@cachly-dev/mcp-server 0.5.34 → 0.5.35

package/dist/index.d.ts

@@ -1,11 +1,16 @@
 #!/usr/bin/env node
 /**
  * Stopwords — filtered out during tokenization.
- * Covers: English, German, French, Spanish, Italian, Portuguese.
+ * Covers: English, German, French, Spanish, Italian, Portuguese,
+ *         Chinese (Simplified + Traditional), Japanese, Korean.
  * Keeps the index small and scores meaningful.
  */
 declare const STOPWORDS: Set<string>;
-/** Tokenize text into meaningful keywords. Handles unicode (accented chars). */
+/**
+ * Tokenize text into meaningful keywords.
+ * Handles accented Latin characters and CJK (Chinese, Japanese, Korean).
+ * CJK text is split into character bigrams (the standard for indexing logographic languages).
+ */
 declare function tokenize(text: string): string[];
 /**
  * Split a multi-topic query into sub-queries.

package/dist/index.js

@@ -288,12 +288,14 @@ function embedProviderHint() {
 //   • Recency boost — newer entries rank higher (exponential decay, 7-day half-life)
 //   • Multi-query splitting — numbered lists, semicolons, conjunctions
 //   • Fuzzy matching with Levenshtein distance ≤ 2
-//   • Multilingual stopwords (EN, DE, FR, ES, IT, PT)
+//   • Multilingual stopwords (EN, DE, FR, ES, IT, PT, ZH, JA, KO)
+//   • CJK character bigram extraction — handles Chinese, Japanese, Korean
 //   • Pipeline Redis reads for performance
 //
 /**
  * Stopwords — filtered out during tokenization.
- * Covers: English, German, French, Spanish, Italian, Portuguese.
+ * Covers: English, German, French, Spanish, Italian, Portuguese,
+ *         Chinese (Simplified + Traditional), Japanese, Korean.
  * Keeps the index small and scores meaningful.
  */
 const STOPWORDS = new Set([
@@ -367,14 +369,61 @@ const STOPWORDS = new Set([
     'eu', 'tu', 'ele', 'ela', 'nós', 'eles', 'elas', 'você', 'vocês',
     'não', 'já', 'sim', 'bem', 'muito', 'também', 'ainda', 'sempre',
     'todo', 'toda', 'todos', 'cada', 'outro', 'outra', 'quando', 'porque',
+    // ── Chinese (Simplified + Traditional) — high-frequency function characters ──
+    '的', '了', '是', '在', '不', '和', '我', '他', '这', '中', '大', '为',
+    '上', '个', '国', '以', '要', '就', '出', '说', '们', '有', '来', '到',
+    '时', '地', '年', '得', '着', '那', '过', '后', '还', '与', '也', '可',
+    '于', '从', '但', '而', '被', '把', '让', '使', '对', '很', '都', '一',
+    '会', '没', '人', '它', '这个', '那个', '什么', '如果', '因为', '所以',
+    '已经', '可以', '这些', '那些', '我们', '他们', '她们', '它们',
+    // ── Japanese — common hiragana particles and auxiliary verbs ──
+    'の', 'は', 'が', 'を', 'に', 'で', 'と', 'も', 'や', 'か', 'な', 'ね',
+    'よ', 'わ', 'て', 'い', 'う', 'え', 'お', 'き', 'く', 'け', 'こ', 'さ',
+    'し', 'す', 'せ', 'そ', 'た', 'ち', 'つ', 'ぬ', 'ん', 'から', 'まで',
+    'より', 'へ', 'です', 'ます', 'ない', 'ある', 'いる', 'する', 'こと',
+    'もの', 'ので', 'では', 'には', 'との', 'への', 'から', 'まで',
+    // ── Korean — common particles and auxiliary forms ──
+    '이', '가', '은', '는', '을', '를', '의', '에', '와', '과', '도',
+    '로', '에서', '한', '하', '있', '없', '것', '수', '않', '들',
+    '이다', '하다', '이고', '이며', '라고', '에게', '에서', '으로',
 ]);
-/** Tokenize text into meaningful keywords. Handles unicode (accented chars). */
+// CJK Unicode ranges used for bigram extraction
+const CJK_RE = /[\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]/;
+const SEGMENT_RE = /([\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]+|[^\u3040-\u30ff\u3400-\u4dbf\u4e00-\u9fff\uf900-\ufaff\uac00-\ud7af]+)/g;
+/**
+ * Tokenize text into meaningful keywords.
+ * Handles accented Latin characters and CJK (Chinese, Japanese, Korean).
+ * CJK text is split into character bigrams (the standard for indexing logographic languages).
