@cachly-dev/mcp-server 0.5.70 → 0.5.80

package/dist/index.js

@@ -47,7 +47,7 @@ import { Redis } from 'ioredis';
 const API_URL = process.env.CACHLY_API_URL ?? 'https://api.cachly.dev';
 const JWT = process.env.CACHLY_JWT ?? '';
 const EMBED_MODEL = process.env.CACHLY_EMBED_MODEL ?? '';
-const CURRENT_VERSION = '0.5.70';
+const CURRENT_VERSION = '0.5.80';
 /**
  * ┌──────────────────────────────────────────────────────────────────────┐
  * │  EMBEDDING PROVIDER — pluggable, client-side first                  │
@@ -454,6 +454,13 @@ const STOPWORDS = new Set([
     'এবং', 'কিন্তু', 'বা', 'না', 'হ্যাঁ', 'যে', 'কি', 'কে', 'কী', 'থেকে',
     'দিয়ে', 'জন্য', 'সাথে', 'মধ্যে', 'উপর', 'নিচে', 'আছে', 'ছিল', 'হয়',
     'করা', 'করে', 'করেছে', 'হবে', 'হয়েছে', 'একটি', 'একটা', 'অনেক', 'সব',
+    // ── Vietnamese — common particles, pronouns, auxiliary words ─────────────
+    'tôi', 'bạn', 'anh', 'chị', 'em', 'họ', 'chúng', 'ta', 'mình',
+    'là', 'có', 'không', 'và', 'với', 'của', 'cho', 'trong', 'về',
+    'từ', 'đến', 'để', 'khi', 'nếu', 'thì', 'mà', 'nhưng', 'vì',
+    'đây', 'đó', 'này', 'kia', 'rất', 'cũng', 'đã', 'sẽ', 'đang',
+    'được', 'bị', 'những', 'các', 'một', 'hai', 'ba', 'nhiều', 'ít',
+    'nào', 'ai', 'gì', 'đâu', 'sao', 'bao', 'giờ', 'lúc', 'sau',
 ]);
 // CJK Unicode ranges used for bigram extraction
 const CJK_RE = /[\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]/;
@@ -473,6 +480,9 @@ const CYRILLIC_RE = /[\u0400-\u04ff]/;
 const TURKISH_CHAR_RE = /[\u011f\u015f\u0131\u0130\u00e7]/; // ğ ş ı İ ç
 // Bengali (Bangla) — U+0980–U+09FF
 const BENGALI_RE = /[\u0980-\u09ff]/;
+// Vietnamese uses Latin + Latin Extended Additional (U+1EA0–U+1EF9) for tone marks
+// e.g. ắ ặ ầ ổ ợ ụ ừ — detected by chars exclusive to Vietnamese diacritics
+const VIETNAMESE_CHAR_RE = /[\u1ea0-\u1ef9]/;
 // ── Katakana → Romaji conversion table (Hepburn system) ──────────────────────
 // Digraphs must be listed before single chars so they match first.
