@tekmidian/pai 0.66.0 → 0.66.2

package/dist/async-C2Bm_Lal.mjs.map

@@ -1 +0,0 @@
-{"version":3,"file":"async-C2Bm_Lal.mjs","names":[],"sources":["../src/memory/indexer/async.ts"],"sourcesContent":["/**\n * Backend-aware async indexer for PAI federation memory.\n *\n * Provides the same functionality as sync.ts but writes through the\n * StorageBackend interface instead of directly to the SQLite driver.\n * Used when the daemon is configured with the Postgres backend.\n *\n * The SQLite path still uses sync.ts directly (which is faster for SQLite\n * due to synchronous transactions).\n */\n\nimport { readFileSync, statSync, existsSync } from \"node:fs\";\nimport { join, relative, basename } from \"node:path\";\nimport type { StorageBackend, ChunkRow } from \"../../storage/interface.js\";\nimport type { RegistryBackend } from \"../../storage/registry-interface.js\";\nimport { chunkMarkdown, fileContentHash } from \"../chunker.js\";\nimport {\n  sha256File,\n  chunkId,\n  detectTier,\n  walkMdFiles,\n  walkContentFiles,\n  isPathTooBroadForContentScan,\n  parseSessionTitleChunk,\n  yieldToEventLoop,\n  INDEX_YIELD_EVERY,\n} from \"./helpers.js\";\nimport type { IndexResult } from \"./types.js\";\n\nexport type { IndexResult };\n\n// ---------------------------------------------------------------------------\n// Single-file indexing via StorageBackend\n// ---------------------------------------------------------------------------\n\n/**\n * Index a single file through the StorageBackend interface.\n * Returns true if the file was re-indexed (changed or new), false if skipped.\n */\nexport async function indexFileWithBackend(\n  backend: StorageBackend,\n  projectId: number,\n  rootPath: string,\n  relativePath: string,\n  source: string,\n  tier: string,\n): Promise<boolean> {\n  const absPath = join(rootPath, relativePath);\n\n  let content: string;\n  let stat: ReturnType<typeof statSync>;\n  try {\n    content = readFileSync(absPath, \"utf8\");\n    stat = statSync(absPath);\n  } catch {\n    return false;\n  }\n\n  const hash = fileContentHash(content);\n  const mtime = Math.floor(stat.mtimeMs);\n  const size = stat.size;\n\n  // Change detection\n  const existingHash = await backend.getFileHash(projectId, relativePath);\n  if (existingHash === hash) return false;\n\n  // Delete old chunks\n  await backend.deleteChunksForFile(projectId, relativePath);\n\n  // Chunk the content\n  const rawChunks = chunkMarkdown(content);\n  const updatedAt = Date.now();\n\n  const chunks: ChunkRow[] = rawChunks.map((c, i) => ({\n    id: chunkId(projectId, relativePath, i, c.startLine, c.endLine),\n    projectId,\n    source,\n    tier,\n    path: relativePath,\n    startLine: c.startLine,\n    endLine: c.endLine,\n    hash: c.hash,\n    text: c.text,\n    updatedAt,\n    embedding: null,\n  }));\n\n  // Insert chunks + update file record\n  await backend.insertChunks(chunks);\n  await backend.upsertFile({ projectId, path: relativePath, source, tier, hash, mtime, size });\n\n  return true;\n}\n\n// ---------------------------------------------------------------------------\n// Project-level indexing via StorageBackend\n// ---------------------------------------------------------------------------\n\nexport async function indexProjectWithBackend(\n  backend: StorageBackend,\n  projectId: number,\n  rootPath: string,\n  claudeNotesDir?: string | null,\n): Promise<IndexResult> {\n  const result: IndexResult = { filesProcessed: 0, chunksCreated: 0, filesSkipped: 0 };\n\n  const filesToIndex: Array<{ absPath: