stellavault 0.4.2 → 0.4.4

package/CLAUDE.md

@@ -1,10 +1,13 @@
 # Stellavault — Project Rules
 
 ## Project Info
-- **Name**: Stellavault (분산 지식 인텔리전스 플랫폼)
+- **Name**: Stellavault — Self-compiling knowledge MCP server
+- **Version**: 0.4.2
 - **GitHub**: https://github.com/Evanciel/stellavault
+- **npm**: stellavault (171KB)
 - **Stack**: Node.js 20+, TypeScript, ESM, Monorepo (npm workspaces)
-- **Packages**: core, cli, graph, sync
+- **Packages**: core (21 MCP tools), cli (39+ commands), graph (3D UI), sync
+- **Architecture**: Karpathy self-compiling KB (ingest→compile→wiki→session→flush loop)
 
 ## Autopilot 추가 규칙
 

package/README.md

@@ -2,7 +2,7 @@
 
 > **Self-compiling knowledge MCP server** — ingest anything, auto-organize into Zettelkasten wiki, and let Claude access your entire knowledge base.
 
-Drop a PDF, paste a YouTube link, type a thought — Stellavault compiles it into structured knowledge, connects the dots, and gives your AI agent full access through 20 MCP tools.
+Drop a PDF, paste a YouTube link, type a thought — Stellavault compiles it into structured knowledge, connects the dots, and gives your AI agent full access through 21 MCP tools.
 
 <p align="center">
   <img src="images/screenshots/graph-dark-full.png" alt="3D Knowledge Graph" width="800" />
@@ -21,7 +21,7 @@ PDF, DOCX, YouTube, URL, text — everything goes through the same pipeline. You
 ```bash
 claude mcp add stellavault -- stellavault serve
 ```
-20 MCP tools give Claude direct access to search, ask, draft, and navigate your entire knowledge base.
+21 MCP tools give Claude direct access to search, ask, draft, and navigate your entire knowledge base.
 
 ## 5-Minute Setup
 
@@ -66,6 +66,27 @@ stellavault draft --format outline        # All-knowledge outline
 
 Or in Claude Code: *"Write a blog post about machine learning from my notes"* — Claude uses MCP `generate-draft` tool (free, no API key).
 
+## Self-Evolving Memory (Karpathy's Compounding Loop)
+
+```
+Session → session-save → daily-log → flush → wiki
+  ↑                                            ↓
+  └──── Claude reads wiki via MCP (20 tools) ←─┘
+```
+
+Every conversation makes your knowledge base smarter:
+
+```bash
+# Auto-capture session summary to daily log
+echo "Decided to use JWT. Lesson: never store tokens in localStorage" | stellavault session-save
+
+# Flush daily logs → extract concepts → rebuild wiki
+stellavault flush
+
+# Or set up Claude Code hooks for full automation
+# See: docs/hooks-setup.md
+```
+
 ## Daily Commands
 
 ```bash
@@ -74,10 +95,11 @@ stellavault brief                              # Morning knowledge briefing
 stellavault decay                              # What's fading from memory?
 stellavault lint                               # Health score (0-100)
 stellavault learn                              # AI learning path
+stellavault flush                              # Daily logs → wiki compilation
 stellavault digest --visual                    # Weekly Mermaid chart report
 ```
 
-## MCP Tools (20)
+## MCP Tools (21)
 
 | Tool | What it does |
 |------|-------------|
@@ -99,6 +121,16 @@ stellavault digest --visual                    # Weekly Mermaid chart report
 | `create-snapshot` / `load-snapshot` | Context snapshots |
 | `generate-claude-md` | Auto-generate CLAUDE.md |
 | `export` | JSON/CSV export |
+| `federated-search` | P2P federated search |
+
+## Self-Evolving Commands
+
+```bash
+stellavault session-save              # Capture session summary to daily log
+stellavault flush                     # Daily logs → wiki (Karpathy compile)
+stellavault promote note.md --to lit  # Upgrade note stage
+stellavault autopilot                 # Full cycle: inbox → compile → lint → archive
+```
 
 ## Zettelkasten (Luhmann + Karpathy)
 
@@ -161,6 +193,20 @@ stellavault autopilot                          # Full cycle: inbox → compile
 | 3D | React Three Fiber + Three.js |
 | AI | MCP (Model Context Protocol) + Anthropic SDK |
 
+## Full Feature List
+
+| Category | Features |
+|----------|----------|
+| **Capture** | ingest (URL/YouTube/PDF/DOCX/PPTX/XLSX/text), fleeting, web drag & drop, mobile PWA |
+| **Organize** | Zettelkasten 3-stage, auto index codes, wikilink auto-connect, configurable folders |
+| **Distill** | compile (raw→wiki), lint (health score), gaps, contradictions, duplicates |
+| **Express** | draft (blog/report/outline), --ai (Claude API), MCP generate-draft (free) |
+| **Memory** | FSRS decay tracking, session-save (daily logs), flush (logs→wiki), compounding loop |
+| **Search** | hybrid (BM25+vector+RRF), multilingual (50+ langs), ask Q&A |
+| **Visualize** | 3D graph, constellation, heatmap, timeline, decay overlay, dark/light |
+| **AI Integration** | 21 MCP tools, Claude Code hooks, Anthropic SDK, generate-draft |
+| **CLI** | 39+ commands, `sv` alias |
+
 ## License
 