+ */
 function tokenize(text) {
-    return text
-        .toLowerCase()
-        .replace(/[^a-záàâãäåæçéèêëíìîïñóòôõöúùûüýÿßœ0-9:_\-./]/g, ' ')
-        .split(/\s+/)
-        .filter(w => w.length > 1 && !STOPWORDS.has(w));
+    const tokens = [];
+    const lower = text.toLowerCase();
+    for (const [seg] of lower.matchAll(SEGMENT_RE)) {
+        if (CJK_RE.test(seg)) {
+            // CJK segment — extract overlapping character bigrams
+            for (let i = 0; i < seg.length - 1; i++) {
+                const bigram = seg[i] + seg[i + 1];
+                if (!STOPWORDS.has(seg[i]) && !STOPWORDS.has(bigram)) {
+                    tokens.push(bigram);
+                }
+            }
+            // Trailing single character (for 1-char CJK terms)
+            if (seg.length >= 1) {
+                const last = seg[seg.length - 1];
+                if (!STOPWORDS.has(last))
+                    tokens.push(last);
+            }
+        }
+        else {
+            // Latin / other — existing logic
+            const words = seg
+                .replace(/[^a-záàâãäåæçéèêëíìîïñóòôõöúùûüýÿßœ0-9:_\-./]/g, ' ')
+                .split(/\s+/)
+                .filter(w => w.length > 1 && !STOPWORDS.has(w));
+            tokens.push(...words);
+        }
+    }
+    return tokens;
 }
 /**
  * Split a multi-topic query into sub-queries.
@@ -4650,6 +4699,8 @@ const EDITOR_FILES = {
     windsurf: '.windsurf/mcp.json',
     copilot: '.vscode/mcp.json',
     continue: '.continue/config.json',
+    cline: '.vscode/mcp.json', // Cline respects VS Code project MCP config
+    zed: '.zed/settings.json', // Zed project-level context_servers
 };
 const CLAUDE_MD_MARKER_START = '<!-- cachly-brain-start -->';
 const CLAUDE_MD_MARKER_END = '<!-- cachly-brain-end -->';
@@ -4664,6 +4715,20 @@ function buildMcpConfig(apiKey, instanceId, editor) {
             },
         }, null, 2);
     }
+    if (editor === 'zed') {
+        return JSON.stringify({
+            context_servers: {
+                cachly: {
+                    command: {
+                        path: 'npx',
+                        args: ['-y', '@cachly-dev/mcp-server@latest'],
+                        env: { CACHLY_API_URL: 'https://api.cachly.dev', CACHLY_JWT: apiKey, CACHLY_BRAIN_INSTANCE_ID: instanceId },
+                    },
+                    settings: {},
+                },
+            },
+        }, null, 2);
+    }
     return JSON.stringify({
         mcpServers: {
             cachly: {
@@ -4992,6 +5057,8 @@ if (process.argv[2] === 'setup') {
     // ── Step 3: Detect editors ────────────────────────────────────────────────
     const cwd = process.cwd();
     const detected = [];
+    const { homedir } = await import('node:os');
+    const home = homedir();
     // Claude Code: always include (CLAUDE.md is universal)
     detected.push('claude');
     if (existsSync(resolve(cwd, '.cursor')))
@@ -5002,7 +5069,21 @@ if (process.argv[2] === 'setup') {
         detected.push('copilot');
     if (existsSync(resolve(cwd, '.continue')))
         detected.push('continue');
-    const editorLabel = (e) => ({ claude: 'Claude Code', cursor: 'Cursor', windsurf: 'Windsurf', copilot: 'GitHub Copilot', continue: 'Continue.dev' })[e] ?? e;
+    // Cline: VS Code extension — detect via globalStorage on the machine
+    const clineGlobalDir = process.platform === 'darwin'
+        ? resolve(home, 'Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev')
+        : resolve(home, '.vscode/extensions');
+    if (existsSync(clineGlobalDir) && !detected.includes('copilot'))
+        detected.push('cline');
+    else if (existsSync(clineGlobalDir))
+        detected.push('cline'); // both copilot + cline share .vscode/mcp.json — fine
+    // Zed editor — detect via app data dir
+    const zedDir = process.platform === 'darwin'
+        ? resolve(home, 'Library/Application Support/Zed')
+        : resolve(home, '.config/zed');
+    if (existsSync(zedDir))
+        detected.push('zed');
+    const editorLabel = (e) => ({ claude: 'Claude Code', cursor: 'Cursor', windsurf: 'Windsurf', copilot: 'GitHub Copilot', continue: 'Continue.dev', cline: 'Cline (VSCode)', zed: 'Zed' })[e] ?? e;
     console.log(`Step 3: Detected editors: ${detected.map(editorLabel).join(', ')}`);
     const editorsRaw = await ask(`   Configure for which? [all / ${detected.join('/')}] (Enter = all): `);
     const editorsToSetup = !editorsRaw || editorsRaw.toLowerCase() === 'all'
@@ -5043,6 +5124,86 @@ if (process.argv[2] === 'setup') {
     rl.close();
     console.log(`\n🧠  Done! Restart your editor — the \`cachly\` MCP tools will appear.`);
     console.log(`    Your AI now has persistent memory across every session.\n`);
+    // ── Step 7: Show first session_start preview (Aha moment) ────────────────
+    // Fetch brain health from the API to show the user what their agent will see.