string; rootBase: string; source: string; tier: string }> = [];\n\n  const rootMemoryMd = join(rootPath, \"MEMORY.md\");\n  if (existsSync(rootMemoryMd)) {\n    filesToIndex.push({ absPath: rootMemoryMd, rootBase: rootPath, source: \"memory\", tier: \"evergreen\" });\n  }\n\n  const memoryDir = join(rootPath, \"memory\");\n  for (const absPath of walkMdFiles(memoryDir)) {\n    const relPath = relative(rootPath, absPath);\n    const tier = detectTier(relPath);\n    filesToIndex.push({ absPath, rootBase: rootPath, source: \"memory\", tier });\n  }\n\n  const notesDir = join(rootPath, \"Notes\");\n  for (const absPath of walkMdFiles(notesDir)) {\n    filesToIndex.push({ absPath, rootBase: rootPath, source: \"notes\", tier: \"session\" });\n  }\n\n  // Synthetic session-title chunks for Notes files\n  {\n    const updatedAt = Date.now();\n    for (const absPath of walkMdFiles(notesDir)) {\n      const fileName = basename(absPath);\n      const text = parseSessionTitleChunk(fileName);\n      if (!text) continue;\n      const relPath = relative(rootPath, absPath);\n      const syntheticPath = `${relPath}::title`;\n      const id = chunkId(projectId, syntheticPath, 0, 0, 0);\n      const hash = sha256File(text);\n      const titleChunk: ChunkRow = {\n        id, projectId, source: \"notes\", tier: \"session\",\n        path: syntheticPath, startLine: 0, endLine: 0,\n        hash, text, updatedAt, embedding: null,\n      };\n      try {\n        await backend.insertChunks([titleChunk]);\n      } catch {\n        // Skip title chunks that cause backend errors\n      }\n    }\n  }\n\n  if (!isPathTooBroadForContentScan(rootPath)) {\n    for (const absPath of walkContentFiles(rootPath)) {\n      filesToIndex.push({ absPath, rootBase: rootPath, source: \"content\", tier: \"topic\" });\n    }\n  }\n\n  if (claudeNotesDir && claudeNotesDir !== notesDir) {\n    for (const absPath of walkMdFiles(claudeNotesDir)) {\n      filesToIndex.push({ absPath, rootBase: claudeNotesDir, source: \"notes\", tier: \"session\" });\n    }\n\n    // Synthetic title chunks for claude notes dir\n    {\n      const updatedAt = Date.now();\n      for (const absPath of walkMdFiles(claudeNotesDir)) {\n        const fileName = basename(absPath);\n        const text = parseSessionTitleChunk(fileName);\n        if (!text) continue;\n        const relPath = relative(claudeNotesDir, absPath);\n        const syntheticPath = `${relPath}::title`;\n        const id = chunkId(projectId, syntheticPath, 0, 0, 0);\n        const hash = sha256File(text);\n        const titleChunk: ChunkRow = {\n          id, projectId, source: \"notes\", tier: \"session\",\n          path: syntheticPath, startLine: 0, endLine: 0,\n          hash, text, updatedAt, embedding: null,\n        };\n        try {\n          await backend.insertChunks([titleChunk]);\n        } catch {\n          // Skip title chunks that cause backend errors\n        }\n      }\n    }\n\n    if (claudeNotesDir.endsWith(\"/Notes\")) {\n      const claudeProjectDir = claudeNotesDir.slice(0, -\"/Notes\".length);\n      const claudeMemoryMd = join(claudeProjectDir, \"MEMORY.md\");\n      if (existsSync(claudeMemoryMd)) {\n        filesToIndex.push({ absPath: claudeMemoryMd, rootBase: claudeProjectDir, source: \"memory\", tier: \"evergreen\" });\n      }\n      const claudeMemoryDir = join(claudeProjectDir, \"memory\");\n      for (const absPath of walkMdFiles(claudeMemoryDir)) {\n        const relPath = relative(claudeProjectDir, absPath);\n        const tier = detectTier(relPath);\n        