 MIT

package/index.html

@@ -4,7 +4,7 @@
 <meta charset="UTF-8">
 <meta name="viewport" content="width=device-width, initial-scale=1.0">
 <title>Stellavault — Self-Compiling Knowledge MCP Server</title>
-<meta name="description" content="Drop anything. It organizes itself. Claude knows everything you know. Self-compiling Zettelkasten with 20 MCP tools.">
+<meta name="description" content="Drop anything. It organizes itself. Claude knows everything you know. Self-compiling Zettelkasten with 21 MCP tools.">
 <meta name="theme-color" content="#6366f1">
 <meta property="og:title" content="Stellavault — Self-Compiling Knowledge MCP Server">
 <meta property="og:description" content="Drop anything. It organizes itself. Claude knows everything you know.">
@@ -233,7 +233,7 @@ footer{padding:60px 0;text-align:center}
   <div class="hero-inner container">
     <div class="hero-badge">open source <span class="sep">&#9679;</span> local-first <span class="sep">&#9679;</span> MIT license</div>
     <h1>Drop anything.<br>It <em>compiles itself</em><br>into knowledge.</h1>
-    <p class="hero-sub">Self-compiling Zettelkasten MCP server. Ingest PDFs, YouTube, documents&mdash;auto-organized into linked wiki. Claude accesses your entire knowledge base through 20 MCP tools.</p>
+    <p class="hero-sub">Self-compiling Zettelkasten MCP server. Ingest PDFs, YouTube, documents&mdash;auto-organized into linked wiki. Claude accesses your entire knowledge base through 21 MCP tools.</p>
     <div class="hero-cta">
       <button class="cta-install" id="ctaInstall">
         <span class="prompt">$</span> npm install -g stellavault
@@ -243,7 +243,7 @@ footer{padding:60px 0;text-align:center}
       <a href="https://github.com/Evanciel/stellavault" class="cta-mobile">View on GitHub &rarr;</a>
     </div>
     <div class="hero-proof">
-      <span><span class="proof-num">157KB</span> on npm</span>
+      <span><span class="proof-num">171KB</span> on npm</span>
       <span class="proof-sep">&#9679;</span>
       <span><span class="proof-num">20</span> MCP tools</span>
       <span class="proof-sep">&#9679;</span>
@@ -285,7 +285,7 @@ footer{padding:60px 0;text-align:center}
         <div class="problem-title">AI doesn't know what you know</div>
         <div class="problem-desc">ChatGPT has the internet. It doesn't have your notes, your decisions, your context.</div>
         <span class="problem-arrow">&darr;</span>
-        <div class="problem-solution">20 MCP tools &mdash; Claude searches, asks, drafts from your vault</div>
+        <div class="problem-solution">21 MCP tools &mdash; Claude searches, asks, drafts from your vault</div>
       </div>
     </div>
   </div>
@@ -308,6 +308,27 @@ footer{padding:60px 0;text-align:center}
   </div>
 </section>
 
+<!-- ═══ COMPOUNDING LOOP ═══ -->
+<section style="padding:80px 0">
+  <div class="container" style="text-align:center">
+    <div class="section-label reveal">The Compounding Loop</div>
+    <h2 class="section-title reveal">Gets smarter every session.</h2>
+    <p class="section-desc reveal" style="margin:0 auto 40px">Every conversation with Claude becomes knowledge. Knowledge improves the next conversation. Karpathy's self-evolving architecture.</p>
+    <div class="reveal" style="max-width:700px;margin:0 auto;padding:32px;background:var(--surface);border:1px solid var(--border);border-radius:14px;font-family:var(--font-mono);font-size:13px;line-height:2.2;color:var(--text-dim);text-align:left">
+      <span style="color:var(--cyan)">Session</span> &rarr; <code style="color:var(--violet-bright)">session-save</code> &rarr; <span style="color:var(--text)">daily-log</span> &rarr; <code style="color:var(--violet-bright)">flush</code> &rarr; <span style="color:var(--emerald)">wiki</span><br>
+      <span style="color:var(--text-muted);margin-left:16px">&uarr;</span><span style="padding-left:280px;color:var(--text-muted)">&darr;</span><br>
+      <span style="color:var(--text-muted);margin-left:16px">&larr;&larr; Claude reads wiki via MCP (20 tools) &larr;&larr;&larr;</span>
+    </div>
+    <div style="display:flex;gap:16px;justify-content:center;margin-top:28px;flex-wrap:wrap" class="reveal">
+      <div style="padding:14px 20px;background:var(--surface);border:1px solid var(--border);border-radius:10px;font-size:12px"><span style="color:var(--violet-bright);font-family:var(--font-mono)">session-save</span><br><span style="color:var(--text-dim)">Auto-capture decisions &amp; lessons</span></div>
+      <div style="padding:14px 20px;background:var(--surface);border:1px solid var(--border);border-radius:10px;font-size:12px"><span style="color:var(--violet-bright);font-family:var(--font-mono)">flush</span><br><span style="color:var(--text-dim)">Daily logs &rarr; wiki compilation</span></div>
+      <div style="padding:14px 20px;background:var(--surface);border:1px solid var(--border);border-radius:10px;font-size:12px"><span style="color:var(--violet-bright);font-family:var(--font-mono)">Claude Code hooks</span><br><span style="color:var(--text-dim)">Fully automatic capture</span></div>
+    </div>
+  </div>
+</section>
+
+<div class="divider"></div>
+
 <!-- ═══ SCREENSHOTS ═══ -->
 <section class="s-screenshots">
   <div class="container">
@@ -345,7 +366,8 @@ footer{padding:60px 0;text-align:center}
 <section class="s-mcp">
   <div class="container">
     <div class="section-label reveal">Claude Integration</div>
-    <h2 class="section-title reveal">20 MCP tools. Claude knows<br>everything you know.</h2>
+    <h2 class="section-title reveal">21 MCP tools. Claude knows<br>everything you know.</h2>
+    <p class="section-desc reveal">One command connects your vault to Claude Code. No API key needed.</p>
     <p class="section-desc reveal">One command connects your vault to Claude Code. No API key needed.</p>
     <div class="mcp-connect reveal"><span class="prompt">$</span> <code>claude mcp add stellavault -- stellavault serve</code></div>
     <div class="mcp-grid reveal">
@@ -365,6 +387,10 @@ footer{padding:60px 0;text-align:center}
       <div class="mcp-tool"><span class="dot"></span><code>get-morning-brief</code><span class="desc">daily briefing</span></div>
       <div class="mcp-tool"><span class="dot"></span><code>list-topics</code><span class="desc">topic cloud</span></div>
       <div class="mcp-tool"><span class="dot"></span><code>export</code><span class="desc">JSON / CSV</span></div>
+      <div class="mcp-tool"><span class="dot"></span><code>find-decisions</code><span class="desc">decision lookup</span></div>
+      <div class="mcp-tool"><span class="dot"></span><code>create-snapshot</code><span class="desc">context save/load</span></div>
+      <div class="mcp-tool"><span class="dot"></span><code>generate-claude-md</code><span class="desc">auto CLAUDE.md</span></div>
+      <div class="mcp-tool"><span class="dot"></span><code>federated-search</code><span class="desc">P2P search</span></div>
     </div>
   </div>
 </section>
@@ -415,9 +441,9 @@ footer{padding:60px 0;text-align:center}
       <div class="tech-item"><span class="label">Search</span><span class="value">BM25 + Cosine + RRF</span></div>
       <div class="tech-item"><span class="label">Memory</span><span class="value">FSRS repetition</span></div>
       <div class="tech-item"><span class="label">3D Graph</span><span class="value">React Three Fiber</span></div>
-      <div class="tech-item"><span class="label">AI Protocol</span><span class="value">MCP (20 tools)</span></div>
+      <div class="tech-item"><span class="label">AI Protocol</span><span class="value">MCP (21 tools)</span></div>
       <div class="tech-item"><span class="label">Runtime</span><span class="value">Node.js 20+ ESM</span></div>
-      <div class="tech-item"><span class="label">Package</span><span class="value">157KB on npm</span></div>
+      <div class="tech-item"><span class="label">Package</span><span class="value">171KB on npm</span></div>
     </div>
   </div>
 </section>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "stellavault",
-  "version": "0.4.2",
+  "version": "0.4.4",
   "description": "Self-compiling knowledge MCP server — ingest anything, auto-organize into Zettelkasten wiki, search with AI, and let Claude access your entire knowledge base.",
   "repository": {
     "type": "git",

