@cachly-dev/mcp-server 0.5.47 → 0.5.49

package/dist/index.d.ts

@@ -14,6 +14,8 @@ declare function katakanaToRomaji(kata: string): string;
 /**
  * Arabic light stemmer — strips definite article and common prefix particles.
  * Handles: ال (al-), و (wa-), ب (bi-), ل (li-), ف (fa-), ك (ka-).
+ * Runs iteratively (max 3 passes) so compound prefixes like فال- are fully
+ * resolved: فالخطأ → الخطأ → خطأ.
  * Only strips when the result is still ≥ 3 chars to avoid over-stemming.
  */
 declare function arabicLightStem(word: string): string;
@@ -31,8 +33,9 @@ declare function expandCrossLingual(token: string): string[];
  * Features:
  *  • CJK → character bigrams
  *  • Katakana → additionally emits Hepburn romaji tokens (enables romaji queries)
- *  • Arabic/Hebrew → word tokenization with light Arabic stemming
- *  • All scripts → cross-lingual synonym expansion (EN↔JA↔ZH↔KO↔AR↔HE)
+ *  • Arabic → word tokenization + iterative prefix stemming (al-, wa-, bi-, li-, fa-)
+ *  • Hebrew → word tokenization + iterative prefix stemming (ha-, ve-, be-, le-, mi-, ke-)
+ *  • All scripts → cross-lingual synonym expansion (EN↔JA↔ZH↔KO↔AR↔HE↔DE↔FR↔ES↔IT↔PT)
  */
 declare function tokenize(text: string): string[];
 /**

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 = '0.5.26';
+const CURRENT_VERSION = '0.5.49';
 /**
  * ┌──────────────────────────────────────────────────────────────────────┐
  * │  EMBEDDING PROVIDER — pluggable, client-side first                  │
@@ -409,6 +409,10 @@ const CJK_RE = /[\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7
 const SEGMENT_RE = /([\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]+|[^\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]+)/g;
 // RTL Unicode ranges: Arabic (U+0600–U+06FF + extended), Hebrew (U+0590–U+05FF)
 const RTL_RE = /[\u0590-\u05ff\u0600-\u06ff\u0750-\u077f\u08a0-\u08ff\ufb1d-\ufb4f\ufb50-\ufdff\ufe70-\ufeff]/;
+// Distinguish Arabic from Hebrew within RTL segments
+const HEBREW_CHAR_RE = /[\u0590-\u05ff\ufb1d-\ufb4f]/;
+const ARABIC_CHAR_RE = /[\u0600-\u06ff\u0750-\u077f\u08a0-\u08ff\ufb50-\ufdff\ufe70-\ufeff]/;
+;
 // ── Katakana → Romaji conversion table (Hepburn system) ──────────────────────
 // Digraphs must be listed before single chars so they match first.
 const KATA_ROMAJI_TABLE = [
@@ -490,14 +494,51 @@ function katakanaToRomaji(kata) {
 /**
  * Arabic light stemmer — strips definite article and common prefix particles.
  * Handles: ال (al-), و (wa-), ب (bi-), ل (li-), ف (fa-), ك (ka-).
+ * Runs iteratively (max 3 passes) so compound prefixes like فال- are fully
+ * resolved: فالخطأ → الخطأ → خطأ.
  * Only strips when the result is still ≥ 3 chars to avoid over-stemming.
  */
 function arabicLightStem(word) {
-    if (word.startsWith('ال') && word.length > 4)
-        return word.slice(2);
-    if (word.length > 3 && 'وبلفك'.includes(word[0]) && RTL_RE.test(word[1]))
-        return word.slice(1);
-    return word;
+    let result = word;
+    for (let i = 0; i < 3; i++) {
+        const prev = result;
+        if (result.startsWith('ال') && result.length > 4) {
+            result = result.slice(2);
+        }
+        else if (result.length > 3 && 'وبلفك'.includes(result[0]) && RTL_RE.test(result[1])) {
+            result = result.slice(1);
+        }
+        if (result === prev)
+            break; // stable — no more prefixes to strip
+    }
+    return result;
+}
+/**
+ * Hebrew light stemmer — strips the definite article and common prefix particles
+ * that attach directly to words (no space) in Hebrew.