 const KATA_ROMAJI_TABLE = [
@@ -928,13 +938,13 @@ const CROSS_LINGUAL_MAP = new Map([
     ['pip', ['python', 'package', 'pypi']],
     ['brew', ['homebrew', 'macos', 'package']],
     // English → all
-    ['deploy', ['デプロイ', '部署', '배포', 'deployment', 'deploying', 'نشر', 'פריסה', 'استقرار', 'तैनाती', 'bereitstellen', 'bereitstellung', 'déployer', 'déploiement', 'desplegar', 'despliegue', 'distribuire', 'distribuzione', 'implantar', 'implantação', 'развёртывание', 'развертывание', 'dağıtım', 'dağıtmak']],
-    ['deployment', ['デプロイ', '部署', '배포', 'deploy', 'نشر', 'פריסה', 'استقرار', 'तैनाती', 'bereitstellung', 'déploiement', 'despliegue', 'distribuzione', 'implantação', 'развёртывание', 'dağıtım']],
+    ['deploy', ['デプロイ', '部署', '배포', 'deployment', 'deploying', 'نشر', 'פריסה', 'استقرار', 'तैनाती', 'bereitstellen', 'bereitstellung', 'déployer', 'déploiement', 'desplegar', 'despliegue', 'distribuire', 'distribuzione', 'implantar', 'implantação', 'развёртывание', 'развертывание', 'dağıtım', 'dağıtmak', 'triển khai']],
+    ['deployment', ['デプロイ', '部署', '배포', 'deploy', 'نشر', 'פריסה', 'استقرار', 'तैनाती', 'bereitstellung', 'déploiement', 'despliegue', 'distribuzione', 'implantação', 'развёртывание', 'dağıtım', 'triển khai']],
     ['container', ['コンテナ', '容器', '컨테이너', 'docker']],
-    ['server', ['サーバー', 'サーバ', '服务器', '서버', 'سيرفر', 'שרת', 'سرور', 'सर्वर', 'сервер', 'sunucu']],
+    ['server', ['サーバー', 'サーバ', '服务器', '서버', 'سيرفر', 'שרת', 'سرور', 'सर्वर', 'сервер', 'sunucu', 'máy chủ']],
     ['database', ['データベース', '数据库', '데이터베이스', 'db', 'قاعدة البيانات', 'datenbank', 'پایگاه داده', 'डेटाबेس', 'base de données', 'base de datos', 'banco de dados']],
-    ['cache', ['キャッシュ', '缓存', '캐시', 'كاش', 'מטמון', 'کش', 'कैश', 'кэш', 'önbellek']],
-    ['error', ['エラー', '错误', '오류', 'خطأ', 'שגיאה', 'خطا', 'त्रुटि', 'गलती', 'exception', 'err', 'fehler', 'erreur', 'fallo', 'errore', 'erro', 'ошибка', 'hata', 'hatalı']],
+    ['cache', ['キャッシュ', '缓存', '캐시', 'كاش', 'מטמון', 'کش', 'कैश', 'кэш', 'önbellek', 'bộ đệm']],
+    ['error', ['エラー', '错误', '오류', 'خطأ', 'שגיאה', 'خطا', 'त्रुटि', 'गलती', 'exception', 'err', 'fehler', 'erreur', 'fallo', 'errore', 'erro', 'ошибка', 'hata', 'hatalı', 'lỗi']],
     ['bug', ['バグ', '缺陷', '버그', 'issue', 'defect', 'خلل', 'اشکال', 'बग']],
     ['fix', ['修正', '修复', '수정', 'bugfix', 'patch', 'hotfix', 'إصلاح', 'תיקון', 'رفع', 'सुधार', 'beheben', 'behoben', 'réparer', 'correction', 'arreglar', 'corrección', 'correggere', 'corrigir', 'исправить', 'исправление', 'düzeltmek', 'düzeltme']],
     ['build', ['ビルド', '构建', '빌드', 'بناء', 'בנייה', 'ساخت', 'निर्माण', 'bauen', 'construire', 'construir', 'costruire', 'сборка', 'собрать', 'derleme', 'derlemek']],
@@ -1509,6 +1519,37 @@ const CROSS_LINGUAL_MAP = new Map([
     ['şifre', ['password']],
     ['izleme', ['monitor', 'monitoring']],
     ['göç', ['migrate', 'migration']],
+    // ── Vietnamese (Latin + Latin Extended Additional U+1EA0–U+1EF9) → English ─
+    ['triển khai', ['deploy', 'deployment']],
+    ['lỗi', ['error', 'err']],
+    ['sửa lỗi', ['fix', 'bugfix', 'debug']],
+    ['xây dựng', ['build']],
+    ['kiểm tra', ['test']],
+    ['xác thực', ['auth', 'authentication']],
+    ['cấu hình', ['config', 'configuration']],
+    ['cài đặt', ['install']],
+    ['cập nhật', ['update', 'upgrade']],
+    ['nhật ký', ['log']],
+    ['máy chủ', ['server']],
+    ['cơ sở dữ liệu', ['database', 'db']],
+    ['mạng', ['network']],
+    ['bộ nhớ', ['memory', 'ram']],
+    ['bộ đệm', ['cache']],
+    ['gỡ lỗi', ['debug', 'debugging']],
+    ['sự cố', ['crash', 'failure', 'error']],
+    ['khởi động lại', ['restart', 'reboot']],
+    ['quyền', ['permission', 'access']],
+    ['sao lưu', ['backup']],
+    ['hiệu suất', ['performance']],
+    ['vấn đề', ['problem', 'issue', 'error']],
+    ['dịch vụ', ['service']],
+    ['kết nối', ['connection', 'conn']],
+    ['vùng chứa', ['container', 'docker']],
+    ['nút', ['node']],
+    ['khóa', ['key', 'token', 'secret']],
+    ['mật khẩu', ['password']],
+    ['giám sát', ['monitor', 'monitoring']],
+    ['di chuyển', ['migrate', 'migration']],
 ]);
 /**
  * Expand a token to its cross-lingual synonyms.
