moflo 4.8.32 → 4.8.34

package/bin/index-guidance.mjs

@@ -1,906 +1,906 @@
 #!/usr/bin/env node
-/**
- * Index guidance files into claude-flow memory with full RAG linked segments
- *
- * Strategy:
- * - Full documents stored as `doc-{name}` for complete retrieval
- * - Semantic chunks stored as `chunk-{name}-{n}` for precise search
- * - FULL RAG LINKING:
- *   - parentDoc: link to full document
- *   - prevChunk/nextChunk: forward/backward navigation
- *   - siblings: all chunk keys from same document
- *   - children: sub-chunks for hierarchical headers (h2 -> h3)
- *   - contextBefore/contextAfter: overlapping text for context continuity
- * - Chunking based on markdown headers (## and ###) for natural boundaries
- * - After indexing, generates embeddings for semantic search (HNSW)
- *
- * Usage:
- *   node node_modules/moflo/bin/index-guidance.mjs                 # Index all + generate embeddings
- *   npx flo-index --force                                        # Force reindex all
- *   npx flo-index --file X                                       # Index specific file
- *   npx flo-index --no-embeddings                                # Skip embedding generation
- *   npx flo-index --overlap 20                                   # Set context overlap % (default: 15)
- */
-
-import { existsSync, readdirSync, readFileSync, statSync, mkdirSync, writeFileSync } from 'fs';
-import { resolve, dirname, basename, extname } from 'path';
-import { fileURLToPath } from 'url';
-import { mofloResolveURL } from './lib/moflo-resolve.mjs';
-const initSqlJs = (await import(mofloResolveURL('sql.js'))).default;
-
-
-const __dirname = dirname(fileURLToPath(import.meta.url));
-
-function findProjectRoot() {
-  let dir = process.cwd();
-  const root = resolve(dir, '/');
-  while (dir !== root) {
-    if (existsSync(resolve(dir, 'package.json'))) return dir;
-    dir = dirname(dir);
-  }
-  return process.cwd();
-}
-
-const projectRoot = findProjectRoot();
-
-// Locate the moflo package root (for bundled guidance that ships with moflo)
-const mofloRoot = resolve(__dirname, '..');
-
-const NAMESPACE = 'guidance';
-const DB_PATH = resolve(projectRoot, '.swarm/memory.db');
-
-// ============================================================================
-// Load guidance directories from moflo.yaml, falling back to defaults
-// ============================================================================
-
-function loadGuidanceDirs() {
-  const dirs = [];
-
-  // 1. Read moflo.yaml / moflo.config.json for user-configured directories
-  let configDirs = null;
-  const yamlPath = resolve(projectRoot, 'moflo.yaml');
-  const jsonPath = resolve(projectRoot, 'moflo.config.json');
-
-  if (existsSync(yamlPath)) {
-    try {
-      const content = readFileSync(yamlPath, 'utf-8');
-      // Simple YAML array extraction — avoids needing js-yaml at runtime
-      // Matches:  guidance:\n    directories:\n      - .claude/guidance\n      - docs/guides
-      const guidanceBlock = content.match(/guidance:\s*\n\s+directories:\s*\n((?:\s+-\s+.+\n?)+)/);
-      if (guidanceBlock) {
-        const items = guidanceBlock[1].match(/-\s+(.+)/g);
-        if (items && items.length > 0) {
-          configDirs = items.map(item => item.replace(/^-\s+/, '').trim());
-        }
-      }
-    } catch { /* ignore parse errors, fall through to defaults */ }
-  } else if (existsSync(jsonPath)) {
-    try {
-      const raw = JSON.parse(readFileSync(jsonPath, 'utf-8'));
-      if (raw.guidance?.directories && Array.isArray(raw.guidance.directories)) {
-        configDirs = raw.guidance.directories;
-      }
-    } catch { /* ignore parse errors */ }
-  }
-
-  // Use config dirs or fall back to defaults
-  // Each directory gets a unique prefix derived from its path to avoid key collisions
-  // when multiple directories contain files with the same name.
-  const userDirs = configDirs || ['.claude/guidance', 'docs/guides'];
-  for (const d of userDirs) {
-    const prefix = d.replace(/\\/g, '/')
-      .replace(/^\.claude\//, '')
-      .replace(/^back-office\/api\/\.claude\//, 'bo-api-')
-      .replace(/^back-office\/ui\/\.claude\//, 'bo-ui-')
-      .replace(/[^a-zA-Z0-9-]/g, '-')
-      .replace(/-+/g, '-')
-      .replace(/^-|-$/g, '') || 'guidance';
-    dirs.push({ path: d, prefix });
-  }
-
-  // 2. Include moflo's own bundled guidance (ships with the package)
-  //    Only when running inside a consumer project (not moflo itself)
-  //    Shipped guidance lives in .claude/guidance/shipped/ — internal/ is excluded from npm
-  const bundledShippedDir = resolve(mofloRoot, '.claude/guidance/shipped');
-  const bundledGuidanceDir = existsSync(bundledShippedDir)
-    ? bundledShippedDir
-    : resolve(mofloRoot, '.claude/guidance');
-  const projectGuidanceDir = resolve(projectRoot, '.claude/guidance');
-  if (
-    existsSync(bundledGuidanceDir) &&
-    resolve(bundledGuidanceDir) !== resolve(projectGuidanceDir) &&
-    resolve(bundledGuidanceDir) !== resolve(projectGuidanceDir, 'shipped')
-  ) {
-    dirs.push({ path: bundledGuidanceDir, prefix: 'moflo-bundled', absolute: true });
-  }
-
-  // 3. CLAUDE.md files are NOT indexed — Claude loads them into context automatically.
-  //    Indexing them wastes vectors and creates duplicate keys across subprojects.
-
-  return dirs;
-}
-
-const GUIDANCE_DIRS = loadGuidanceDirs();
-
-// Chunking config - optimized for Claude's retrieval
-const MIN_CHUNK_SIZE = 50;    // Lower minimum to avoid mega-chunks
-const MAX_CHUNK_SIZE = 4000;  // Larger chunks for code-heavy docs (fits context better)
-const FORCE_CHUNK_THRESHOLD = 6000; // Force paragraph-split if file > this and < 3 chunks
-const DEFAULT_OVERLAP_PERCENT = 20; // Increased context overlap for better continuity
-
-// Parse args
-const args = process.argv.slice(2);
-const force = args.includes('--force');
-const specificFile = args.includes('--file') ? args[args.indexOf('--file') + 1] : null;
-const verbose = args.includes('--verbose') || args.includes('-v');
-const skipEmbeddings = args.includes('--no-embeddings');
-const overlapPercent = args.includes('--overlap')
-  ? parseInt(args[args.indexOf('--overlap') + 1], 10) || DEFAULT_OVERLAP_PERCENT
-  : DEFAULT_OVERLAP_PERCENT;
-
-function log(msg) {
-  console.log(`[index-guidance] ${msg}`);
-}
-
-function debug(msg) {
-  if (verbose) console.log(`[index-guidance]   ${msg}`);
-}
-
-function ensureDbDir() {
-  const dir = dirname(DB_PATH);
-  if (!existsSync(dir)) {
-    mkdirSync(dir, { recursive: true });
-  }
-}
-
-async function getDb() {
-  ensureDbDir();
-  const SQL = await initSqlJs();
-  let db;
-  if (existsSync(DB_PATH)) {
-    const buffer = readFileSync(DB_PATH);
-    db = new SQL.Database(buffer);
-  } else {
-    db = new SQL.Database();
-  }
-
-  // Ensure table exists with unique constraint
-  db.run(`
-    CREATE TABLE IF NOT EXISTS memory_entries (
-      id TEXT PRIMARY KEY,
-      key TEXT NOT NULL,
-      namespace TEXT DEFAULT 'default',
-      content TEXT NOT NULL,
-      type TEXT DEFAULT 'semantic',
-      embedding TEXT,
-      embedding_model TEXT DEFAULT 'local',
-      embedding_dimensions INTEGER,
-      tags TEXT,
-      metadata TEXT,
-      owner_id TEXT,
-      created_at INTEGER NOT NULL DEFAULT (strftime('%s', 'now') * 1000),
-      updated_at INTEGER NOT NULL DEFAULT (strftime('%s', 'now') * 1000),
-      expires_at INTEGER,
-      last_accessed_at INTEGER,
-      access_count INTEGER DEFAULT 0,
-      status TEXT DEFAULT 'active',
-      UNIQUE(namespace, key)
-    )
-  `);
-
-  db.run(`CREATE INDEX IF NOT EXISTS idx_memory_key_ns ON memory_entries(key, namespace)`);
-  db.run(`CREATE INDEX IF NOT EXISTS idx_memory_namespace ON memory_entries(namespace)`);
-
-  return db;
-}
-
-function saveDb(db) {
-  const data = db.export();
-  writeFileSync(DB_PATH, Buffer.from(data));
-}
-
-function generateId() {
-  return `mem_${Date.now()}_${Math.random().toString(36).substr(2, 9)}`;
-}
-