+    process.stdout.write('\n⏳ Fetching your Brain health preview...');
+    try {
+        const brainRes = await fetch(`${API_URL}/api/v1/instances/${instance.id}/brain/stats`, {
+            headers: { Authorization: `Bearer ${token}` },
+            signal: AbortSignal.timeout(6000),
+        });
+        if (brainRes.ok) {
+            const brainData = await brainRes.json();
+            const lessons = brainData.lesson_count ?? 0;
+            const contexts = brainData.context_count ?? 0;
+            const score = brainData.quality_score ?? 0;
+            const level = lessons === 0 ? 'Intern 🌱' :
+                lessons < 10 ? 'Junior Dev 🔧' :
+                    lessons < 30 ? 'Mid Dev ⚡' :
+                        lessons < 60 ? 'Senior Dev 🧠' :
+                            lessons < 100 ? 'Staff Eng 🚀' : 'Principal Eng 🏆';
+            console.log(' ✓\n');
+            console.log('┌──────────────────────────────────────────────────────┐');
+            console.log(`│  🧠  Brain Health Report                            │`);
+            console.log('├──────────────────────────────────────────────────────┤');
+            console.log(`│  Lessons stored    : ${String(lessons).padEnd(6)} │  Level: ${level.padEnd(20)}│`);
+            console.log(`│  Context entries   : ${String(contexts).padEnd(6)} │  Quality score: ${String(Math.round(score * 100)).padEnd(3)}%       │`);
+            if (lessons === 0) {
+                console.log('├──────────────────────────────────────────────────────┤');
+                console.log('│  Your Brain is empty and ready to learn.            │');
+                console.log('│  After your first coding session it will contain:   │');
+                console.log('│    • Lessons from every bug you fixed               │');
+                console.log('│    • Your project\'s indexed files                   │');
+                console.log('│    • A session summary for next time                │');
+            }
+            console.log('└──────────────────────────────────────────────────────┘');
+        }
+        else {
+            console.log(' (skipped)');
+        }
+    }
+    catch {
+        console.log(' (skipped)');
+    }
+    console.log(`\n   Your AI will call session_start automatically when it uses the first tool.`);
+    console.log(`   To see your Brain at any time: https://cachly.dev/instances/${instance.id}\n`);
+    process.exit(0);
+}
+// ── CLI: cachly index <dir> ───────────────────────────────────────────────────
+// Usage: npx @cachly-dev/mcp-server@latest index [./path/to/project]
+// Indexes the project directory into the Brain — perfect for CI/CD cron jobs.
+if (process.argv[2] === 'index') {
+    const { resolve } = await import('node:path');
+    const argv = process.argv.slice(3);
+    const flag = (name) => { const i = argv.indexOf(`--${name}`); return i !== -1 ? argv[i + 1] : undefined; };
+    const dir = resolve(flag('dir') ?? argv.find(a => !a.startsWith('--')) ?? '.');
+    const instanceId = flag('instance-id') ?? process.env.CACHLY_BRAIN_INSTANCE_ID;
+    const maxFiles = parseInt(flag('max-files') ?? '500', 10);
+    const namespace = flag('namespace') ?? 'cachly:sem:code';
+    if (!instanceId || !JWT) {
+        console.error('\n❌  CACHLY_BRAIN_INSTANCE_ID and CACHLY_JWT must be set\n');
+        console.error('   export CACHLY_BRAIN_INSTANCE_ID=<uuid>');
+        console.error('   export CACHLY_JWT=<cky_live_...>');
+        console.error('   npx @cachly-dev/mcp-server@latest index ./my-project\n');
+        process.exit(1);
+    }
+    console.log(`\n📂  Indexing: ${dir}`);
+    console.log(`    Instance: ${instanceId.slice(0, 8)}…  Max files: ${maxFiles}\n`);
+    try {
+        const result = await handleTool('index_project', {
+            instance_id: instanceId,
+            dir,
+            max_files: maxFiles,
+            ttl: 86400 * 7,
+            namespace,
+        });
+        console.log(result);
+        console.log('\n✅  Indexing complete.\n');
+    }
+    catch (err) {
+        console.error(`\n❌  Indexing failed: ${err.message}\n`);
+        process.exit(1);
+    }
     process.exit(0);
 }
 // ── Start ─────────────────────────────────────────────────────────────────────