package/packages/cli/dist/commands/adr-cmd.d.ts

@@ -0,0 +1,7 @@
+export declare function adrCommand(title: string, options: {
+    context?: string;
+    options?: string;
+    decision?: string;
+    consequences?: string;
+}): Promise<void>;
+//# sourceMappingURL=adr-cmd.d.ts.map

package/packages/cli/dist/commands/adr-cmd.js

@@ -0,0 +1,45 @@
+// stellavault adr — Architecture Decision Record 구조화 인제스트
+// 의사결정 기록: 제목/맥락/옵션/결정/결과
+import chalk from 'chalk';
+import { loadConfig, ingest } from '@stellavault/core';
+export async function adrCommand(title, options) {
+    if (!title) {
+        console.error(chalk.yellow('Usage: stellavault adr "Decision Title" --context "..." --options "..." --decision "..." --consequences "..."'));
+        process.exit(1);
+    }
+    const config = loadConfig();
+    const now = new Date().toISOString().split('T')[0];
+    const content = [
+        `# ADR: ${title}`,
+        '',
+        `**Date:** ${now}`,
+        `**Status:** Accepted`,
+        '',
+        '## Context',
+        options.context ?? '<!-- Why is this decision needed? -->',
+        '',
+        '## Options Considered',
+        options.options ?? '<!-- What alternatives were evaluated? -->',
+        '',
+        '## Decision',
+        options.decision ?? '<!-- What was decided and why? -->',
+        '',
+        '## Consequences',
+        options.consequences ?? '<!-- What are the implications? -->',
+        '',
+    ].join('\n');
+    const result = ingest(config.vaultPath, {
+        type: 'text',
+        content,
+        tags: ['adr', 'decision'],
+        title: `ADR: ${title}`,
+        stage: 'literature',
+    }, config.folders);
+    console.log(chalk.green(`ADR created: ${title}`));
+    console.log(chalk.dim(`  Saved: ${result.savedTo}`));
+    console.log(chalk.dim(`  Stage: literature`));
+    console.log(chalk.dim(`  Tags: adr, decision`));
+    console.log('');
+    console.log(chalk.dim(`Find later: stellavault ask "why did we choose ${title}?"`));
+}
+//# sourceMappingURL=adr-cmd.js.map

package/packages/cli/dist/commands/ask-cmd.d.ts

@@ -1,4 +1,5 @@
 export declare function askCommand(question: string, options: {
     save?: boolean;
+    quotes?: boolean;
 }): Promise<void>;
 //# sourceMappingURL=ask-cmd.d.ts.map

package/packages/cli/dist/commands/ask-cmd.js

@@ -17,6 +17,7 @@ export async function askCommand(question, options) {
         limit: 10,
         save: options.save ?? false,
         vaultPath: config.vaultPath,
+        mode: options.quotes ? 'quotes' : 'default',
     });
     // 출력
     console.log('');

package/packages/cli/dist/commands/draft-cmd.d.ts

@@ -1,5 +1,6 @@
 export declare function draftCommand(topic: string | undefined, options: {
     format?: string;
     ai?: boolean;
+    blueprint?: string;
 }): Promise<void>;
 //# sourceMappingURL=draft-cmd.d.ts.map

package/packages/cli/dist/commands/draft-cmd.js

@@ -13,7 +13,15 @@ export async function draftCommand(topic, options) {
     const format = (options.format ?? 'blog');
     try {
         // Step 1: rule-based 초안 생성 (스캐폴딩 + 소스 수집)
-        const result = generateDraft(config.vaultPath, { topic, format }, config.folders);
+        // Blueprint 파싱: "Chapter 1:tag1,tag2; Chapter 2:tag3" 형식
+        let blueprint;
+        if (options.blueprint) {
+            blueprint = options.blueprint.split(';').map(s => {
+                const [title, tagsStr] = s.split(':').map(p => p.trim());
+                return { title, tags: tagsStr?.split(',').map(t => t.trim()) };
+            });
+        }
+        const result = generateDraft(config.vaultPath, { topic, format, blueprint }, config.folders);
         if (options.ai) {
             // Step 2: --ai 모드 → Claude API로 실제 글 생성
             console.log(chalk.dim('  AI mode: generating with Claude...'));