+ *
+ * Handles:
+ *   ה (ha-) — definite article:   הפריסה → פריסה
+ *   ו (ve-/u-) — conjunction “and”: ופריסה → פריסה
+ *   ב (be-/bi-) — preposition “in”: בסביבה → סביבה
+ *   ל (le-/li-) — preposition “for”: לשרת → שרת
+ *   מ (mi-/me-) — preposition “from”: מהשרת → השרת → שרת
+ *   כ (ke-/ki-) — preposition “like”: כשרת → שרת
+ *
+ * Iterative (max 3 passes): מהפריסה → הפריסה → פריסה.
+ * Only strips when result is still ≥ 3 chars.
+ */
+function hebrewLightStem(word) {
+    let result = word;
+    for (let i = 0; i < 3; i++) {
+        const prev = result;
+        if (result.length > 3 && 'הובלמכ'.includes(result[0]) && HEBREW_CHAR_RE.test(result[1])) {
+            result = result.slice(1);
+        }
+        if (result === prev)
+            break;
+    }
+    return result;
 }
 // ── Cross-lingual tech term dictionary ────────────────────────────────────────
 // Maps tokens (English, Japanese katakana, Chinese, Korean, Arabic, Hebrew)
@@ -507,8 +548,8 @@ function arabicLightStem(word) {
 // Keys must be lowercase. CJK values will be bigram-expanded at use time.
 const CROSS_LINGUAL_MAP = new Map([
     // English → all
-    ['deploy', ['デプロイ', '部署', '배포', 'deployment', 'deploying', 'bereitstellen', 'bereitstellung', 'déployer', 'déploiement', 'desplegar', 'despliegue', 'distribuire', 'distribuzione', 'implantar', 'implantação']],
-    ['deployment', ['デプロイ', '部署', '배포', 'deploy', 'bereitstellung', 'déploiement', 'despliegue', 'distribuzione', 'implantação']],
+    ['deploy', ['デプロイ', '部署', '배포', 'deployment', 'deploying', 'نشر', 'פריסה', 'bereitstellen', 'bereitstellung', 'déployer', 'déploiement', 'desplegar', 'despliegue', 'distribuire', 'distribuzione', 'implantar', 'implantação']],
+    ['deployment', ['デプロイ', '部署', '배포', 'deploy', 'نشر', 'פריסה', 'bereitstellung', 'déploiement', 'despliegue', 'distribuzione', 'implantação']],
     ['container', ['コンテナ', '容器', '컨테이너', 'docker']],
     ['server', ['サーバー', 'サーバ', '服务器', '서버', 'سيرفر', 'שרת']],
     ['database', ['データベース', '数据库', '데이터베이스', 'db', 'قاعدة البيانات', 'datenbank', 'base de données', 'base de datos', 'banco de dados']],
@@ -519,6 +560,10 @@ const CROSS_LINGUAL_MAP = new Map([
     ['build', ['ビルド', '构建', '빌드', 'بناء', 'בנייה', 'bauen', 'construire', 'construir', 'costruire']],
     ['test', ['テスト', '测试', '테스트', 'اختبار', 'בדיקה', 'testen', 'tester', 'probar', 'testare', 'testar']],
     ['auth', ['認証', '认证', '인증', 'authentication', 'login', 'oauth', 'مصادقة', 'אימות', 'authentifizierung', 'authentification', 'autenticación', 'autenticazione', 'autenticação']],
+    ['authentication', ['auth', 'مصادقة', 'אימות', '認証', '认证', '인증', 'login', 'oauth']],
+    ['environment', ['環境', '环境', '환경', 'env', 'بيئة', 'סביבה', 'umgebung', 'environnement', 'entorno', 'ambiente']],
+    ['secret', ['key', 'token', 'password', 'jwt', 'مفتاح', 'מפתח', 'geheimnis', 'schlüssel', 'clave', 'clé', 'chiave', 'chave']],
+    ['problem', ['error', 'issue', 'bug', 'failure', 'مشكلة', 'بعيا', '問題', '問題', '오류', 'fehler', 'problème', 'problema']],
     ['kubernetes', ['クベルネテス', 'k8s', 'kube']],