filesToIndex.push({ absPath, rootBase: claudeProjectDir, source: \"memory\", tier });\n      }\n    }\n  }\n\n  await yieldToEventLoop();\n\n  let filesSinceYield = 0;\n\n  for (const { absPath, rootBase, source, tier } of filesToIndex) {\n    if (filesSinceYield >= INDEX_YIELD_EVERY) {\n      await yieldToEventLoop();\n      filesSinceYield = 0;\n    }\n    filesSinceYield++;\n\n    const relPath = relative(rootBase, absPath);\n    try {\n      const changed = await indexFileWithBackend(backend, projectId, rootBase, relPath, source, tier);\n\n      if (changed) {\n        const ids = await backend.getChunkIds(projectId, relPath);\n        result.filesProcessed++;\n        result.chunksCreated += ids.length;\n      } else {\n        result.filesSkipped++;\n      }\n    } catch {\n      // Skip files that cause backend errors (e.g. null bytes in Postgres)\n      result.filesSkipped++;\n    }\n  }\n\n  // Prune stale paths\n  const livePaths = new Set<string>();\n  for (const { absPath, rootBase } of filesToIndex) {\n    livePaths.add(relative(rootBase, absPath));\n  }\n\n  const dbChunkPaths = await backend.getDistinctChunkPaths(projectId);\n\n  const stalePaths: string[] = [];\n  for (const p of dbChunkPaths) {\n    const basePath = p.endsWith(\"::title\") ? p.slice(0, -\"::title\".length) : p;\n    if (!livePaths.has(basePath)) {\n      stalePaths.push(p);\n    }\n  }\n\n  if (stalePaths.length > 0) {\n    await backend.deletePaths(projectId, stalePaths);\n  }\n\n  return result;\n}\n\n// ---------------------------------------------------------------------------\n// Embedding generation via StorageBackend\n// ---------------------------------------------------------------------------\n\nconst EMBED_BATCH_SIZE = 50;\nconst EMBED_YIELD_EVERY = 1;\n\n/** Default ceiling on one pass. See EmbedPassOptions. */\nconst DEFAULT_MAX_CHUNKS_PER_PASS = 5_000;\n/** Default wall-clock ceiling on one pass, in milliseconds. */\nconst DEFAULT_MAX_MILLIS_PER_PASS = 120_000;\n\nexport interface EmbedPassOptions {\n  /**\n   * Stop the pass after this many chunks. Unbounded passes are the reason a\n   * six-figure backlog starves the indexer: the daemon serialises indexing and\n   * embedding against each other, so a pass that runs for hours means nothing\n   * new gets indexed for hours. Bounding the pass costs nothing — every\n   * embedding is written as it is produced, so the next pass resumes exactly\n   * where this one stopped.\n   */\n  maxChunks?: number;\n  /** Stop the pass after this much wall-clock time, whichever bound hits first. */\n  maxMillis?: number;\n}\n\n/**\n * Generate and store embeddings for unembedded chunks via the StorageBackend.\n *\n * The pass is deliberately bounded (see EmbedPassOptions) and resumable: it\n * takes a slice of the backlog, embeds it, and returns so the scheduler can run\n * an index pass before coming back. Draining the whole backlog in one call\n * starves indexing for as long as the call runs.\n *\n * The optional `shouldStop` callback is checked between every batch. When it\n * returns true the embed loop exits early so the caller (e.g. the daemon\n * shutdown handler) can close the pool without racing against active queries.\n *\n * Returns the number of newly embedded chunks.