package/packages/cli/dist/commands/ingest-cmd.js

@@ -1,13 +1,47 @@
 // stellavault ingest — 통합 인제스트 (URL, 텍스트, 파일 → 자동 분류 저장)
 import chalk from 'chalk';
 import { loadConfig, ingest, promoteNote } from '@stellavault/core';
-import { readFileSync, existsSync } from 'node:fs';
-import { extname, resolve } from 'node:path';
+import { readFileSync, existsSync, readdirSync, statSync } from 'node:fs';
+import { extname, resolve, join } from 'node:path';
 export async function ingestCommand(input, options) {
     if (!input) {
-        console.error(chalk.yellow('Usage: stellavault ingest <url|file|text> [--tags t1,t2] [--stage fleeting|literature|permanent]'));
+        console.error(chalk.yellow('Usage: stellavault ingest <url|file|text|folder/> [--tags t1,t2]'));
         process.exit(1);
     }
+    // 배치 모드: 폴더 경로이면 내부 파일 전부 처리
+    if (existsSync(input) && statSync(input).isDirectory()) {
+        const files = readdirSync(input)
+            .filter(f => /\.(md|txt|pdf|docx|pptx|xlsx|xls|csv)$/i.test(f))
+            .map(f => join(input, f));
+        if (files.length === 0) {
+            console.error(chalk.yellow(`No supported files found in ${input}`));
+            process.exit(1);
+        }
+        console.log(chalk.dim(`Batch ingest: ${files.length} files from ${input}\n`));
+        let success = 0;
+        const failed = [];
+        for (let i = 0; i < files.length; i++) {
+            const file = files[i];
+            const name = file.split(/[/\\]/).pop() ?? file;
+            const progress = `[${i + 1}/${files.length}]`;
+            process.stderr.write(`\r${chalk.dim(progress)} ${name}...`);
+            try {
+                await ingestCommand(file, { ...options, title: undefined });
+                success++;
+            }
+            catch (err) {
+                failed.push(`${name}: ${err instanceof Error ? err.message : 'error'}`);
+            }
+        }
+        process.stderr.write('\r' + ' '.repeat(80) + '\r');
+        console.log(chalk.green(`Batch complete: ${success}/${files.length} files ingested`));
+        if (failed.length > 0) {
+            console.log(chalk.yellow(`\nFailed (${failed.length}):`));
+            for (const f of failed)
+                console.log(chalk.yellow(`  - ${f}`));
+        }
+        return;
+    }
     const config = loadConfig();
     const tags = options.tags?.split(',').map(t => t.trim()) ?? [];
     const stage = (options.stage ?? 'fleeting');

package/packages/cli/dist/index.js

@@ -26,6 +26,7 @@ import { compileCommand } from './commands/compile-cmd.js';
 import { draftCommand } from './commands/draft-cmd.js';
 import { sessionSaveCommand } from './commands/session-cmd.js';
 import { flushCommand } from './commands/flush-cmd.js';
+import { adrCommand } from './commands/adr-cmd.js';
 import { lintCommand } from './commands/lint-cmd.js';
 import { fleetingCommand } from './commands/fleeting-cmd.js';
 import { ingestCommand, promoteCommand } from './commands/ingest-cmd.js';
@@ -134,6 +135,7 @@ program
     .command('ask <question>')
     .description('Ask a question about your knowledge base — search, compose answer, optionally save')
     .option('-s, --save', 'Save answer as a new note in your vault')
+    .option('-q, --quotes', 'Show direct quotes from sources (Insight Extraction mode)')
     .action((question, opts) => askCommand(question, opts));
 program
     .command('compile')
@@ -144,9 +146,10 @@ program
     .action((opts) => compileCommand(opts));
 program
     .command('draft [topic]')
-    .description('Express: Generate a blog post, report, or outline draft from your knowledge')
-    .option('--format <type>', 'Output format: blog, report, outline (default: blog)')
+    .description('Express: Generate draft from knowledge (blog/report/outline/instagram/thread/script)')
+    .option('--format <type>', 'Output format: blog, report, outline, instagram, thread, script')
     .option('--ai', 'Use Claude API for AI-enhanced draft (requires ANTHROPIC_API_KEY)')
+    .option('--blueprint <spec>', 'Chapter structure: "Ch1:tag1,tag2; Ch2:tag3"')
     .action((topic, opts) => draftCommand(topic, opts));
 program
     .command('session-save')
@@ -160,6 +163,14 @@ program
     .command('flush')
     .description('Flush daily logs → wiki: extract concepts, rebuild connections (Karpathy compile)')
     .action(() => flushCommand());
+program
+    .command('adr <title>')
+    .description('Create an Architecture Decision Record (structured decision log)')
+    .option('--context <text>', 'Why is this decision needed?')
+    .option('--options <text>', 'What alternatives were considered?')
+    .option('--decision <text>', 'What was decided and why?')
+    .option('--consequences <text>', 'What are the implications?')
+    .action((title, opts) => adrCommand(title, opts));
 program
     .command('lint')
     .description('Knowledge health check — find gaps, duplicates, contradictions, stale notes')

package/packages/core/dist/api/graph-data.js

@@ -130,7 +130,7 @@ export async function buildGraphData(store, options = {}) {
         const size = 1 + 6 * Math.pow(ratio, 0.5);
         return {
             id: doc.id,
-            label: doc.title.slice(0, 40),
+            label: doc.title.replace(/&#39;/g, "'").replace(/&#x27;/g, "'").replace(/&quot;/g, '"').replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/g, '>').replace(/\s*-\s*YouTube$/, '').slice(0, 40),
             filePath: doc.filePath,
             tags: doc.tags,
             clusterId: assignmentMap.get(doc.id) ?? 0,

package/packages/core/dist/api/server.js

@@ -41,22 +41,32 @@ export function createApiServer(options) {
             res.status(500).json({ error: 'Internal server error' });
         }
     });
+    // GET /api/reindex/status — 인덱싱 진행률 조회
+    let reindexProgress = { active: false, current: 0, total: 0, phase: '' };
+    app.get('/api/reindex/status', (_req, res) => {
+        res.json(reindexProgress);
+    });
     // POST /api/reindex — 웹에서 인덱싱 트리거
     let isReindexing = false;
     app.post('/api/reindex', async (_req, res) => {
         if (isReindexing) {
-            res.json({ success: false, error: '인덱싱이 이미 진행 중입니다' });
+            res.json({ success: false, error: 'Reindexing already in progress', progress: reindexProgress });
             return;
         }
         isReindexing = true;
+        reindexProgress = { active: true, current: 0, total: 0, phase: 'initializing' };
         try {
             const indexer = await import('../indexer/index.js');
+            reindexProgress.phase = 'loading embedder';
             const embedder = indexer.createLocalEmbedder('all-MiniLM-L6-v2');
             await embedder.initialize();
+            reindexProgress.phase = 'indexing';
             const result = await indexer.indexVault(vaultPath, {
                 store,
                 embedder,
                 onProgress: (current, total) => {
+                    reindexProgress.current = current;
+                    reindexProgress.total = total;
                     if (current % 50 === 0)
                         console.error(`[reindex] ${current}/${total}`);
                 },
@@ -76,6 +86,7 @@ export function createApiServer(options) {
         }
         finally {
             isReindexing = false;
+            reindexProgress = { active: false, current: 0, total: 0, phase: 'done' };
         }
     });
     // GET /api/search?q=&limit=