-function hashContent(content) {
-  let hash = 0;
-  for (let i = 0; i < content.length; i++) {
-    const char = content.charCodeAt(i);
-    hash = ((hash << 5) - hash) + char;
-    hash = hash & hash;
-  }
-  return hash.toString(16);
-}
-
-function storeEntry(db, key, content, metadata = {}, tags = []) {
-  const now = Date.now();
-  const id = generateId();
-  const metaJson = JSON.stringify(metadata);
-  const tagsJson = JSON.stringify(tags);
-
-  db.run(`
-    INSERT OR REPLACE INTO memory_entries
-    (id, key, namespace, content, metadata, tags, created_at, updated_at, status)
-    VALUES (?, ?, ?, ?, ?, ?, ?, ?, 'active')
-  `, [id, key, NAMESPACE, content, metaJson, tagsJson, now, now]);
-
-  return true;
-}
-
-function deleteByPrefix(db, prefix) {
-  db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key LIKE ?`, [NAMESPACE, `${prefix}%`]);
-}
-
-function getEntryHash(db, key) {
-  const stmt = db.prepare('SELECT metadata FROM memory_entries WHERE key = ? AND namespace = ?');
-  stmt.bind([key, NAMESPACE]);
-  const entry = stmt.step() ? stmt.getAsObject() : null;
-  stmt.free();
-  if (entry?.metadata) {
-    try {
-      const meta = JSON.parse(entry.metadata);
-      return meta.contentHash;
-    } catch { /* ignore */ }
-  }
-  return null;
-}
-
-/**
- * Extract overlapping context from adjacent text
- * @param {string} text - The text to extract from
- * @param {number} percent - Percentage of text to extract
- * @param {string} position - 'start' or 'end'
- * @returns {string} - The extracted context
- */
-function extractOverlapContext(text, percent, position) {
-  if (!text || percent <= 0) return '';
-
-  const targetLength = Math.floor(text.length * (percent / 100));
-  if (targetLength < 20) return ''; // Too short to be useful
-
-  if (position === 'start') {
-    // Get first N% of text, try to break at sentence/paragraph
-    let end = targetLength;
-    const nextPara = text.indexOf('\n\n', targetLength - 50);
-    const nextSentence = text.indexOf('. ', targetLength - 30);
-
-    if (nextPara > 0 && nextPara < targetLength + 100) {
-      end = nextPara;
-    } else if (nextSentence > 0 && nextSentence < targetLength + 50) {
-      end = nextSentence + 1;
-    }
-
-    return text.substring(0, end).trim();
-  } else {
-    // Get last N% of text, try to break at sentence/paragraph
-    let start = text.length - targetLength;
-    const prevPara = text.lastIndexOf('\n\n', start + 50);
-    const prevSentence = text.lastIndexOf('. ', start + 30);
-
-    if (prevPara > 0 && prevPara > start - 100) {
-      start = prevPara + 2;
-    } else if (prevSentence > 0 && prevSentence > start - 50) {
-      start = prevSentence + 2;
-    }
-
-    return text.substring(start).trim();
-  }
-}
-
-/**
- * Split markdown content into semantic chunks based on headers
- * Returns array of { title, content, level, headerLine }
- */
-function chunkMarkdown(content, fileName) {
-  const lines = content.split('\n');
-  const chunks = [];
-  let currentChunk = { title: fileName, content: [], level: 0, headerLine: 0 };
-
-  for (let lineNum = 0; lineNum < lines.length; lineNum++) {
-    // Strip CRLF carriage returns for Windows compatibility
-    const line = lines[lineNum].replace(/\r$/, '');
-
-    // Check for headers (## and ###)
-    const h2Match = line.match(/^## (.+)$/);
-    const h3Match = line.match(/^### (.+)$/);
-
-    if (h2Match || h3Match) {
-      // Save current chunk if it has content
-      if (currentChunk.content.length > 0) {
-        const chunkContent = currentChunk.content.join('\n').trim();
-        if (chunkContent.length >= MIN_CHUNK_SIZE) {
-          chunks.push({
-            title: currentChunk.title,
-            content: chunkContent,
-            level: currentChunk.level,
-            headerLine: currentChunk.headerLine
-          });
-        }
-      }
-
-      // Start new chunk
-      currentChunk = {
-        title: h2Match ? h2Match[1] : h3Match[1],
-        content: [line],
-        level: h2Match ? 2 : 3,
-        headerLine: lineNum
-      };
-    } else {
-      currentChunk.content.push(line);
-    }
-  }
-
-  // Don't forget the last chunk
-  if (currentChunk.content.length > 0) {
-    const chunkContent = currentChunk.content.join('\n').trim();
-    if (chunkContent.length >= MIN_CHUNK_SIZE) {
-      chunks.push({
-        title: currentChunk.title,
-        content: chunkContent,
-        level: currentChunk.level,
-        headerLine: currentChunk.headerLine
-      });
-    }
-  }
-
-  // Handle chunks that are too large - split by paragraphs
-  const finalChunks = [];
-  for (const chunk of chunks) {
-    if (chunk.content.length > MAX_CHUNK_SIZE) {
-      const paragraphs = chunk.content.split(/\n\n+/);
-      let currentPart = [];
-      let currentLength = 0;
-      let partNum = 1;
-
-      for (const para of paragraphs) {
-        if (currentLength + para.length > MAX_CHUNK_SIZE && currentPart.length > 0) {
-          finalChunks.push({
-            title: `${chunk.title} (part ${partNum})`,
-            content: currentPart.join('\n\n'),
-            level: chunk.level,
-            headerLine: chunk.headerLine,
-            isPart: true,
-            partNum
-          });
-          currentPart = [para];
-          currentLength = para.length;
-          partNum++;
-        } else {
-          currentPart.push(para);
-          currentLength += para.length;
-        }
-      }
-
-      if (currentPart.length > 0) {
-        finalChunks.push({
-          title: partNum > 1 ? `${chunk.title} (part ${partNum})` : chunk.title,
-          content: currentPart.join('\n\n'),
-          level: chunk.level,
-          headerLine: chunk.headerLine,
-          isPart: partNum > 1,
-          partNum: partNum > 1 ? partNum : undefined
-        });
-      }
-    } else {
-      finalChunks.push(chunk);
-    }
-  }
-
-  // FORCE CHUNKING: If file is large but resulted in few chunks, split by sections
-  const totalContent = finalChunks.reduce((acc, c) => acc + c.content.length, 0);
-  if (totalContent > FORCE_CHUNK_THRESHOLD && finalChunks.length < 3) {
-    debug(`  Force-chunking: ${totalContent} bytes in ${finalChunks.length} chunks - splitting by sections`);
-    const allContent = finalChunks.map(c => c.content).join('\n\n');
-
-    // Split on --- horizontal rules first, then on ## headers, then on paragraphs
-    const TARGET_CHUNK_SIZE = 2500;
-    const rawSections = allContent.split(/\n---+\n/);
-    let sections = [];
-
-    for (const raw of rawSections) {
-      // Further split on ## headers if section is too large
-      if (raw.length > TARGET_CHUNK_SIZE) {
-        const headerSplit = raw.split(/\n(?=## )/);
-        for (const hSect of headerSplit) {
-          if (hSect.length > TARGET_CHUNK_SIZE) {
-            // Split very long sections on single newlines as last resort
-            const lines = hSect.split('\n');
-            let chunk = '';
-            for (const line of lines) {
-              if (chunk.length + line.length > TARGET_CHUNK_SIZE && chunk.length > 100) {
-                sections.push(chunk.trim());
-                chunk = line;
-              } else {
-                chunk += (chunk ? '\n' : '') + line;
-              }
-            }
-            if (chunk.trim().length > 30) sections.push(chunk.trim());
-          } else if (hSect.trim().length > 30) {
-            sections.push(hSect.trim());
-          }
-        }
-      } else if (raw.trim().length > 30) {
-        sections.push(raw.trim());
-      }
-    }
-
-    // Now group sections into chunks
-    const forcedChunks = [];
-    let currentGroup = [];
-    let currentLength = 0;
-    let groupNum = 1;
-
-    const flushGroup = () => {
-      if (currentGroup.length === 0) return;
-      const firstLine = currentGroup[0].split('\n')[0].trim();
-      const title = firstLine.startsWith('#')
-        ? firstLine.replace(/^#+\s*/, '').slice(0, 60)
-        : `${fileName} Section ${groupNum}`;