@@ -1624,9 +1665,9 @@ function tokenize(text) {
         }
         else {
             // Latin / other — includes Turkish (ğ/ş/ı), Polish (ą/ę/ł/ń/ó/ś/ź/ż/ć),
-            // Czech (č/š/ž/ě/ů/ř), Hungarian (ő/ű), Romanian (ș/ț)
+            // Czech (č/š/ž/ě/ů/ř), Hungarian (ő/ű), Romanian (ș/ț), Vietnamese (ắặầổợụừ…)
             const words = seg
-                .replace(/[^a-záàâãäåæçéèêëíìîïñóòôõöúùûüýÿßœğşıİąćęłńśźżčšžěůřőűșț0-9:\-./]/g, ' ')
+                .replace(/[^a-záàâãäåæçéèêëíìîïñóòôõöúùûüýÿßœğşıİąćęłńśźżčšžěůřőűșț\u1ea0-\u1ef90-9:\-./]/g, ' ')
                 .split(/\s+/)
                 .filter(w => w.length > 1 && !STOPWORDS.has(w));
             // Turkish: apply light stemmer to words with Turkish-specific chars
@@ -1763,6 +1804,17 @@ const BM25_B = 0.75; // length normalization
 const BM25_DELTA = 1.0; // BM25+ lower-bound guarantee (0 = classic BM25)
 /** Recency boost: half-life in days. Entry from 7 days ago gets 0.5× boost. */
 const RECENCY_HALF_LIFE_DAYS = 7;
+const ZERO_RESULTS_LOG = [];
+const ZERO_RESULTS_MAX = 500;
+let zeroResultsTotal = 0;
+function logZeroResult(query) {
+    zeroResultsTotal++;
+    if (ZERO_RESULTS_LOG.length >= ZERO_RESULTS_MAX)
+        ZERO_RESULTS_LOG.shift();
+    ZERO_RESULTS_LOG.push({ query, ts: Date.now() });
+}
+/** Global index vocabulary — rebuilt on each keywordSearch call from the doc set. */
+let _indexVocab = new Set();
 /**
  * Levenshtein distance — edit distance between two strings.
  * Used for typo-tolerant fuzzy matching (distance ≤ 2 = match).
@@ -1903,13 +1955,34 @@ async function keywordSearch(redis, patterns, query, topK = 10) {
         return idf(term) * (tfNorm + BM25_DELTA);
     }
     /**
-     * Fuzzy match: tries exact → substring → Levenshtein ≤ 2.
+     * Fuzzy match: tries exact → prefix → substring → Levenshtein ≤ 2.
      * Returns [matchedTerm, weight] or null.