package/packages/core/dist/indexer/scanner.js

@@ -46,9 +46,20 @@ function parseDocument(vaultPath, filePath) {
         ?? extractFirstHeading(content)
         ?? relativePath.replace(/\.md$/, '');
     const tags = extractTags(frontmatter, content);
+    // Obsidian frontmatter 호환: aliases를 tags에 추가
+    if (frontmatter.aliases) {
+        const aliases = Array.isArray(frontmatter.aliases) ? frontmatter.aliases : [frontmatter.aliases];
+        for (const alias of aliases) {
+            if (typeof alias === 'string' && alias.length > 1)
+                tags.push(alias);
+        }
+    }
     // source/type 자동 추출 (원본 파일 수정 없이 DB에만 저장)
     const source = inferSource(frontmatter, relativePath);
     const type = inferType(frontmatter, relativePath);
+    // Obsidian frontmatter 호환: date/created 우선 사용
+    const fmDate = frontmatter.date ?? frontmatter.created ?? frontmatter.created_at;
+    const lastModified = fmDate ? new Date(fmDate).toISOString() : stat.mtime.toISOString();
     return {
         id,
         filePath: relativePath,
@@ -56,7 +67,7 @@ function parseDocument(vaultPath, filePath) {
         content,
         frontmatter,
         tags,
-        lastModified: stat.mtime.toISOString(),
+        lastModified,
         contentHash,
         source,
         type,

package/packages/core/dist/intelligence/ask-engine.d.ts

@@ -19,5 +19,6 @@ export declare function askVault(searchEngine: SearchEngine, question: string, o
     save?: boolean;
     vaultPath?: string;
     outputDir?: string;
+    mode?: 'default' | 'quotes';
 }): Promise<AskResult>;
 //# sourceMappingURL=ask-engine.d.ts.map

package/packages/core/dist/intelligence/ask-engine.js

@@ -7,7 +7,7 @@ import { join, resolve } from 'node:path';
  * LLM 없이 검색 결과를 구조화하는 버전 (LLM 연동은 MCP ask tool에서 처리).
  */
 export async function askVault(searchEngine, question, options = {}) {
-    const { limit = 10, save = false, vaultPath, outputDir = '_stellavault/answers' } = options;
+    const { limit = 10, save = false, vaultPath, outputDir = '_stellavault/answers', mode = 'default' } = options;
     // 1. 검색
     const results = await searchEngine.search({ query: question, limit });
     // 2. 소스 정리
@@ -18,7 +18,9 @@ export async function askVault(searchEngine, question, options = {}) {
         snippet: r.chunk?.content?.substring(0, 200) ?? '',
     }));
     // 3. 답변 구성 (검색 결과 기반 구조화)
-    const answer = composeAnswer(question, results);
+    const answer = mode === 'quotes'
+        ? composeQuotes(question, results)
+        : composeAnswer(question, results);
     // 4. vault에 저장 (선택)
     let savedTo = null;
     if (save && vaultPath) {
@@ -66,6 +68,38 @@ function composeAnswer(question, results) {
     lines.push(`- Find knowledge gaps: \`stellavault gaps\``);
     return lines.join('\n');
 }
+/**
+ * Quotes 모드: 원문 인용을 카드 형태로 나열.
+ */
+function composeQuotes(question, results) {
+    if (results.length === 0) {
+        return `No quotes found for "${question}".`;
+    }
+    const lines = [];
+    lines.push(`## Quotes: "${question}"\n`);
+    lines.push(`*${results.length} sources found. Each quote is a direct excerpt.*\n`);
+    for (let i = 0; i < Math.min(8, results.length); i++) {
+        const r = results[i];
+        const quote = r.chunk?.content
+            ?.replace(/^---[\s\S]*?---\n?/, '')
+            ?.replace(/^#+\s+.+\n/m, '')
+            ?.trim()
+            ?.substring(0, 300) ?? '';
+        if (!quote)
+            continue;
+        const score = Math.round(r.score * 100);
+        lines.push(`---`);
+        lines.push(`**[${i + 1}] ${r.document.title}** (${score}% match)`);
+        lines.push(`> ${quote.replace(/\n/g, '\n> ')}`);
+        if (r.document.tags.length > 0) {
+            lines.push(`Tags: ${r.document.tags.slice(0, 5).map(t => `#${t}`).join(' ')}`);
+        }
+        lines.push('');
+    }
+    lines.push('---');
+    lines.push(`*Use \`stellavault ask "${question}" --save\` to save these quotes as a note.*`);
+    return lines.join('\n');
+}
 /**
  * 답변을 vault에 .md 파일로 저장.
  * "_stellavault/answers/YYYY-MM-DD-question.md" 형식.

package/packages/core/dist/intelligence/auto-linker.js

@@ -95,10 +95,14 @@ export function insertWikilinks(body, vaultTitles, selfTitle) {
     // 긴 구문부터 매칭 (짧은 구문이 긴 구문의 부분 매칭되는 것 방지)
     const sortedPhrases = [...phraseToTitle.keys()].sort((a, b) => b.length - a.length);
     const linkedPhrases = new Set(); // 이미 링크된 구문 (중복 방지)
+    const contentLower = content.toLowerCase();
     for (const phrase of sortedPhrases) {
         const targetTitle = phraseToTitle.get(phrase);
         if (linkedPhrases.has(targetTitle))
-            continue; // 같은 노트에 여러 구문 매칭 방지
+            continue;
+        // 사전 필터: 본문에 포함되지 않으면 regex 생성 스킵 (성능 최적화)
+        if (!contentLower.includes(phrase.toLowerCase()))
+            continue;
         const escaped = phrase.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
         const regex = new RegExp(`(?<!\\[\\[)(?<!#\\s)\\b(${escaped})\\b(?!\\]\\])(?![^\\[]*\\]\\])`, 'gi');
         let replaced = false;

package/packages/core/dist/intelligence/draft-generator.d.ts

@@ -1,8 +1,14 @@
 import { type FolderNames } from '../config.js';
+export interface BlueprintSection {
+    title: string;
+    tags?: string[];
+    description?: string;
+}
 export interface DraftOptions {
     topic?: string;
-    format?: 'blog' | 'report' | 'outline';
+    format?: 'blog' | 'report' | 'outline' | 'instagram' | 'thread' | 'script';
     maxSections?: number;
+    blueprint?: BlueprintSection[];
 }
 export interface DraftResult {
     title: string;