-
-      forcedChunks.push({
-        title,
-        content: currentGroup.join('\n\n'),
-        level: 2,
-        headerLine: 0,
-        isForced: true,
-        forceNum: groupNum
-      });
-      groupNum++;
-      currentGroup = [];
-      currentLength = 0;
-    };
-
-    for (const section of sections) {
-      if (currentLength + section.length > TARGET_CHUNK_SIZE && currentGroup.length > 0) {
-        flushGroup();
-      }
-      currentGroup.push(section);
-      currentLength += section.length;
-    }
-    flushGroup();
-
-    // Always use force-chunked results if we got multiple chunks
-    if (forcedChunks.length >= 2) {
-      debug(`  Force-chunking produced ${forcedChunks.length} chunks (was ${finalChunks.length})`);
-      return forcedChunks;
-    }
-  }
-
-  return finalChunks;
-}
-
-/**
- * Build hierarchical relationships between chunks
- * H2 chunks are parents of subsequent H3 chunks
- */
-function buildHierarchy(chunks, chunkPrefix) {
-  const hierarchy = {};
-  let currentH2Index = null;
-
-  for (let i = 0; i < chunks.length; i++) {
-    const chunk = chunks[i];
-    const chunkKey = `${chunkPrefix}-${i}`;
-
-    hierarchy[chunkKey] = {
-      parent: null,
-      children: []
-    };
-
-    if (chunk.level === 2) {
-      currentH2Index = i;
-    } else if (chunk.level === 3 && currentH2Index !== null) {
-      const parentKey = `${chunkPrefix}-${currentH2Index}`;
-      hierarchy[chunkKey].parent = parentKey;
-      hierarchy[parentKey].children.push(chunkKey);
-    }
-  }
-
-  return hierarchy;
-}
-
-function indexFile(db, filePath, keyPrefix) {
-  const fileName = basename(filePath, extname(filePath));
-  const docKey = `doc-${keyPrefix}-${fileName}`;
-  const chunkPrefix = `chunk-${keyPrefix}-${fileName}`;
-
-  try {
-    const content = readFileSync(filePath, 'utf-8');
-    const contentHash = hashContent(content);
-
-    // Check if content changed (skip if same hash unless --force)
-    if (!force) {
-      const existingHash = getEntryHash(db, docKey);
-      if (existingHash === contentHash) {
-        return { docKey, status: 'unchanged', chunks: 0 };
-      }
-    }
-
-    const stats = statSync(filePath);
-    const relativePath = filePath.replace(projectRoot, '').replace(/\\/g, '/');
-
-    // Delete old chunks for this file before re-indexing
-    deleteByPrefix(db, chunkPrefix);
-
-    // 1. Store full document
-    const docMetadata = {
-      type: 'document',
-      filePath: relativePath,
-      fileSize: stats.size,
-      lastModified: stats.mtime.toISOString(),
-      contentHash,
-      indexedAt: new Date().toISOString(),
-      ragVersion: '2.0',  // Mark as full RAG indexed
-    };
-
-    storeEntry(db, docKey, content, docMetadata, [keyPrefix, 'document']);
-    debug(`Stored document: ${docKey}`);
-
-    // 2. Chunk and store semantic pieces with full RAG linking
-    const chunks = chunkMarkdown(content, fileName);
-
-    if (chunks.length === 0) {
-      return { docKey, status: 'indexed', chunks: 0 };
-    }
-
-    // Build hierarchy and sibling list
-    const hierarchy = buildHierarchy(chunks, chunkPrefix);
-    const siblings = chunks.map((_, i) => `${chunkPrefix}-${i}`);
-
-    // Update document with children references
-    const docChildrenMeta = {
-      ...docMetadata,
-      children: siblings,
-      chunkCount: chunks.length,
-    };
-    storeEntry(db, docKey, content, docChildrenMeta, [keyPrefix, 'document']);
-
-    for (let i = 0; i < chunks.length; i++) {
-      const chunk = chunks[i];
-      const chunkKey = `${chunkPrefix}-${i}`;
-
-      // Build prev/next links
-      const prevChunk = i > 0 ? `${chunkPrefix}-${i - 1}` : null;
-      const nextChunk = i < chunks.length - 1 ? `${chunkPrefix}-${i + 1}` : null;
-
-      // Extract overlapping context from adjacent chunks
-      const contextBefore = i > 0
-        ? extractOverlapContext(chunks[i - 1].content, overlapPercent, 'end')
-        : null;
-      const contextAfter = i < chunks.length - 1
-        ? extractOverlapContext(chunks[i + 1].content, overlapPercent, 'start')
-        : null;
-
-      // Get hierarchical relationships
-      const hierInfo = hierarchy[chunkKey];
-
-      const chunkMetadata = {
-        type: 'chunk',
-        ragVersion: '2.0',
-
-        // Document relationship
-        parentDoc: docKey,
-        parentPath: relativePath,
-
-        // Sequential navigation (forward/backward links)
-        chunkIndex: i,
-        totalChunks: chunks.length,
-        prevChunk,
-        nextChunk,
-
-        // Sibling awareness
-        siblings,
-
-        // Hierarchical relationships (h2 -> h3)
-        hierarchicalParent: hierInfo.parent,
-        hierarchicalChildren: hierInfo.children.length > 0 ? hierInfo.children : null,
-
-        // Chunk info
-        chunkTitle: chunk.title,
-        headerLevel: chunk.level,
-        headerLine: chunk.headerLine,
-        isPart: chunk.isPart || false,
-        partNum: chunk.partNum || null,
-
-        // Overlapping context for continuity
-        contextOverlapPercent: overlapPercent,
-        hasContextBefore: !!contextBefore,
-        hasContextAfter: !!contextAfter,
-
-        // Content metadata
-        contentLength: chunk.content.length,
-        contentHash: hashContent(chunk.content),
-        indexedAt: new Date().toISOString(),
-      };
-
-      // Build searchable content with title context
-      // Include overlap context for better retrieval
-      let searchableContent = `# ${chunk.title}\n\n`;
-
-      if (contextBefore) {
-        searchableContent += `[Context from previous section:]\n${contextBefore}\n\n---\n\n`;
-      }
-
-      searchableContent += chunk.content;
-
-      if (contextAfter) {
-        searchableContent += `\n\n---\n\n[Context from next section:]\n${contextAfter}`;
-      }
-
-      // Store chunk with full metadata
-      storeEntry(
-        db,
-        chunkKey,
-        searchableContent,
-        chunkMetadata,
-        [keyPrefix, 'chunk', `level-${chunk.level}`, chunk.title.toLowerCase().replace(/[^a-z0-9]+/g, '-')]
-      );
-
-      debug(`  Stored chunk ${i}: ${chunk.title} (${chunk.content.length} chars, prev=${!!prevChunk}, next=${!!nextChunk})`);
-    }
-
-    return { docKey, status: 'indexed', chunks: chunks.length };
-  } catch (err) {
-    return { docKey, status: 'error', error: err.message, chunks: 0 };
-  }
-}
-
-/**
- * Recursively collect all .md files under a directory.
- * Skips node_modules, .git, and other non-content directories.
- */
-function walkMdFiles(dir) {
-  const SKIP_DIRS = new Set(['node_modules', '.git', 'dist', 'build', 'coverage', '.next', '.reports']);
-  // CLAUDE.md is loaded into context by Claude automatically — skip to avoid duplicate vectors
-  const SKIP_FILES = new Set(['CLAUDE.md']);
-  const files = [];
-
-  function walk(current) {
-    if (!existsSync(current)) return;
-    for (const entry of readdirSync(current, { withFileTypes: true })) {
-      if (entry.isDirectory()) {
-        if (!SKIP_DIRS.has(entry.name)) walk(resolve(current, entry.name));
-      } else if (entry.isFile() && entry.name.endsWith('.md') && !SKIP_FILES.has(entry.name)) {
-        files.push(resolve(current, entry.name));
-      }
-    }
-  }
-
-  walk(dir);
-  return files;
-}
-
-function indexDirectory(db, dirConfig) {
-  const dirPath = dirConfig.absolute ? dirConfig.path : resolve(projectRoot, dirConfig.path);
-  const results = [];
-
-  if (!existsSync(dirPath)) {
-    if (verbose) debug(`Directory not found: ${dirConfig.path}`);
-    return results;
-  }
-
-  const allMdFiles = walkMdFiles(dirPath);
-  const filtered = dirConfig.fileFilter
-    ? allMdFiles.filter(f => dirConfig.fileFilter.includes(basename(f)))
-    : allMdFiles;
-
-  for (const filePath of filtered) {
-    const result = indexFile(db, filePath, dirConfig.prefix);
-    results.push(result);
-  }
-
-  return results;
-}
-
-/**
- * Remove stale entries for files that no longer exist on disk.