\n */\nexport async function embedChunksWithBackend(\n  backend: StorageBackend,\n  shouldStop?: () => boolean,\n  projectNames?: Map<number, string>,\n  options?: EmbedPassOptions,\n): Promise<number> {\n  const { generateEmbeddings, serializeEmbedding } = await import(\"../embeddings.js\");\n\n  const maxChunks = options?.maxChunks ?? DEFAULT_MAX_CHUNKS_PER_PASS;\n  const maxMillis = options?.maxMillis ?? DEFAULT_MAX_MILLIS_PER_PASS;\n  const deadline = Date.now() + maxMillis;\n\n  const rows = await backend.getUnembeddedChunkIds(undefined, maxChunks);\n  if (rows.length === 0) return 0;\n\n  const total = rows.length;\n  let embedded = 0;\n\n  // Build a summary of what needs embedding: count chunks per project_id\n  const projectChunkCounts = new Map<number, { count: number; samplePath: string }>();\n  for (const row of rows) {\n    const entry = projectChunkCounts.get(row.project_id);\n    if (entry) {\n      entry.count++;\n    } else {\n      projectChunkCounts.set(row.project_id, { count: 1, samplePath: row.path });\n    }\n  }\n  const pName = (pid: number) => projectNames?.get(pid) ?? `project ${pid}`;\n  const projectSummary = Array.from(projectChunkCounts.entries())\n    .map(([pid, { count, samplePath }]) => `  ${pName(pid)}: ${count} chunks (e.g. ${samplePath})`)\n    .join(\"\\n\");\n  process.stderr.write(\n    `[pai-daemon] Embed pass: ${total} unembedded chunks across ${projectChunkCounts.size} project(s)\\n${projectSummary}\\n`\n  );\n\n  // Track current project for transition logging\n  let currentProjectId = -1;\n  let projectEmbedded = 0;\n\n  for (let i = 0; i < rows.length; i += EMBED_BATCH_SIZE) {\n    // Check cancellation between every batch before touching the pool again\n    if (shouldStop?.()) {\n      process.stderr.write(\n        `[pai-daemon] Embed pass cancelled after ${embedded}/${total} chunks (shutdown requested)\\n`\n      );\n      break;\n    }\n\n    // Yield the daemon back to the indexer rather than run past the deadline.\n    // The remaining rows are simply picked up by the next scheduled pass.\n    if (Date.now() >= deadline) {\n      process.stderr.write(\n        `[pai-daemon] Embed pass yielding after ${embedded}/${total} chunks (${maxMillis}ms budget spent); resuming next pass\\n`\n      );\n      break;\n    }\n\n    const batch = rows.slice(i, i + EMBED_BATCH_SIZE);\n\n    // Keep IPC responsive: the forward pass below is synchronous inside the\n    // model, so yield before entering it rather than between chunks.\n    await yieldToEventLoop();\n\n    const vecs = await generateEmbeddings(batch.map((r) => r.text));\n    // Issue the writes together rather than one round-trip at a time. Measured\n    // on this machine the model embeds 40-67 chunks/s in isolation while the\n    // daemon managed ~5/s: the difference was one sequential UPDATE per chunk,\n    // so the pass spent most of its budget waiting on the network, not working.\n    // Concurrency is bounded by the batch size, which the pool handles; the\n    // SQLite backend is synchronous and simply ignores the difference.\n    await Promise.all(\n      batch.map((row, j) => backend.updateEmbedding(row.id, serializeEmbedding(vecs[j])))\n    );\n\n    // Attribute the batch to projects only after it is durably stored, and walk\n    // it in order so a batch straddling a project boundary credits each side\n    // correctly. Rows arrive ordered by project, so a boundary is a real\n    // transition, not the per-chunk flapping this logging used to produce.\n    for (const { project_id, path } of batch) {\n      if (project_id !== currentProjectId) {\n        if (currentProjectId !== -1) {\n          process.stderr.write(\n            `[pai-daemon] Finished ${pName(currentProjectId)}: ${projectEmbedded} chunks embedded\\n`\n          );\n        }\n        const info = projectChunkCounts.get(project_id);\n        process.stderr.write(\n          `[pai-daemon] Embedding ${pName(project_id)} (${info?.count ?? \"?