package/packages/core/dist/intelligence/draft-generator.js

@@ -10,7 +10,7 @@ import { DEFAULT_FOLDERS } from '../config.js';
  * Express 단계: 지식이 vault에서 나가는 출구.
  */
 export function generateDraft(vaultPath, options = {}, folders = DEFAULT_FOLDERS) {
-    const { topic, format = 'blog', maxSections = 8 } = options;
+    const { topic, format = 'blog', maxSections = 8, blueprint } = options;
     // raw + wiki 문서 스캔
     const rawDir = resolve(vaultPath, folders.fleeting);
     const wikiDir = resolve(vaultPath, folders.wiki);
@@ -43,18 +43,53 @@ export function generateDraft(vaultPath, options = {}, folders = DEFAULT_FOLDERS
         ? `Draft: ${topic}`
         : `Knowledge Draft — ${new Date().toISOString().split('T')[0]}`;
     let body;
-    switch (format) {
-        case 'outline':
-            body = generateOutline(draftTitle, topConcepts, filteredDocs);
-            break;
-        case 'report':
-            body = generateReport(draftTitle, topConcepts, filteredDocs);
-            break;
-        case 'blog':
-        default:
-            body = generateBlog(draftTitle, topConcepts, filteredDocs);
-            break;
+    // Blueprint 모드: 사용자 정의 챕터 구성
+    if (blueprint && blueprint.length > 0) {
+        const bpLines = [`# ${draftTitle}\n`];
+        bpLines.push(`> Blueprint draft from ${filteredDocs.length} sources\n`);
+        for (const section of blueprint) {
+            bpLines.push(`## ${section.title}\n`);
+            if (section.description)
+                bpLines.push(`*${section.description}*\n`);
+            // 태그 필터로 관련 문서 찾기
+            const sectionDocs = section.tags
+                ? filteredDocs.filter(d => d.tags.some(t => section.tags.includes(t)) || d.title.toLowerCase().includes(section.title.toLowerCase()))
+                : filteredDocs;
+            const usedDocs = sectionDocs.slice(0, 5);
+            for (const doc of usedDocs) {
+                const excerpt = extractExcerpt(doc.content, 200);
+                if (excerpt)
+                    bpLines.push(`> ${excerpt}\n> — *${doc.title}*\n`);
+            }
+            if (usedDocs.length > 0) {
+                bpLines.push(`*Sources: ${usedDocs.map(d => d.title).join(', ')}*\n`);
+            }
+        }
+        bpLines.push(`---\n*Generated by \`stellavault draft --blueprint\` at ${new Date().toISOString()}*`);
+        body = bpLines.join('\n');
     }
+    else
+        switch (format) {
+            case 'outline':
+                body = generateOutline(draftTitle, topConcepts, filteredDocs);
+                break;
+            case 'report':
+                body = generateReport(draftTitle, topConcepts, filteredDocs);
+                break;
+            case 'instagram':
+                body = generateInstagram(draftTitle, topConcepts, filteredDocs);
+                break;
+            case 'thread':
+                body = generateThread(draftTitle, topConcepts, filteredDocs);
+                break;
+            case 'script':
+                body = generateScript(draftTitle, topConcepts, filteredDocs);
+                break;
+            case 'blog':
+            default:
+                body = generateBlog(draftTitle, topConcepts, filteredDocs);
+                break;
+        }
     const wordCount = body.split(/\s+/).filter(Boolean).length;
     // _drafts/ 폴더에 저장
     const draftsDir = resolve(vaultPath, '_drafts');
@@ -97,11 +132,23 @@ function generateBlog(title, concepts, docs) {
         }
         lines.push('<!-- TODO: Add your analysis and insights here -->', '');
     }
-    // 참고 자료
+    // 참고 자료 (APA-style citations)
     lines.push('## References', '');
     const uniqueDocs = [...new Map(docs.map(d => [d.filePath, d])).values()].slice(0, 20);
-    for (const doc of uniqueDocs) {
-        lines.push(`- [[${basename(doc.filePath, extname(doc.filePath))}|${doc.title}]]`);
+    const now = new Date();
+    for (let i = 0; i < uniqueDocs.length; i++) {
+        const doc = uniqueDocs[i];
+        const dateMatch = doc.content.match(/(?:created|date):\s*(\d{4}-\d{2}-\d{2})/);
+        const year = dateMatch ? dateMatch[1].split('-')[0] : now.getFullYear().toString();
+        const sourceMatch = doc.content.match(/source:\s*(.+)/);
+        const source = sourceMatch ? sourceMatch[1].trim() : '';
+        // APA-style: Author/Title (Year). Source.
+        if (source.startsWith('http')) {
+            lines.push(`[${i + 1}] ${doc.title} (${year}). Retrieved from ${source}`);
+        }
+        else {
+            lines.push(`[${i + 1}] ${doc.title} (${year}). [[${basename(doc.filePath, extname(doc.filePath))}|Link]]`);
+        }
     }
     lines.push('');
     lines.push(`---`, `*Generated by \`stellavault draft\` at ${new Date().toISOString()}*`);
@@ -146,6 +193,75 @@ function generateOutline(title, concepts, docs) {
     lines.push('', `---`, `*Generated by \`stellavault draft\` at ${new Date().toISOString()}*`);
     return lines.join('\n');
 }
+// ─── SNS 포맷 ───
+function generateInstagram(title, concepts, docs) {
+    const lines = [];
+    const topicWords = concepts.slice(0, 3).map(([c]) => c);
+    lines.push(`${title}`, '');
+    lines.push('---', '');
+    // 슬라이드 1: Hook
+    lines.push('**Slide 1 (Hook)**', '');
+    lines.push(`Did you know about ${topicWords.join(', ')}?`, '');
+    // 슬라이드 2-4: Key Points
+    for (let i = 0; i < Math.min(3, concepts.length); i++) {
+        const [concept] = concepts[i];
+        const doc = docs.find(d => d.tags.includes(concept) || d.title.toLowerCase().includes(concept.toLowerCase()));
+        const excerpt = doc ? extractExcerpt(doc.content, 80) : '';
+        lines.push(`**Slide ${i + 2} (${capitalize(concept)})**`, '');
+        lines.push(excerpt || `Key insight about ${concept}`, '');
+    }
+    // CTA
+    lines.push('**Last Slide (CTA)**', '');
+    lines.push('Save this for later! Follow for more knowledge drops.', '');
+    // Caption
+    lines.push('---', '', '**Caption:**', '');
+    lines.push(`${topicWords.map(w => `#${w}`).join(' ')} #knowledge #secondbrain #stellavault`, '');
+    lines.push(`---`, `*Generated by \`stellavault draft --format instagram\`*`);
+    return lines.join('\n');
+}
+function generateThread(title, concepts, docs) {
+    const lines = [];
+    // Tweet 1: Hook
+    lines.push(`**1/** ${title}`, '');
+    lines.push(`Here's what I learned from ${docs.length} sources:`, '');
+    // Tweet 2-N: One per concept
+    for (let i = 0; i < Math.min(6, concepts.length); i++) {
+        const [concept, docPaths] = concepts[i];
+        const doc = docs.find(d => docPaths.includes(d.filePath));
+        const excerpt = doc ? extractExcerpt(doc.content, 200) : '';
+        lines.push(`**${i + 2}/** ${capitalize(concept)}`, '');
+        lines.push(excerpt || `Key point about ${concept}`, '');
+    }
+    // Final tweet
+    lines.push(`**${Math.min(6, concepts.length) + 2}/** That's a wrap!`, '');
+    lines.push('If this was useful, repost the first tweet.', '');
+    lines.push(`---`, `*Generated by \`stellavault draft --format thread\`*`);
+    return lines.join('\n');
+}
+function generateScript(title, concepts, docs) {
+    const lines = [];
+    lines.push(`# ${title} — Video Script`, '');
+    lines.push(`**Duration:** ~${Math.max(3, concepts.length * 2)} minutes`, '');
+    // Intro
+    lines.push('## Intro (30s)', '');
+    lines.push(`"Hey everyone, today we're diving into ${concepts.slice(0, 2).map(([c]) => c).join(' and ')}."`, '');
+    // Sections
+    for (let i = 0; i < Math.min(5, concepts.length); i++) {
+        const [concept, docPaths] = concepts[i];
+        const relatedDocs = docs.filter(d => docPaths.includes(d.filePath)).slice(0, 2);
+        lines.push(`## Section ${i + 1}: ${capitalize(concept)} (~${Math.max(1, Math.floor(relatedDocs.length))}min)`, '');
+        for (const doc of relatedDocs) {
+            const excerpt = extractExcerpt(doc.content, 150);
+            if (excerpt)
+                lines.push(`[Visual: ${doc.title}]`, excerpt, '');
+        }
+    }
+    // Outro
+    lines.push('## Outro (15s)', '');
+    lines.push('"If you found this helpful, like and subscribe!"', '');
+    lines.push(`---`, `*Generated by \`stellavault draft --format script\`*`);
+    return lines.join('\n');
+}
 // ─── 유틸 ───
 function extractExcerpt(content, maxLen) {
     // frontmatter 제거

