseRanker {
 
   #computeFreshness(candidate) {
     const updated = candidate.updatedAt || candidate.lastModified || candidate.createdAt;
-    if (!updated) return 0.5;
+    if (!updated) {
+      return 0.5;
+    }
     // 自动识别秒级/毫秒级 Unix 时间戳 (秒级 ≤ 9999999999 即 2286 年)
-    const ts = typeof updated === 'number' && updated > 0 && updated <= 9999999999
-      ? updated * 1000
-      : (typeof updated === 'number' ? updated : new Date(updated).getTime());
+    const ts =
+      typeof updated === 'number' && updated > 0 && updated <= 9999999999
+        ? updated * 1000
+        : typeof updated === 'number'
+          ? updated
+          : new Date(updated).getTime();
     const ageDays = (Date.now() - ts) / 86400000;
-    if (ageDays < 0) return 1.0; // 未来时间戳视为最新
-    return Math.exp(-0.693 * ageDays / 180); // 半衰期 180 天
+    if (ageDays < 0) {
+      return 1.0; // 未来时间戳视为最新
+    }
+    return Math.exp((-Math.LN2 * ageDays) / 180); // 半衰期 180 天
   }
 
   #computePopularity(candidate) {

package/lib/service/search/CrossEncoderReranker.js

@@ -15,11 +15,11 @@
  *   - AI 不可用时自动降级到 Jaccard
  */
 
-import { tokenize } from './InvertedIndex.js';
 import { jaccardSimilarity } from '../../shared/similarity.js';
+import { tokenize } from './InvertedIndex.js';
 
-const MAX_CANDIDATES = 40;  // 超过此数量截断（控制 prompt 大小）
-const MAX_DOC_LEN = 300;    // 每个文档最大字符数
+const MAX_CANDIDATES = 40; // 超过此数量截断（控制 prompt 大小）
+const MAX_DOC_LEN = 300; // 每个文档最大字符数
 
 export class CrossEncoderReranker {
   #aiProvider;
@@ -43,8 +43,12 @@ export class CrossEncoderReranker {
    * @returns {Promise<Array<object>>} — 附带 semanticScore 的候选列表（降序）
    */
   async rerank(query, candidates) {
-    if (!candidates || candidates.length === 0) return [];
-    if (!query) return candidates;
+    if (!candidates || candidates.length === 0) {
+      return [];
+    }
+    if (!query) {
+      return candidates;
+    }
 
     // 如果 AI Provider 不可用，降级到 Jaccard
     if (!this.#aiProvider || typeof this.#aiProvider.chatWithStructuredOutput !== 'function') {
@@ -58,16 +62,17 @@ export class CrossEncoderReranker {
     try {
       const scored = await this.#batchScore(query, head);
       // tail 部分给一个递减的低分以保持稳定排序
-      const minScore = scored.length > 0
-        ? Math.min(...scored.map(s => s.semanticScore)) * 0.5
-        : 0;
+      const minScore =
+        scored.length > 0 ? Math.min(...scored.map((s) => s.semanticScore)) * 0.5 : 0;
       const tailScored = tail.map((c, i) => ({
         ...c,
         semanticScore: Math.max(minScore - (i + 1) * 0.001, 0),
       }));
       return [...scored, ...tailScored];
     } catch (err) {
-      this.#logger.warn?.(`[CrossEncoderReranker] AI scoring failed, falling back to Jaccard: ${err.message}`);
+      this.#logger.warn?.(
+        `[CrossEncoderReranker] AI scoring failed, falling back to Jaccard: ${err.message}`
+      );
       return this.#jaccardFallback(query, candidates);
     }
   }
@@ -126,10 +131,12 @@ Return ONLY a JSON array, no markdown or explanation.`;
     }
 
     // 合并分数，未评分的给 0
-    return candidates.map((c, i) => ({
-      ...c,
-      semanticScore: scoreMap.get(i) ?? 0,
-    })).sort((a, b) => b.semanticScore - a.semanticScore);
+    return candidates
+      .map((c, i) => ({
+        ...c,
+        semanticScore: scoreMap.get(i) ?? 0,
+      }))
+      .sort((a, b) => b.semanticScore - a.semanticScore);
   }
 
   /**
@@ -151,13 +158,17 @@ Return ONLY a JSON array, no markdown or explanation.`;
    */
   #jaccardFallback(query, candidates) {
     const queryTokens = new Set(tokenize(query));
-    if (queryTokens.size === 0) return candidates;
-
-    return candidates.map(candidate => {
-      const text = this.#extractDocText(candidate);
-      const docTokens = new Set(tokenize(text));
-      const score = jaccardSimilarity(queryTokens, docTokens);
-      return { ...candidate, semanticScore: score };
-    }).sort((a, b) => b.semanticScore - a.semanticScore);
+    if (queryTokens.size === 0) {
+      return candidates;
+    }
+
+    return candidates
+      .map((candidate) => {
+        const text = this.#extractDocText(candidate);
+        const docTokens = new Set(tokenize(text));
+        const score = jaccardSimilarity(queryTokens, docTokens);
+        return { ...candidate, semanticScore: score };
+      })
+      .sort((a, b) => b.semanticScore - a.semanticScore);
   }
 }

package/lib/service/search/InvertedIndex.js

@@ -18,11 +18,15 @@ export function buildInvertedIndex(documents) {
 
   for (let i = 0; i < documents.length; i++) {
     const doc = documents[i];
-    const text = [doc.title, doc.trigger, doc.content, doc.code, doc.description].filter(Boolean).join(' ');
+    const text = [doc.title, doc.trigger, doc.content, doc.code, doc.description]
+      .filter(Boolean)
+      .join(' ');
     const tokens = tokenize(text);
 
     for (const token of tokens) {
-      if (!index.has(token)) index.set(token, new Set());
+      if (!index.has(token)) {
+        index.set(token, new Set());
+      }
       index.get(token).add(i);
     }
   }
@@ -38,13 +42,17 @@ export function buildInvertedIndex(documents) {
  */
 export function lookup(invertedIndex, query) {
   const queryTokens = tokenize(query);
-  if (queryTokens.length === 0) return [];
+  if (queryTokens.length === 0) {
+    return [];
+  }
 
   const resultSet = new Set();
   for (const token of queryTokens) {
     const docs = invertedIndex.get(token);
     if (docs) {
-      for (const docIdx of docs) resultSet.add(docIdx);
+      for (const docIdx of docs) {
+        resultSet.add(docIdx);
+      }
     }
   }
 
@@ -59,17 +67,23 @@ export function lookup(invertedIndex, query) {
  */
 export function lookupAll(invertedIndex, query) {
   const queryTokens = tokenize(query);
-  if (queryTokens.length === 0) return [];
+  if (queryTokens.length === 0) {
+    return [];
+  }
 
   let result = null;
   for (const token of queryTokens) {
     const docs = invertedIndex.get(token);
-    if (!docs) return [];
+    if (!docs) {
+      return [];
+    }
     if (result === null) {
       result = new Set(docs);
     } else {
       for (const idx of result) {
-        if (!docs.has(idx)) result.delete(idx);
+        if (!docs.has(idx)) {
+          result.delete(idx);
+        }
       }
     }
   }