- * Uses the set of docKeys seen during the current indexing run to determine
- * which entries are stale, rather than reconstructing file paths from keys
- * (which breaks for files in subdirectories).
- */
-function cleanStaleEntries(db, currentDocKeys) {
-  const docsStmt = db.prepare(
-    `SELECT DISTINCT key FROM memory_entries WHERE namespace = ? AND key LIKE 'doc-%'`
-  );
-  docsStmt.bind([NAMESPACE]);
-  const docs = [];
-  while (docsStmt.step()) docs.push(docsStmt.getAsObject());
-  docsStmt.free();
-
-  let staleCount = 0;
-
-  for (const { key } of docs) {
-    // If this doc key was seen during the current indexing run, it's not stale
-    if (currentDocKeys.has(key)) continue;
-
-    const chunkPrefix = key.replace('doc-', 'chunk-');
-    const countBefore = db.exec(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = '${NAMESPACE}'`)[0]?.values[0][0] || 0;
-    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key LIKE ?`, [NAMESPACE, `${chunkPrefix}%`]);
-    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key = ?`, [NAMESPACE, key]);
-    const countAfter = db.exec(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = '${NAMESPACE}'`)[0]?.values[0][0] || 0;
-    const removed = countBefore - countAfter;
-    if (removed > 0) {
-      log(`  Removed ${removed} stale entries for deleted file: ${key}`);
-      staleCount += removed;
-    }
-  }
-
-  // Also clean any orphaned entries not matching doc-/chunk- patterns
-  const orphanStmt = db.prepare(
-    `SELECT key FROM memory_entries WHERE namespace = ? AND key NOT LIKE 'doc-%' AND key NOT LIKE 'chunk-%'`
-  );
-  orphanStmt.bind([NAMESPACE]);
-  const orphans = [];
-  while (orphanStmt.step()) orphans.push(orphanStmt.getAsObject());
-  orphanStmt.free();
-  for (const { key } of orphans) {
-    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key = ?`, [NAMESPACE, key]);
-    staleCount++;
-    log(`  Removed orphan entry: ${key}`);
-  }
-
-  return staleCount;
-}
-
-// Main
-console.log('');
-log('Indexing guidance files with FULL RAG linked segments...');
-log(`  Context overlap: ${overlapPercent}%`);
-log(`  Directories (${GUIDANCE_DIRS.length}):`);
-for (const d of GUIDANCE_DIRS) {
-  const dirPath = d.absolute ? d.path : resolve(projectRoot, d.path);
-  const exists = existsSync(dirPath);
-  log(`    ${exists ? '✓' : '✗'} ${d.absolute ? dirPath : d.path} [${d.prefix}]`);
-}
-console.log('');
-
-const db = await getDb();
-let docsIndexed = 0;
-let chunksIndexed = 0;
-let unchanged = 0;
-let errors = 0;
-const currentDocKeys = new Set();
-
-if (specificFile) {
-  // Index single file
-  const filePath = resolve(projectRoot, specificFile);
-  if (!existsSync(filePath)) {
-    log(`File not found: ${specificFile}`);
-    process.exit(1);
-  }
-
-  let prefix = 'docs';
-  if (specificFile.includes('.claude/guidance/')) {
-    prefix = 'guidance';
-  }
-
-  const result = indexFile(db, filePath, prefix);
-  log(`${result.docKey}: ${result.status} (${result.chunks} chunks)`);
-
-  if (result.status === 'indexed') {
-    docsIndexed++;
-    chunksIndexed += result.chunks;
-  } else if (result.status === 'unchanged') {
-    unchanged++;
-  } else {
-    errors++;
-  }
-} else {
-  // Index all directories
-  for (const dir of GUIDANCE_DIRS) {
-    log(`Scanning ${dir.path}/...`);
-    const results = indexDirectory(db, dir);
-
-    for (const result of results) {
-      if (result.status === 'indexed' || result.status === 'unchanged') {
-        currentDocKeys.add(result.docKey);
-      }
-      if (result.status === 'indexed') {
-        log(`  ✅ ${result.docKey} (${result.chunks} chunks)`);
-        docsIndexed++;
-        chunksIndexed += result.chunks;
-      } else if (result.status === 'unchanged') {
-        unchanged++;
-      } else {
-        log(`  ❌ ${result.docKey}: ${result.error}`);
-        errors++;
-      }
-    }
-  }
-}
-
-// Clean stale entries for deleted files (unless indexing a specific file)
-let staleRemoved = 0;
-if (!specificFile) {
-  log('Cleaning stale entries for deleted files...');
-  staleRemoved = cleanStaleEntries(db, currentDocKeys);
-  if (staleRemoved === 0) {
-    log('  No stale entries found');
-  }
-}
-
-// Write changes back to disk
-if (docsIndexed > 0 || chunksIndexed > 0 || staleRemoved > 0) saveDb(db);
-
-// Check for entries missing embeddings (e.g. prior background run failed)
-let missingEmbeddings = 0;
-{
-  const stmt = db.prepare(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = ? AND (embedding IS NULL OR embedding = '')`);
-  stmt.bind([NAMESPACE]);
-  if (stmt.step()) missingEmbeddings = stmt.getAsObject().cnt;
-  stmt.free();
-}
-db.close();
-
-console.log('');
-log('═══════════════════════════════════════════════════════════');
-log('  FULL RAG INDEXING COMPLETE');
-log('═══════════════════════════════════════════════════════════');
-log(`  Documents indexed:    ${docsIndexed}`);
-log(`  Chunks created:       ${chunksIndexed}`);
-log(`  Unchanged:            ${unchanged}`);
-log(`  Stale removed:        ${staleRemoved}`);
-log(`  Errors:               ${errors}`);
-log('');
-log('  RAG Features Enabled:');
-log(`    • Forward/backward links (prevChunk/nextChunk)`);
-log(`    • Sibling awareness (all chunks from same doc)`);
-log(`    • Hierarchical links (h2 -> h3 parent/children)`);
-log(`    • Context overlap: ${overlapPercent}% (contextBefore/contextAfter)`);
-log('═══════════════════════════════════════════════════════════');
-
-// Generate embeddings for new entries or backfill missing ones
-// Runs in BACKGROUND to avoid blocking startup
-const needsEmbeddings = (docsIndexed > 0 || chunksIndexed > 0 || missingEmbeddings > 0);
-if (!skipEmbeddings && needsEmbeddings) {
-  if (missingEmbeddings > 0 && docsIndexed === 0 && chunksIndexed === 0) {
-    log(`${missingEmbeddings} entries missing embeddings — backfilling...`);
-  }
-  console.log('');
-  log('Spawning embedding generation in background...');
-
-  const { spawn } = await import('child_process');
-
-  // Look for build-embeddings script in multiple locations:
-  // 1. Shipped with moflo (node_modules/moflo/bin/)
-  // 2. Project-local (.claude/scripts/)
-  const mofloScript = resolve(__dirname, 'build-embeddings.mjs');
-  const projectLocalScript = resolve(projectRoot, '.claude/scripts/build-embeddings.mjs');
-  const embeddingScript = existsSync(mofloScript) ? mofloScript : projectLocalScript;
-
-  if (existsSync(embeddingScript)) {
-    const embeddingArgs = ['--namespace', NAMESPACE];
-
-    // Create log file for background process output
-    const logDir = resolve(projectRoot, '.swarm/logs');
-    if (!existsSync(logDir)) {
-      mkdirSync(logDir, { recursive: true });
-    }
-    const logFile = resolve(logDir, 'embeddings.log');
-    const { openSync } = await import('fs');
-    const out = openSync(logFile, 'a');
-    const err = openSync(logFile, 'a');
-
-    // Spawn in background - don't wait for completion
-    const proc = spawn('node', [embeddingScript, ...embeddingArgs], {
-      stdio: ['ignore', out, err],
-      cwd: projectRoot,
-      detached: true,
-      windowsHide: true  // Suppress command windows on Windows
-    });
-    proc.unref();  // Allow parent to exit independently
-
-    log(`Background embedding started (PID: ${proc.pid})`);
-    log(`Log file: .swarm/logs/embeddings.log`);
-  } else {
-    log('⚠️  Embedding script not found, skipping embedding generation');
-  }
-} else if (skipEmbeddings) {
-  log('Skipping embedding generation (--no-embeddings)');
-} else {
-  log('No new content indexed, skipping embedding generation');
-}
-
-if (errors > 0) {
-  process.exit(1);
-}
+/**
+ * Index guidance files into claude-flow memory with full RAG linked segments
+ *
+ * Strategy:
+ * - Full documents stored as `doc-{name}` for complete retrieval
+ * - Semantic chunks stored as `chunk-{name}-{n}` for precise search