package/packages/core/dist/intelligence/file-extractors.d.ts

@@ -9,10 +9,5 @@ export interface ExtractedContent {
     sourceFormat: 'pdf' | 'docx' | 'pptx' | 'xlsx' | 'xls' | 'text';
 }
 export declare function isBinaryFormat(filePath: string): boolean;
-/**
- * 파일 경로에서 텍스트 추출. 확장자 기반 파서 디스패치.
- * 지원: .pdf, .docx, .pptx, .xlsx, .xls
- * 미지원 확장자: utf-8 텍스트로 읽기
- */
 export declare function extractFileContent(filePath: string): Promise<ExtractedContent>;
 //# sourceMappingURL=file-extractors.d.ts.map

package/packages/core/dist/intelligence/file-extractors.js

@@ -1,6 +1,6 @@
 // Design Ref: §file-ingest-v2 — Binary file text extraction dispatchers
 // Plan SC: SC1-SC5 (format-specific extraction + fallback)
-import { readFileSync } from 'node:fs';
+import { readFileSync, statSync } from 'node:fs';
 import { extname, basename } from 'node:path';
 const BINARY_EXTS = new Set(['.pdf', '.docx', '.pptx', '.xlsx', '.xls']);
 export function isBinaryFormat(filePath) {
@@ -11,8 +11,13 @@ export function isBinaryFormat(filePath) {
  * 지원: .pdf, .docx, .pptx, .xlsx, .xls
  * 미지원 확장자: utf-8 텍스트로 읽기
  */
+const MAX_FILE_SIZE = 50 * 1024 * 1024; // 50MB
 export async function extractFileContent(filePath) {
     const ext = extname(filePath).toLowerCase();
+    const { size } = statSync(filePath);
+    if (size > MAX_FILE_SIZE) {
+        throw new Error(`File too large (${Math.round(size / 1024 / 1024)}MB > 50MB limit)`);
+    }
     const buffer = readFileSync(filePath);
     switch (ext) {
         case '.pdf': return extractPdf(buffer, filePath);
@@ -86,12 +91,35 @@ async function extractXlsx(buffer, filePath) {
     try {
         const XLSX = await import('xlsx');
         const workbook = XLSX.read(buffer);
-        const text = workbook.SheetNames
-            .map((name) => {
-            const csv = XLSX.utils.sheet_to_csv(workbook.Sheets[name]);
-            return `## ${name}\n\n${csv}`;
-        })
-            .join('\n\n');
+        const parts = [];
+        for (const name of workbook.SheetNames) {
+            const sheet = workbook.Sheets[name];
+            // 마크다운 테이블 형식 (헤더-값 구조 보존)
+            const rows = XLSX.utils.sheet_to_json(sheet, { header: 1 });
+            if (rows.length === 0)
+                continue;
+            parts.push(`## ${name}\n`);
+            const headers = rows[0].map((h) => String(h ?? ''));
+            parts.push(`| ${headers.join(' | ')} |`);
+            parts.push(`| ${headers.map(() => '---').join(' | ')} |`);
+            for (const row of rows.slice(1)) {
+                const cells = headers.map((_, i) => String(row[i] ?? ''));
+                parts.push(`| ${cells.join(' | ')} |`);
+            }
+            parts.push('');
+            // JSON 구조도 포함 (AI 검색/ask에서 수치 활용 가능)
+            if (rows.length <= 100) {
+                const jsonRows = XLSX.utils.sheet_to_json(sheet);
+                if (jsonRows.length > 0) {
+                    parts.push(`<details><summary>Structured Data (${jsonRows.length} rows)</summary>\n`);
+                    parts.push('```json');
+                    parts.push(JSON.stringify(jsonRows.slice(0, 50), null, 2));
+                    parts.push('```');
+                    parts.push('</details>\n');
+                }
+            }
+        }
+        const text = parts.join('\n');
         return {
             text,
             metadata: {