     ['network', ['ネットワーク', '网络', '네트워크', 'شبكة', 'רשת', 'netzwerk', 'réseau', 'red', 'rete', 'rede']],
     ['timeout', ['タイムアウト', '超时', '타임아웃', 'مهلة']],
@@ -557,7 +602,7 @@ const CROSS_LINGUAL_MAP = new Map([
     ['ci', ['pipeline', 'github actions', 'gitlab', 'jenkins', 'build']],
     ['cd', ['deploy', 'deployment', 'release']],
     ['debug', ['デバッグ', '调试', '디버그', 'debugging', 'breakpoint', 'تصحيح', 'איתור באגים', 'debuggen', 'déboguer', 'depurar']],
-    ['crash', ['クラッシュ', '崩溃', '크래시', 'panic', 'segfault', 'انهيار', 'absturz', 'panne', 'plantage', 'caída', 'falha']],
+    ['crash', ['クラッシュ', '崩溃', '크래시', 'panic', 'segfault', 'انهيار', 'קריסה', 'absturz', 'panne', 'plantage', 'caída', 'falha']],
     ['restart', ['再起動', '重启', '재시작', 'reboot', 'إعادة تشغيل', 'הפעלה מחדש', 'neustart', 'redémarrer', 'reiniciar', 'riavviare']],
     ['permission', ['権限', '权限', '권한', 'access', 'acl', 'chmod', 'صلاحية', 'הרשאה', 'berechtigung', 'berechtigungen', 'permiso', 'permesso', 'permissão']],
     // Japanese katakana → English
@@ -662,39 +707,89 @@ const CROSS_LINGUAL_MAP = new Map([
     ['إصلاح', ['fix']],
     ['بناء', ['build']],
     ['اختبار', ['test']],
-    ['مصادقة', ['auth']],
-    ['إعدادات', ['config']],
+    ['مصادقة', ['auth', 'authentication']],
+    ['إعدادات', ['config', 'configuration']],
     ['تثبيت', ['install']],
-    ['تحديث', ['update']],
+    ['تحديث', ['update', 'upgrade']],
     ['سجل', ['log']],
     ['صورة', ['image']],
     ['خدمة', ['service']],
     ['نسخ احتياطي', ['backup']],
-    ['مراقبة', ['monitor']],
+    ['مراقبة', ['monitor', 'monitoring']],
     ['اتصال', ['connection']],
     ['ذاكرة', ['memory']],
-    ['تصحيح', ['debug']],
+    ['تصحيح', ['debug', 'debugging']],
     ['شبكة', ['network']],
+    // Additional Arabic → English (covering blog post examples + common terms)
+    ['نشر', ['deploy', 'deployment']],
+    ['انهيار', ['crash', 'panic']],
+    ['طابور', ['queue']],
+    ['رمز', ['token', 'secret']],
+    ['كاش', ['cache']],
+    ['منفذ', ['port']],
+    ['سيرفر', ['server']],
+    ['قاعدة البيانات', ['database', 'db']],
+    ['إعادة تشغيل', ['restart', 'reboot']],
+    ['صلاحية', ['permission', 'access']],
+    ['تطبيق', ['application', 'app']],
+    ['مفتاح', ['key', 'token', 'secret']],
+    ['واجهة برمجية', ['api', 'interface']],
+    ['فحص الصحة', ['healthcheck', 'health']],
+    ['حاوية', ['container', 'docker']],
+    ['عقدة', ['node']],
+    ['سرعة', ['performance', 'speed']],
+    ['قرص', ['disk', 'storage']],
+    ['خادم', ['server']],
+    ['برنامج', ['application', 'software', 'app']],
+    // Arabic terms from blog examples
+    ['مشكلة', ['error', 'issue', 'bug', 'problem', 'failure']],
+    ['بيئة', ['environment', 'env']],
+    ['متغيرات', ['variables', 'env', 'environment']],
+    ['إنتاج', ['production', 'prod']],
+    ['سري', ['secret', 'key', 'token']],
+    ['مفتاح سري', ['secret', 'key']],
     // Hebrew → English
     ['שגיאה', ['error']],
     ['תיקון', ['fix']],
     ['בנייה', ['build']],
     ['בדיקה', ['test']],
-    ['אימות', ['auth']],
-    ['הגדרות', ['config']],
+    ['אימות', ['auth', 'authentication']],
+    ['הגדרות', ['config', 'configuration']],