+ * - FULL RAG LINKING:
+ *   - parentDoc: link to full document
+ *   - prevChunk/nextChunk: forward/backward navigation
+ *   - siblings: all chunk keys from same document
+ *   - children: sub-chunks for hierarchical headers (h2 -> h3)
+ *   - contextBefore/contextAfter: overlapping text for context continuity
+ * - Chunking based on markdown headers (## and ###) for natural boundaries
+ * - After indexing, generates embeddings for semantic search (HNSW)
+ *
+ * Usage:
+ *   node node_modules/moflo/bin/index-guidance.mjs                 # Index all + generate embeddings
+ *   npx flo-index --force                                        # Force reindex all
+ *   npx flo-index --file X                                       # Index specific file
+ *   npx flo-index --no-embeddings                                # Skip embedding generation
+ *   npx flo-index --overlap 20                                   # Set context overlap % (default: 15)
+ */
+
+import { existsSync, readdirSync, readFileSync, statSync, mkdirSync, writeFileSync } from 'fs';
+import { resolve, dirname, basename, extname } from 'path';
+import { fileURLToPath } from 'url';
+import { mofloResolveURL } from './lib/moflo-resolve.mjs';
+const initSqlJs = (await import(mofloResolveURL('sql.js'))).default;
+
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+
+function findProjectRoot() {
+  let dir = process.cwd();
+  const root = resolve(dir, '/');
+  while (dir !== root) {
+    if (existsSync(resolve(dir, 'package.json'))) return dir;
+    dir = dirname(dir);
+  }
+  return process.cwd();
+}
+
+const projectRoot = findProjectRoot();
+
+// Locate the moflo package root (for bundled guidance that ships with moflo)
+const mofloRoot = resolve(__dirname, '..');
+
+const NAMESPACE = 'guidance';
+const DB_PATH = resolve(projectRoot, '.swarm/memory.db');
+
+// ============================================================================
+// Load guidance directories from moflo.yaml, falling back to defaults
+// ============================================================================
+
+function loadGuidanceDirs() {
+  const dirs = [];
+
+  // 1. Read moflo.yaml / moflo.config.json for user-configured directories
+  let configDirs = null;
+  const yamlPath = resolve(projectRoot, 'moflo.yaml');
+  const jsonPath = resolve(projectRoot, 'moflo.config.json');
+
+  if (existsSync(yamlPath)) {
+    try {
+      const content = readFileSync(yamlPath, 'utf-8');
+      // Simple YAML array extraction — avoids needing js-yaml at runtime
+      // Matches:  guidance:\n    directories:\n      - .claude/guidance\n      - docs/guides
+      const guidanceBlock = content.match(/guidance:\s*\n\s+directories:\s*\n((?:\s+-\s+.+\n?)+)/);
+      if (guidanceBlock) {
+        const items = guidanceBlock[1].match(/-\s+(.+)/g);
+        if (items && items.length > 0) {
+          configDirs = items.map(item => item.replace(/^-\s+/, '').trim());
+        }
+      }
+    } catch { /* ignore parse errors, fall through to defaults */ }
+  } else if (existsSync(jsonPath)) {
+    try {
+      const raw = JSON.parse(readFileSync(jsonPath, 'utf-8'));
+      if (raw.guidance?.directories && Array.isArray(raw.guidance.directories)) {
+        configDirs = raw.guidance.directories;
+      }
+    } catch { /* ignore parse errors */ }
+  }
+
+  // Use config dirs or fall back to defaults
+  // Each directory gets a unique prefix derived from its path to avoid key collisions
+  // when multiple directories contain files with the same name.
+  const userDirs = configDirs || ['.claude/guidance', 'docs/guides'];
+  for (const d of userDirs) {
+    const prefix = d.replace(/\\/g, '/')
+      .replace(/^\.claude\//, '')
+      .replace(/^back-office\/api\/\.claude\//, 'bo-api-')
+      .replace(/^back-office\/ui\/\.claude\//, 'bo-ui-')
+      .replace(/[^a-zA-Z0-9-]/g, '-')
+      .replace(/-+/g, '-')
+      .replace(/^-|-$/g, '') || 'guidance';
+    dirs.push({ path: d, prefix });
+  }
+
+  // 2. Include moflo's own bundled guidance (ships with the package)
+  //    Only when running inside a consumer project (not moflo itself)
+  //    Shipped guidance lives in .claude/guidance/shipped/ — internal/ is excluded from npm
+  const bundledShippedDir = resolve(mofloRoot, '.claude/guidance/shipped');
+  const bundledGuidanceDir = existsSync(bundledShippedDir)
+    ? bundledShippedDir
+    : resolve(mofloRoot, '.claude/guidance');
+  const projectGuidanceDir = resolve(projectRoot, '.claude/guidance');
+  if (
+    existsSync(bundledGuidanceDir) &&
+    resolve(bundledGuidanceDir) !== resolve(projectGuidanceDir) &&
+    resolve(bundledGuidanceDir) !== resolve(projectGuidanceDir, 'shipped')
+  ) {
+    dirs.push({ path: bundledGuidanceDir, prefix: 'moflo-bundled', absolute: true });
+  }
+
+  // 3. CLAUDE.md files are NOT indexed — Claude loads them into context automatically.
+  //    Indexing them wastes vectors and creates duplicate keys across subprojects.
+
+  return dirs;
+}
+
+const GUIDANCE_DIRS = loadGuidanceDirs();
+
+// Chunking config - optimized for Claude's retrieval
+const MIN_CHUNK_SIZE = 50;    // Lower minimum to avoid mega-chunks
+const MAX_CHUNK_SIZE = 4000;  // Larger chunks for code-heavy docs (fits context better)
+const FORCE_CHUNK_THRESHOLD = 6000; // Force paragraph-split if file > this and < 3 chunks
+const DEFAULT_OVERLAP_PERCENT = 20; // Increased context overlap for better continuity
+
+// Parse args
+const args = process.argv.slice(2);
+const force = args.includes('--force');
+const specificFile = args.includes('--file') ? args[args.indexOf('--file') + 1] : null;
+const verbose = args.includes('--verbose') || args.includes('-v');
+const skipEmbeddings = args.includes('--no-embeddings');
+const overlapPercent = args.includes('--overlap')
+  ? parseInt(args[args.indexOf('--overlap') + 1], 10) || DEFAULT_OVERLAP_PERCENT
+  : DEFAULT_OVERLAP_PERCENT;
+
+function log(msg) {
+  console.log(`[index-guidance] ${msg}`);
+}
+
+function debug(msg) {
+  if (verbose) console.log(`[index-guidance]   ${msg}`);
+}
+
+function ensureDbDir() {
+  const dir = dirname(DB_PATH);
+  if (!existsSync(dir)) {
+    mkdirSync(dir, { recursive: true });
+  }
+}
+
+async function getDb() {
+  ensureDbDir();
+  const SQL = await initSqlJs();
+  let db;
+  if (existsSync(DB_PATH)) {
+    const buffer = readFileSync(DB_PATH);
+    db = new SQL.Database(buffer);
+  } else {
+    db = new SQL.Database();
+  }
+
+  // Ensure table exists with unique constraint
+  db.run(`
+    CREATE TABLE IF NOT EXISTS memory_entries (
+      id TEXT PRIMARY KEY,
+      key TEXT NOT NULL,
+      namespace TEXT DEFAULT 'default',
+      content TEXT NOT NULL,
+      type TEXT DEFAULT 'semantic',
+      embedding TEXT,
+      embedding_model TEXT DEFAULT 'local',
+      embedding_dimensions INTEGER,
+      tags TEXT,
+      metadata TEXT,
+      owner_id TEXT,
+      created_at INTEGER NOT NULL DEFAULT (strftime('%s', 'now') * 1000),
+      updated_at INTEGER NOT NULL DEFAULT (strftime('%s', 'now') * 1000),
+      expires_at INTEGER,
+      last_accessed_at INTEGER,
+      access_count INTEGER DEFAULT 0,
+      status TEXT DEFAULT 'active',
+      UNIQUE(namespace, key)
+    )
+  `);
+
+  db.run(`CREATE INDEX IF NOT EXISTS idx_memory_key_ns ON memory_entries(key, namespace)`);
+  db.run(`CREATE INDEX IF NOT EXISTS idx_memory_namespace ON memory_entries(namespace)`);
+
+  return db;
+}
+
+function saveDb(db) {
+  const data = db.export();
+  writeFileSync(DB_PATH, Buffer.from(data));
+}
+
+function generateId() {
+  return `mem_${Date.now()}_${Math.random().toString(36).substr(2, 9)}`;
+}
+
+function hashContent(content) {
+  let hash = 0;
+  for (let i = 0; i < content.length; i++) {
+    const char = content.charCodeAt(i);
+    hash = ((hash << 5) - hash) + char;
+    hash = hash & hash;
+  }
+  return hash.toString(16);