package/packages/core/dist/intelligence/ingest-pipeline.js

@@ -9,16 +9,24 @@ import { scanFrontmatter, assignIndexCodes, archiveFile } from './zettelkasten.j
 import { compileWiki } from './wiki-compiler.js';
 import { autoLink } from './auto-linker.js';
 import { DEFAULT_FOLDERS } from '../config.js';
+/** HTML 엔티티 디코딩 */
+function decodeHtmlEntities(text) {
+    return text
+        .replace(/'/g, "'").replace(/'/g, "'")
+        .replace(/"/g, '"').replace(/&/g, '&')
+        .replace(/&lt;/g, '<').replace(/&gt;/g, '>')
+        .replace(/&nbsp;/g, ' ');
+}
 /** YAML 값에서 위험한 문자를 이스케이프 */
 function sanitizeYaml(val) {
-    return val.replace(/["\\]/g, '\\$&').replace(/\n/g, ' ').slice(0, 200);
+    return decodeHtmlEntities(val).replace(/["\\]/g, '\\$&').replace(/\n/g, ' ').slice(0, 200);
 }
 /**
  * 어떤 입력이든 Stellavault 표준 포맷으로 변환하여 저장.
  */
 export function ingest(vaultPath, input, folders = DEFAULT_FOLDERS) {
     const stage = input.stage ?? 'fleeting';
-    const title = input.title ?? extractTitleFromContent(input.content, input.type);
+    const title = decodeHtmlEntities(input.title ?? extractTitleFromContent(input.content, input.type));
     const tags = input.tags ?? extractAutoTags(input.content, input.type);
     const source = input.source ?? (input.type === 'url' || input.type === 'youtube' ? input.content.split('\n')[0] : 'manual');
     // 본문 정리
@@ -57,7 +65,9 @@ export function ingest(vaultPath, input, folders = DEFAULT_FOLDERS) {
             }]);
         indexCode = assignments.get(filePath);
     }
-    catch { /* index code is optional */ }
+    catch (err) {
+        console.warn('[ingest] Index code skipped:', err instanceof Error ? err.message : err);
+    }
     // Stellavault 표준 포맷으로 저장
     let md = buildStandardNote({
         title,
@@ -73,7 +83,9 @@ export function ingest(vaultPath, input, folders = DEFAULT_FOLDERS) {
     try {
         md = autoLink(md, vaultPath, title, folders);
     }
-    catch { /* autoLink 실패해도 저장은 진행 */ }
+    catch (err) {
+        console.warn('[ingest] Auto-link skipped:', err instanceof Error ? err.message : err);
+    }
     writeFileSync(fullPath, md, 'utf-8');
     // 자동 compile: fleeting → wiki (rule-based, <100ms)
     try {
@@ -166,7 +178,25 @@ function extractAutoTags(content, type) {
     // 인라인 #태그 추출
     const inline = content.match(/#([a-zA-Z가-힣][a-zA-Z0-9가-힣_-]{2,})/g) ?? [];
     inline.forEach(t => tags.add(t.slice(1)));
-    return [...tags].slice(0, 10);
+    // 스마트 자동 태깅: 문서 내용 분석 → 카테고리 분류
+    const lc = content.toLowerCase();
+    if (/회의|meeting|minutes|참석자|agenda/.test(lc))
+        tags.add('meeting-notes');
+    if (/기획|prd|요구사항|spec|feature|유저\s*스토리/.test(lc))
+        tags.add('planning');
+    if (/api|endpoint|서버|backend|database|쿼리/.test(lc))
+        tags.add('technical');
+    if (/디자인|design|ui|ux|figma|wireframe|mockup/.test(lc))
+        tags.add('design');
+    if (/논문|paper|abstract|methodology|conclusion|참고문헌/.test(lc))
+        tags.add('research');
+    if (/tutorial|강의|강좌|배우|learn|course/.test(lc))
+        tags.add('learning');
+    if (/경쟁|competitor|시장|market|swot|분석/.test(lc))
+        tags.add('analysis');
+    if (/일기|diary|journal|오늘|today|daily/.test(lc))
+        tags.add('journal');
+    return [...tags].slice(0, 15);
 }
 function cleanContent(content, type) {
     if (type === 'url' || type === 'youtube') {
@@ -195,10 +225,10 @@ function buildStandardNote(params) {
         '---',
         `title: "${sanitizeYaml(params.title)}"`,
         `type: ${params.stage}`,
-        `source: ${params.source}`,
+        `source: "${sanitizeYaml(params.source)}"`,
         `input_type: ${params.inputType}`,
         params.indexCode ? `zettel_id: "${params.indexCode}"` : null,
-        `tags: [${params.tags.map(t => `"${t}"`).join(', ')}]`,
+        `tags: [${params.tags.map(t => `"${sanitizeYaml(t)}"`).join(', ')}]`,
         `created: ${params.created}`,
         `summary: "${sanitizeYaml(params.body.slice(0, 100))}"`,
         '---',

package/packages/core/dist/intelligence/youtube-extractor.js

@@ -49,7 +49,17 @@ export function formatYouTubeNote(content) {
         lines.push(`> ${nt('views')}: ${Number(content.viewCount).toLocaleString()}`);
     lines.push(`> ${content.url}`, '');
     if (content.summary) {
-        lines.push(`## ${nt('summary')}`, '', content.summary, '');
+        // 핵심 포인트를 bullet list로 포맷
+        const points = content.summary.split(/[.。!?]\s+/).filter(s => s.length > 15).slice(0, 5);
+        if (points.length > 1) {
+            lines.push(`## Key Points`, '');
+            for (const p of points)
+                lines.push(`- ${p.trim()}`);
+            lines.push('');
+        }
+        else {
+            lines.push(`## ${nt('summary')}`, '', content.summary, '');
+        }
     }
     if (content.description.length > 30) {
         lines.push(`## ${nt('description')}`, '', content.description.slice(0, 3000), '');

package/packages/core/dist/mcp/tools/ask.js

@@ -24,12 +24,15 @@ export function createAskTool(searchEngine, vaultPath) {
                 save: args.save ?? false,
                 vaultPath,
             });
+            const sourceList = result.sources.slice(0, 5).map((s, i) => `${i + 1}. **${s.title}** (score: ${s.score})\n   > ${s.snippet.slice(0, 120)}...`).join('\n');
             const text = [
                 result.answer,
                 '',
-                result.savedTo ? `Saved to: ${result.savedTo}` : '',
+                '---',
+                `### Sources (${result.sources.length} documents)`,
+                sourceList,
                 '',
-                `Sources: ${result.sources.length} documents found`,
+                result.savedTo ? `Saved to: ${result.savedTo}` : '',
             ].filter(Boolean).join('\n');
             return {
                 content: [{ type: 'text', text }],