     ['התקנה', ['install']],
-    ['עדכון', ['update']],
+    ['עדכון', ['update', 'upgrade']],
     ['יומן', ['log']],
     ['שרת', ['server']],
     ['שירות', ['service']],
     ['גיבוי', ['backup']],
-    ['ניטור', ['monitor']],
+    ['ניטור', ['monitor', 'monitoring']],
     ['חיבור', ['connection']],
     ['זיכרון', ['memory']],
     ['רשת', ['network']],
     ['תמונה', ['image']],
     ['סיסמה', ['password']],
-    ['הרשאה', ['permission']],
+    ['הרשאה', ['permission', 'access']],
+    // Additional Hebrew → English (covering blog post examples + common terms)
+    ['קריסה', ['crash', 'panic']],
+    ['תור', ['queue']],
+    ['מטמון', ['cache']],
+    ['אסימון', ['token', 'secret']],
+    ['פורט', ['port']],
+    ['איתור באגים', ['debug', 'debugging']],
+    ['הפעלה מחדש', ['restart', 'reboot']],
+    ['מיכל', ['container', 'docker']],
+    ['יישום', ['application', 'app']],
+    ['מפתח', ['key', 'token', 'secret']],
+    ['בדיקת תקינות', ['healthcheck', 'health']],
+    ['ממשק תכנות', ['api', 'interface']],
+    ['ביצועים', ['performance']],
+    ['אחסון', ['storage', 'disk']],
+    // Hebrew terms from blog examples
+    ['פריסה', ['deploy', 'deployment']],
+    ['בעיה', ['error', 'issue', 'bug', 'problem']],
+    ['סביבה', ['environment', 'env']],
+    ['ייצור', ['production', 'prod']],
+    ['סודי', ['secret', 'key', 'token']],
+    ['מפתח סודי', ['secret', 'key']],
     // ── European languages → English ─────────────────────────────────────────
     // German (Deutsch) → English
     ['fehler', ['error', 'err']],
@@ -892,8 +987,9 @@ function expandCrossLingual(token) {
  * Features:
  *  • CJK → character bigrams
  *  • Katakana → additionally emits Hepburn romaji tokens (enables romaji queries)
- *  • Arabic/Hebrew → word tokenization with light Arabic stemming
- *  • All scripts → cross-lingual synonym expansion (EN↔JA↔ZH↔KO↔AR↔HE)
+ *  • Arabic → word tokenization + iterative prefix stemming (al-, wa-, bi-, li-, fa-)
+ *  • Hebrew → word tokenization + iterative prefix stemming (ha-, ve-, be-, le-, mi-, ke-)
+ *  • All scripts → cross-lingual synonym expansion (EN↔JA↔ZH↔KO↔AR↔HE↔DE↔FR↔ES↔IT↔PT)
  */
 function tokenize(text) {
     const tokens = [];
@@ -932,12 +1028,15 @@ function tokenize(text) {
             }
         }
         else if (RTL_RE.test(seg)) {
-            // RTL segment (Arabic / Hebrew) — space-separated words with light stemming
+            // RTL segment — split on spaces, apply language-specific prefix stemming
+            // Arabic: strips ال/و/ب/ل/ف/ك prefixes iteratively
+            // Hebrew: strips ה/ו/ב/ל/מ/כ prefixes iteratively
+            const isHebrew = HEBREW_CHAR_RE.test(seg);
             const words = seg.trim().split(/\s+/);
             for (const w of words) {
                 if (w.length <= 1 || STOPWORDS.has(w))
                     continue;
-                const stemmed = arabicLightStem(w);
+                const stemmed = isHebrew ? hebrewLightStem(w) : arabicLightStem(w);
                 if (stemmed.length > 1 && !STOPWORDS.has(stemmed))
                     tokens.push(stemmed);
             }