+}
+
+function storeEntry(db, key, content, metadata = {}, tags = []) {
+  const now = Date.now();
+  const id = generateId();
+  const metaJson = JSON.stringify(metadata);
+  const tagsJson = JSON.stringify(tags);
+
+  db.run(`
+    INSERT OR REPLACE INTO memory_entries
+    (id, key, namespace, content, metadata, tags, created_at, updated_at, status)
+    VALUES (?, ?, ?, ?, ?, ?, ?, ?, 'active')
+  `, [id, key, NAMESPACE, content, metaJson, tagsJson, now, now]);
+
+  return true;
+}
+
+function deleteByPrefix(db, prefix) {
+  db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key LIKE ?`, [NAMESPACE, `${prefix}%`]);
+}
+
+function getEntryHash(db, key) {
+  const stmt = db.prepare('SELECT metadata FROM memory_entries WHERE key = ? AND namespace = ?');
+  stmt.bind([key, NAMESPACE]);
+  const entry = stmt.step() ? stmt.getAsObject() : null;
+  stmt.free();
+  if (entry?.metadata) {
+    try {
+      const meta = JSON.parse(entry.metadata);
+      return meta.contentHash;
+    } catch { /* ignore */ }
+  }
+  return null;
+}
+
+/**
+ * Extract overlapping context from adjacent text
+ * @param {string} text - The text to extract from
+ * @param {number} percent - Percentage of text to extract
+ * @param {string} position - 'start' or 'end'
+ * @returns {string} - The extracted context
+ */
+function extractOverlapContext(text, percent, position) {
+  if (!text || percent <= 0) return '';
+
+  const targetLength = Math.floor(text.length * (percent / 100));
+  if (targetLength < 20) return ''; // Too short to be useful
+
+  if (position === 'start') {
+    // Get first N% of text, try to break at sentence/paragraph
+    let end = targetLength;
+    const nextPara = text.indexOf('\n\n', targetLength - 50);
+    const nextSentence = text.indexOf('. ', targetLength - 30);
+
+    if (nextPara > 0 && nextPara < targetLength + 100) {
+      end = nextPara;
+    } else if (nextSentence > 0 && nextSentence < targetLength + 50) {
+      end = nextSentence + 1;
+    }
+
+    return text.substring(0, end).trim();
+  } else {
+    // Get last N% of text, try to break at sentence/paragraph
+    let start = text.length - targetLength;
+    const prevPara = text.lastIndexOf('\n\n', start + 50);
+    const prevSentence = text.lastIndexOf('. ', start + 30);
+
+    if (prevPara > 0 && prevPara > start - 100) {
+      start = prevPara + 2;
+    } else if (prevSentence > 0 && prevSentence > start - 50) {
+      start = prevSentence + 2;
+    }
+
+    return text.substring(start).trim();
+  }
+}
+
+/**
+ * Split markdown content into semantic chunks based on headers
+ * Returns array of { title, content, level, headerLine }
+ */
+function chunkMarkdown(content, fileName) {
+  const lines = content.split('\n');
+  const chunks = [];
+  let currentChunk = { title: fileName, content: [], level: 0, headerLine: 0 };
+
+  for (let lineNum = 0; lineNum < lines.length; lineNum++) {
+    // Strip CRLF carriage returns for Windows compatibility
+    const line = lines[lineNum].replace(/\r$/, '');
+
+    // Check for headers (## and ###)
+    const h2Match = line.match(/^## (.+)$/);
+    const h3Match = line.match(/^### (.+)$/);
+
+    if (h2Match || h3Match) {
+      // Save current chunk if it has content
+      if (currentChunk.content.length > 0) {
+        const chunkContent = currentChunk.content.join('\n').trim();
+        if (chunkContent.length >= MIN_CHUNK_SIZE) {
+          chunks.push({
+            title: currentChunk.title,
+            content: chunkContent,
+            level: currentChunk.level,
+            headerLine: currentChunk.headerLine
+          });
+        }
+      }
+
+      // Start new chunk
+      currentChunk = {
+        title: h2Match ? h2Match[1] : h3Match[1],
+        content: [line],
+        level: h2Match ? 2 : 3,
+        headerLine: lineNum
+      };
+    } else {
+      currentChunk.content.push(line);
+    }
+  }
+
+  // Don't forget the last chunk
+  if (currentChunk.content.length > 0) {
+    const chunkContent = currentChunk.content.join('\n').trim();
+    if (chunkContent.length >= MIN_CHUNK_SIZE) {
+      chunks.push({
+        title: currentChunk.title,
+        content: chunkContent,
+        level: currentChunk.level,
+        headerLine: currentChunk.headerLine
+      });
+    }
+  }
+
+  // Handle chunks that are too large - split by paragraphs
+  const finalChunks = [];
+  for (const chunk of chunks) {
+    if (chunk.content.length > MAX_CHUNK_SIZE) {
+      const paragraphs = chunk.content.split(/\n\n+/);
+      let currentPart = [];
+      let currentLength = 0;
+      let partNum = 1;
+
+      for (const para of paragraphs) {
+        if (currentLength + para.length > MAX_CHUNK_SIZE && currentPart.length > 0) {
+          finalChunks.push({
+            title: `${chunk.title} (part ${partNum})`,
+            content: currentPart.join('\n\n'),
+            level: chunk.level,
+            headerLine: chunk.headerLine,
+            isPart: true,
+            partNum
+          });
+          currentPart = [para];
+          currentLength = para.length;
+          partNum++;
+        } else {
+          currentPart.push(para);
+          currentLength += para.length;
+        }
+      }
+
+      if (currentPart.length > 0) {
+        finalChunks.push({
+          title: partNum > 1 ? `${chunk.title} (part ${partNum})` : chunk.title,
+          content: currentPart.join('\n\n'),
+          level: chunk.level,
+          headerLine: chunk.headerLine,
+          isPart: partNum > 1,
+          partNum: partNum > 1 ? partNum : undefined
+        });
+      }
+    } else {
+      finalChunks.push(chunk);
+    }
+  }
+
+  // FORCE CHUNKING: If file is large but resulted in few chunks, split by sections
+  const totalContent = finalChunks.reduce((acc, c) => acc + c.content.length, 0);
+  if (totalContent > FORCE_CHUNK_THRESHOLD && finalChunks.length < 3) {
+    debug(`  Force-chunking: ${totalContent} bytes in ${finalChunks.length} chunks - splitting by sections`);
+    const allContent = finalChunks.map(c => c.content).join('\n\n');
+
+    // Split on --- horizontal rules first, then on ## headers, then on paragraphs
+    const TARGET_CHUNK_SIZE = 2500;
+    const rawSections = allContent.split(/\n---+\n/);
+    let sections = [];
+
+    for (const raw of rawSections) {
+      // Further split on ## headers if section is too large
+      if (raw.length > TARGET_CHUNK_SIZE) {
+        const headerSplit = raw.split(/\n(?=## )/);
+        for (const hSect of headerSplit) {
+          if (hSect.length > TARGET_CHUNK_SIZE) {
+            // Split very long sections on single newlines as last resort
+            const lines = hSect.split('\n');
+            let chunk = '';
+            for (const line of lines) {
+              if (chunk.length + line.length > TARGET_CHUNK_SIZE && chunk.length > 100) {
+                sections.push(chunk.trim());
+                chunk = line;
+              } else {
+                chunk += (chunk ? '\n' : '') + line;
+              }
+            }
+            if (chunk.trim().length > 30) sections.push(chunk.trim());
+          } else if (hSect.trim().length > 30) {
+            sections.push(hSect.trim());
+          }
+        }
+      } else if (raw.trim().length > 30) {
+        sections.push(raw.trim());
+      }
+    }
+
+    // Now group sections into chunks
+    const forcedChunks = [];
+    let currentGroup = [];
+    let currentLength = 0;
+    let groupNum = 1;
+
+    const flushGroup = () => {
+      if (currentGroup.length === 0) return;
+      const firstLine = currentGroup[0].split('\n')[0].trim();
+      const title = firstLine.startsWith('#')
+        ? firstLine.replace(/^#+\s*/, '').slice(0, 60)
+        : `${fileName} Section ${groupNum}`;
+
+      forcedChunks.push({
+        title,
+        content: currentGroup.join('\n\n'),
+        level: 2,