+     *
+     * Weights:
+     *  1.0 — exact match
+     *  0.85 — doc token starts with query (e.g. "dockerf" → "dockerfile")
+     *  0.75 — query starts with doc token (e.g. "kubernetes" → "kube")
+     *  0.6  — substring (either direction)
+     *  0.4  — Levenshtein ≤ 2 (typo tolerance)
      */
     function fuzzyMatch(qt, docTermSet) {
         // Exact
         if (docTermSet.has(qt))
             return [qt, 1.0];
+        // Prefix: query is prefix of a doc token (user typed partial word)
+        if (qt.length >= 4) {
+            for (const dt of docTermSet) {
+                if (dt.length > qt.length && dt.startsWith(qt))
+                    return [dt, 0.85];
+            }
+        }
+        // Reverse-prefix: doc token is prefix of query (doc has abbreviated form)
+        if (qt.length >= 4) {
+            for (const dt of docTermSet) {
+                if (dt.length >= 4 && dt.length < qt.length && qt.startsWith(dt))
+                    return [dt, 0.75];
+            }
+        }
         // Substring (partial)
         for (const dt of docTermSet) {
             if (dt.length > 3 && qt.length > 3 && (dt.includes(qt) || qt.includes(dt))) {
@@ -2003,7 +2076,17 @@ async function keywordSearch(redis, patterns, query, topK = 10) {
     }
     // ── Step 4: Sort by score, return top-K ──
     const sorted = [...allMatches.values()].sort((a, b) => b.score - a.score);
-    return sorted.slice(0, topK);
+    const topResults = sorted.slice(0, topK);
+    // ── Step 5: Zero-results logging + Did-You-Mean ───────────────────────────
+    if (topResults.length === 0) {
+        logZeroResult(query);
+        // Rebuild index vocab for Did-You-Mean suggestions
+        _indexVocab = new Set();
+        for (const doc of docs)
+            for (const t of doc.tokens)
+                _indexVocab.add(t);
+    }
+    return topResults;
 }
 // ── Exported for testing ──────────────────────────────────────────────────────
 export { tokenize, splitMultiQuery, levenshtein, recencyBoost, extractTimestamp, STOPWORDS, katakanaToRomaji, arabicLightStem, expandCrossLingual, CROSS_LINGUAL_MAP };
@@ -4308,10 +4391,37 @@ async function handleTool(name, args) {
             }
             if (kwMatches.length === 0) {
                 lines.push(`⚠️ No matches found for: "${query}"`);
-                lines.push(`\n💡 Tips:`);
-                lines.push(`  • Try different keywords`);
-                lines.push(`  • Use \`list_remembered\` to see available context`);
-                lines.push(`  • Use \`recall_best_solution("topic")\` for exact topic lookup`);
+                // Did-You-Mean: find nearest token in index vocab
+                const queryTokens = tokenize(query);
+                const suggestions = [];
+                if (_indexVocab.size > 0 && queryTokens.length > 0) {
+                    for (const qt of queryTokens.slice(0, 3)) {
+                        if (qt.length < 4)
+                            continue;
+                        let bestDist = 3;
+                        let bestTok = '';
+                        for (const v of _indexVocab) {
+                            if (v.length < 3 || Math.abs(v.length - qt.length) > 4)
+                                continue;
+                            const d = levenshtein(qt, v);
+                            if (d > 0 && d < bestDist) {
+                                bestDist = d;
+                                bestTok = v;
+                            }
+                        }
+                        if (bestTok)
+                            suggestions.push(`"${bestTok}" (instead of "${qt}")`);
+                    }
+                }
+                if (suggestions.length > 0) {
+                    lines.push(`💡 **Did you mean:** ${suggestions.join(', ')}?`);
+                }
+                else {
+                    lines.push(`\n💡 Tips:`);
+                    lines.push(`  • Try different keywords`);
+                    lines.push(`  • Use \`list_remembered\` to see available context`);
+                    lines.push(`  • Use \`recall_best_solution("topic")\` for exact topic lookup`);
+                }
             }
             return lines.join('\n');
         }
@@ -6999,7 +7109,14 @@ if (httpPort) {
         const url = new URL(req.url ?? '/', `http://localhost:${httpPort}`);
         if (url.pathname === '/health') {
             res.writeHead(200, { 'Content-Type': 'application/json' });
-            res.end(JSON.stringify({ status: 'ok', version: CURRENT_VERSION }));
+            res.end(JSON.stringify({
+                status: 'ok',
+                version: CURRENT_VERSION,
+                zeroResults: {
+                    total: zeroResultsTotal,
+                    last10: ZERO_RESULTS_LOG.slice(-10).map(e => ({ query: e.query, ts: new Date(e.ts).toISOString() })),
+                },
+            }));
             return;
         }
         if (url.pathname === '/mcp' || url.pathname === '/') {