+        headerLine: 0,
+        isForced: true,
+        forceNum: groupNum
+      });
+      groupNum++;
+      currentGroup = [];
+      currentLength = 0;
+    };
+
+    for (const section of sections) {
+      if (currentLength + section.length > TARGET_CHUNK_SIZE && currentGroup.length > 0) {
+        flushGroup();
+      }
+      currentGroup.push(section);
+      currentLength += section.length;
+    }
+    flushGroup();
+
+    // Always use force-chunked results if we got multiple chunks
+    if (forcedChunks.length >= 2) {
+      debug(`  Force-chunking produced ${forcedChunks.length} chunks (was ${finalChunks.length})`);
+      return forcedChunks;
+    }
+  }
+
+  return finalChunks;
+}
+
+/**
+ * Build hierarchical relationships between chunks
+ * H2 chunks are parents of subsequent H3 chunks
+ */
+function buildHierarchy(chunks, chunkPrefix) {
+  const hierarchy = {};
+  let currentH2Index = null;
+
+  for (let i = 0; i < chunks.length; i++) {
+    const chunk = chunks[i];
+    const chunkKey = `${chunkPrefix}-${i}`;
+
+    hierarchy[chunkKey] = {
+      parent: null,
+      children: []
+    };
+
+    if (chunk.level === 2) {
+      currentH2Index = i;
+    } else if (chunk.level === 3 && currentH2Index !== null) {
+      const parentKey = `${chunkPrefix}-${currentH2Index}`;
+      hierarchy[chunkKey].parent = parentKey;
+      hierarchy[parentKey].children.push(chunkKey);
+    }
+  }
+
+  return hierarchy;
+}
+
+function indexFile(db, filePath, keyPrefix) {
+  const fileName = basename(filePath, extname(filePath));
+  const docKey = `doc-${keyPrefix}-${fileName}`;
+  const chunkPrefix = `chunk-${keyPrefix}-${fileName}`;
+
+  try {
+    const content = readFileSync(filePath, 'utf-8');
+    const contentHash = hashContent(content);
+
+    // Check if content changed (skip if same hash unless --force)
+    if (!force) {
+      const existingHash = getEntryHash(db, docKey);
+      if (existingHash === contentHash) {
+        return { docKey, status: 'unchanged', chunks: 0 };
+      }
+    }
+
+    const stats = statSync(filePath);
+    const relativePath = filePath.replace(projectRoot, '').replace(/\\/g, '/');
+
+    // Delete old chunks for this file before re-indexing
+    deleteByPrefix(db, chunkPrefix);
+
+    // 1. Store full document
+    const docMetadata = {
+      type: 'document',
+      filePath: relativePath,
+      fileSize: stats.size,
+      lastModified: stats.mtime.toISOString(),
+      contentHash,
+      indexedAt: new Date().toISOString(),
+      ragVersion: '2.0',  // Mark as full RAG indexed
+    };
+
+    storeEntry(db, docKey, content, docMetadata, [keyPrefix, 'document']);
+    debug(`Stored document: ${docKey}`);
+
+    // 2. Chunk and store semantic pieces with full RAG linking
+    const chunks = chunkMarkdown(content, fileName);
+
+    if (chunks.length === 0) {
+      return { docKey, status: 'indexed', chunks: 0 };
+    }
+
+    // Build hierarchy and sibling list
+    const hierarchy = buildHierarchy(chunks, chunkPrefix);
+    const siblings = chunks.map((_, i) => `${chunkPrefix}-${i}`);
+
+    // Update document with children references
+    const docChildrenMeta = {
+      ...docMetadata,
+      children: siblings,
+      chunkCount: chunks.length,
+    };
+    storeEntry(db, docKey, content, docChildrenMeta, [keyPrefix, 'document']);
+
+    for (let i = 0; i < chunks.length; i++) {
+      const chunk = chunks[i];
+      const chunkKey = `${chunkPrefix}-${i}`;
+
+      // Build prev/next links
+      const prevChunk = i > 0 ? `${chunkPrefix}-${i - 1}` : null;
+      const nextChunk = i < chunks.length - 1 ? `${chunkPrefix}-${i + 1}` : null;
+
+      // Extract overlapping context from adjacent chunks
+      const contextBefore = i > 0
+        ? extractOverlapContext(chunks[i - 1].content, overlapPercent, 'end')
+        : null;
+      const contextAfter = i < chunks.length - 1
+        ? extractOverlapContext(chunks[i + 1].content, overlapPercent, 'start')
+        : null;
+
+      // Get hierarchical relationships
+      const hierInfo = hierarchy[chunkKey];
+
+      const chunkMetadata = {
+        type: 'chunk',
+        ragVersion: '2.0',
+
+        // Document relationship
+        parentDoc: docKey,
+        parentPath: relativePath,
+
+        // Sequential navigation (forward/backward links)
+        chunkIndex: i,
+        totalChunks: chunks.length,
+        prevChunk,
+        nextChunk,
+
+        // Sibling awareness
+        siblings,
+
+        // Hierarchical relationships (h2 -> h3)
+        hierarchicalParent: hierInfo.parent,
+        hierarchicalChildren: hierInfo.children.length > 0 ? hierInfo.children : null,
+
+        // Chunk info
+        chunkTitle: chunk.title,
+        headerLevel: chunk.level,
+        headerLine: chunk.headerLine,
+        isPart: chunk.isPart || false,
+        partNum: chunk.partNum || null,
+
+        // Overlapping context for continuity
+        contextOverlapPercent: overlapPercent,
+        hasContextBefore: !!contextBefore,
+        hasContextAfter: !!contextAfter,
+
+        // Content metadata
+        contentLength: chunk.content.length,
+        contentHash: hashContent(chunk.content),
+        indexedAt: new Date().toISOString(),
+      };
+
+      // Build searchable content with title context
+      // Include overlap context for better retrieval
+      let searchableContent = `# ${chunk.title}\n\n`;
+
+      if (contextBefore) {
+        searchableContent += `[Context from previous section:]\n${contextBefore}\n\n---\n\n`;
+      }
+
+      searchableContent += chunk.content;
+
+      if (contextAfter) {
+        searchableContent += `\n\n---\n\n[Context from next section:]\n${contextAfter}`;
+      }
+
+      // Store chunk with full metadata
+      storeEntry(
+        db,
+        chunkKey,
+        searchableContent,
+        chunkMetadata,
+        [keyPrefix, 'chunk', `level-${chunk.level}`, chunk.title.toLowerCase().replace(/[^a-z0-9]+/g, '-')]
+      );
+
+      debug(`  Stored chunk ${i}: ${chunk.title} (${chunk.content.length} chars, prev=${!!prevChunk}, next=${!!nextChunk})`);
+    }
+
+    return { docKey, status: 'indexed', chunks: chunks.length };
+  } catch (err) {
+    return { docKey, status: 'error', error: err.message, chunks: 0 };
+  }
+}
+
+/**
+ * Recursively collect all .md files under a directory.
+ * Skips node_modules, .git, and other non-content directories.
+ */
+function walkMdFiles(dir) {
+  const SKIP_DIRS = new Set(['node_modules', '.git', 'dist', 'build', 'coverage', '.next', '.reports']);
+  // CLAUDE.md is loaded into context by Claude automatically — skip to avoid duplicate vectors
+  const SKIP_FILES = new Set(['CLAUDE.md']);
+  const files = [];
+
+  function walk(current) {
+    if (!existsSync(current)) return;
+    for (const entry of readdirSync(current, { withFileTypes: true })) {
+      if (entry.isDirectory()) {
+        if (!SKIP_DIRS.has(entry.name)) walk(resolve(current, entry.name));
+      } else if (entry.isFile() && entry.name.endsWith('.md') && !SKIP_FILES.has(entry.name)) {
+        files.push(resolve(current, entry.name));
+      }
+    }
+  }
+
+  walk(dir);
+  return files;
+}
+
+function indexDirectory(db, dirConfig) {
+  const dirPath = dirConfig.absolute ? dirConfig.path : resolve(projectRoot, dirConfig.path);
+  const results = [];
+
+  if (!existsSync(dirPath)) {
+    if (verbose) debug(`Directory not found: ${dirConfig.path}`);
+    return results;
+  }
+
+  const allMdFiles = walkMdFiles(dirPath);
+  const filtered = dirConfig.fileFilter
+    ? allMdFiles.filter(f => dirConfig.fileFilter.includes(basename(f)))
+    : allMdFiles;
+
+  for (const filePath of filtered) {
+    const result = indexFile(db, filePath, dirConfig.prefix);
+    results.push(result);
+  }
+
+  return results;
+}
+
+/**
+ * Remove stale entries for files that no longer exist on disk.
+ * Uses the set of docKeys seen during the current indexing run to determine
+ * which entries are stale, rather than reconstructing file paths from keys
+ * (which breaks for files in subdirectories).
+ */
+function cleanStaleEntries(db, currentDocKeys) {
+  const docsStmt = db.prepare(
+    `SELECT DISTINCT key FROM memory_entries WHERE namespace = ? AND key LIKE 'doc-%'`
+  );
+  docsStmt.bind([NAMESPACE]);
+  const docs = [];
+  while (docsStmt.step()) docs.push(docsStmt.getAsObject());
+  docsStmt.free();
+
+  let staleCount = 0;
+
+  for (const { key } of docs) {
+    // If this doc key was seen during the current indexing run, it's not stale
+    if (currentDocKeys.has(key)) continue;
+
+    const chunkPrefix = key.replace('doc-', 'chunk-');
+    const countBefore = db.exec(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = '${NAMESPACE}'`)[0]?.values[0][0] || 0;
+    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key LIKE ?`, [NAMESPACE, `${chunkPrefix}%`]);
+    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key = ?`, [NAMESPACE, key]);
+    const countAfter = db.exec(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = '${NAMESPACE}'`)[0]?.values[0][0] || 0;
+    const removed = countBefore - countAfter;
+    if (removed > 0) {
+      log(`  Removed ${removed} stale entries for deleted file: ${key}`);
+      staleCount += removed;
+    }
+  }
+
+  // Also clean any orphaned entries not matching doc-/chunk- patterns
+  const orphanStmt = db.prepare(
+    `SELECT key FROM memory_entries WHERE namespace = ? AND key NOT LIKE 'doc-%' AND key NOT LIKE 'chunk-%'`
+  );
+  orphanStmt.bind([NAMESPACE]);
+  const orphans = [];
+  while (orphanStmt.step()) orphans.push(orphanStmt.getAsObject());
+  orphanStmt.free();
+  for (const { key } of orphans) {
+    db.run(`DELETE FROM memory_entries WHERE namespace = ? AND key = ?`, [NAMESPACE, key]);
+    staleCount++;
+    log(`  Removed orphan entry: ${key}`);
+  }
+
+  return staleCount;
+}
+
+// Main
+console.log('');
+log('Indexing guidance files with FULL RAG linked segments...');
+log(`  Context overlap: ${overlapPercent}%`);
+log(`  Directories (${GUIDANCE_DIRS.length}):`);
+for (const d of GUIDANCE_DIRS) {
+  const dirPath = d.absolute ? d.path : resolve(projectRoot, d.path);
+  const exists = existsSync(dirPath);
+  log(`    ${exists ? '✓' : '✗'} ${d.absolute ? dirPath : d.path} [${d.prefix}]`);
+}
+console.log('');
+
+const db = await getDb();
+let docsIndexed = 0;
+let chunksIndexed = 0;
+let unchanged = 0;
+let errors = 0;
+const currentDocKeys = new Set();
+
+if (specificFile) {
+  // Index single file
+  const filePath = resolve(projectRoot, specificFile);
+  if (!existsSync(filePath)) {
+    log(`File not found: ${specificFile}`);
+    process.exit(1);
+  }
+
+  let prefix = 'docs';
+  if (specificFile.includes('.claude/guidance/')) {
+    prefix = 'guidance';
+  }
+
+  const result = indexFile(db, filePath, prefix);
+  log(`${result.docKey}: ${result.status} (${result.chunks} chunks)`);
+
+  if (result.status === 'indexed') {
+    docsIndexed++;
+    chunksIndexed += result.chunks;
+  } else if (result.status === 'unchanged') {
+    unchanged++;
+  } else {
+    errors++;
+  }
+} else {
+  // Index all directories
+  for (const dir of GUIDANCE_DIRS) {
+    log(`Scanning ${dir.path}/...`);
+    const results = indexDirectory(db, dir);
+
+    for (const result of results) {
+      if (result.status === 'indexed' || result.status === 'unchanged') {
+        currentDocKeys.add(result.docKey);
+      }
+      if (result.status === 'indexed') {
+        log(`  ✅ ${result.docKey} (${result.chunks} chunks)`);
+        docsIndexed++;
+        chunksIndexed += result.chunks;
+      } else if (result.status === 'unchanged') {
+        unchanged++;
+      } else {
+        log(`  ❌ ${result.docKey}: ${result.error}`);
+        errors++;
+      }
+    }
+  }
+}
+
+// Clean stale entries for deleted files (unless indexing a specific file)
+let staleRemoved = 0;
+if (!specificFile) {
+  log('Cleaning stale entries for deleted files...');
+  staleRemoved = cleanStaleEntries(db, currentDocKeys);
+  if (staleRemoved === 0) {
+    log('  No stale entries found');
+  }
+}
+
+// Write changes back to disk
+if (docsIndexed > 0 || chunksIndexed > 0 || staleRemoved > 0) saveDb(db);
+
+// Check for entries missing embeddings (e.g. prior background run failed)
+let missingEmbeddings = 0;
+{
+  const stmt = db.prepare(`SELECT COUNT(*) as cnt FROM memory_entries WHERE namespace = ? AND (embedding IS NULL OR embedding = '')`);
+  stmt.bind([NAMESPACE]);
+  if (stmt.step()) missingEmbeddings = stmt.getAsObject().cnt;
+  stmt.free();
+}
+db.close();
+
+console.log('');
+log('═══════════════════════════════════════════════════════════');
+log('  FULL RAG INDEXING COMPLETE');
+log('═══════════════════════════════════════════════════════════');
+log(`  Documents indexed:    ${docsIndexed}`);
+log(`  Chunks created:       ${chunksIndexed}`);
+log(`  Unchanged:            ${unchanged}`);
+log(`  Stale removed:        ${staleRemoved}`);
+log(`  Errors:               ${errors}`);
+log('');
+log('  RAG Features Enabled:');
+log(`    • Forward/backward links (prevChunk/nextChunk)`);
+log(`    • Sibling awareness (all chunks from same doc)`);
+log(`    • Hierarchical links (h2 -> h3 parent/children)`);
+log(`    • Context overlap: ${overlapPercent}% (contextBefore/contextAfter)`);
+log('═══════════════════════════════════════════════════════════');
+
+// Generate embeddings for new entries or backfill missing ones
+// Runs in BACKGROUND to avoid blocking startup
+const needsEmbeddings = (docsIndexed > 0 || chunksIndexed > 0 || missingEmbeddings > 0);
+if (!skipEmbeddings && needsEmbeddings) {
+  if (missingEmbeddings > 0 && docsIndexed === 0 && chunksIndexed === 0) {
+    log(`${missingEmbeddings} entries missing embeddings — backfilling...`);
+  }
+  console.log('');
+  log('Spawning embedding generation in background...');
+
+  const { spawn } = await import('child_process');
+
+  // Look for build-embeddings script in multiple locations:
+  // 1. Shipped with moflo (node_modules/moflo/bin/)
+  // 2. Project-local (.claude/scripts/)
+  const mofloScript = resolve(__dirname, 'build-embeddings.mjs');
+  const projectLocalScript = resolve(projectRoot, '.claude/scripts/build-embeddings.mjs');
+  const embeddingScript = existsSync(mofloScript) ? mofloScript : projectLocalScript;
+
+  if (existsSync(embeddingScript)) {
+    const embeddingArgs = ['--namespace', NAMESPACE];
+
+    // Create log file for background process output
+    const logDir = resolve(projectRoot, '.swarm/logs');
+    if (!existsSync(logDir)) {
+      mkdirSync(logDir, { recursive: true });
+    }
+    const logFile = resolve(logDir, 'embeddings.log');
+    const { openSync } = await import('fs');
+    const out = openSync(logFile, 'a');
+    const err = openSync(logFile, 'a');
+
+    // Spawn in background - don't wait for completion
+    const proc = spawn('node', [embeddingScript, ...embeddingArgs], {
+      stdio: ['ignore', out, err],
+      cwd: projectRoot,
+      detached: true,
+      windowsHide: true  // Suppress command windows on Windows
+    });
+    proc.unref();  // Allow parent to exit independently
+
+    log(`Background embedding started (PID: ${proc.pid})`);
+    log(`Log file: .swarm/logs/embeddings.log`);
+  } else {
+    log('⚠️  Embedding script not found, skipping embedding generation');
+  }
+} else if (skipEmbeddings) {
+  log('Skipping embedding generation (--no-embeddings)');
+} else {
+  log('No new content indexed, skipping embedding generation');
+}
+
+if (errors > 0) {
+  process.exit(1);
+}