\"} chunks, starting at ${path})\\n`\n        );\n        currentProjectId = project_id;\n        projectEmbedded = 0;\n      }\n      projectEmbedded++;\n    }\n\n    embedded += batch.length;\n\n    // Log progress with current file path for context\n    const lastChunk = batch[batch.length - 1];\n    process.stderr.write(\n      `[pai-daemon] Embedded ${embedded}/${total} chunks (${pName(lastChunk.project_id)}: ${lastChunk.path})\\n`\n    );\n  }\n\n  // Log final project completion\n  if (currentProjectId !== -1) {\n    process.stderr.write(\n      `[pai-daemon] Finished ${pName(currentProjectId)}: ${projectEmbedded} chunks embedded\\n`\n    );\n  }\n\n  return embedded;\n}\n\n// ---------------------------------------------------------------------------\n// Global indexing via StorageBackend\n// ---------------------------------------------------------------------------\n\nexport async function indexAllWithBackend(\n  backend: StorageBackend,\n  registry: RegistryBackend,\n): Promise<{ projects: number; result: IndexResult }> {\n  const projects = await registry.listProjects({ status: \"active\" });\n\n  const totals: IndexResult = { filesProcessed: 0, chunksCreated: 0, filesSkipped: 0 };\n\n  for (const project of projects) {\n    await yieldToEventLoop();\n    const r = await indexProjectWithBackend(backend, project.id, project.root_path, project.claude_notes_dir);\n    totals.filesProcessed += r.filesProcessed;\n    totals.chunksCreated += r.chunksCreated;\n    totals.filesSkipped += r.filesSkipped;\n  }\n\n  return { projects: projects.length, result: totals };\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAuCA,eAAsB,qBACpB,SACA,WACA,UACA,cACA,QACA,MACkB;CAClB,MAAM,UAAU,KAAK,UAAU,aAAa;CAE5C,IAAI;CACJ,IAAI;AACJ,KAAI;AACF,YAAU,aAAa,SAAS,OAAO;AACvC,SAAO,SAAS,QAAQ;SAClB;AACN,SAAO;;CAGT,MAAM,OAAO,gBAAgB,QAAQ;CACrC,MAAM,QAAQ,KAAK,MAAM,KAAK,QAAQ;CACtC,MAAM,OAAO,KAAK;AAIlB,KADqB,MAAM,QAAQ,YAAY,WAAW,aAAa,KAClD,KAAM,QAAO;AAGlC,OAAM,QAAQ,oBAAoB,WAAW,aAAa;CAG1D,MAAM,YAAY,cAAc,QAAQ;CACxC,MAAM,YAAY,KAAK,KAAK;CAE5B,MAAM,SAAqB,UAAU,KAAK,GAAG,OAAO;EAClD,IAAI,QAAQ,WAAW,cAAc,GAAG,EAAE,WAAW,EAAE,QAAQ;EAC/D;EACA;EACA;EACA,MAAM;EACN,WAAW,EAAE;EACb,SAAS,EAAE;EACX,MAAM,EAAE;EACR,MAAM,EAAE;EACR;EACA,WAAW;EACZ,EAAE;AAGH,OAAM,QAAQ,aAAa,OAAO;AAClC,OAAM,QAAQ,WAAW;EAAE;EAAW,MAAM;EAAc;EAAQ;EAAM;EAAM;EAAO;EAAM,CAAC;AAE5F,QAAO;;AAOT,eAAsB,wBACpB,SACA,WACA,UACA,gBACsB;CACtB,MAAM,SAAsB;EAAE,gBAAgB;EAAG,eAAe;EAAG,cAAc;EAAG;CAEpF,MAAM,eAA2F,EAAE;CAEnG,MAAM,eAAe,KAAK,UAAU,YAAY;AAChD,KAAI,WAAW,aAAa,CAC1B,cAAa,KAAK;EAAE,SAAS;EAAc,UAAU;EAAU,QAAQ;EAAU,MAAM;EAAa,CAAC;CAGvG,MAAM,YAAY,KAAK,UAAU,SAAS;AAC1C,MAAK,MAAM,WAAW,YAAY,UAAU,EAAE;EAE5C,MAAM,OAAO,WADG,SAAS,UAAU,QAAQ,CACX;AAChC,eAAa,KAAK;GAAE;GAAS,UAAU;GAAU,QAAQ;GAAU;GAAM,CAAC;;CAG5E,MAAM,WAAW,KAAK,UAAU,QAAQ;AACxC,MAAK,MAAM,WAAW,YAAY,SAAS,CACzC,cAAa,KAAK;EAAE;EAAS,UAAU;EAAU,QAAQ;EAAS,MAAM;EAAW,CAAC;CAItF;EACE,MAAM,YAAY,KAAK,KAAK;AAC5B,OAAK,MAAM,WAAW,YAAY,SAAS,EAAE;GAE3C,MAAM,OAAO,uBADI,SAAS,QAAQ,CACW;AAC7C,OAAI,CAAC,KAAM;GAEX,MAAM,gBAAgB,GADN,SAAS,UAAU,QAAQ,CACV;GAGjC,MAAM,aAAuB;IAC3B,IAHS,QAAQ,WAAW,eAAe,GAAG,GAAG,EAAE;IAG/C;IAAW,QAAQ;IAAS,MAAM;IACtC,MAAM;IAAe,WAAW;IAAG,SAAS;IAC5C,MAJW,WAAW,KAAK;IAIrB;IAAM;IAAW,WAAW;IACnC;AACD,OAAI;AACF,UAAM,QAAQ,aAAa,CAAC,WAAW,CAAC;WAClC;;;AAMZ,KAAI,CAAC,6BAA6B,SAAS,CACzC,MAAK,MAAM,WAAW,iBAAiB,SAAS,CAC9C,cAAa,KAAK;EAAE;EAAS,UAAU;EAAU,QAAQ;EAAW,MAAM;EAAS,CAAC;AAIxF,KAAI,kBAAkB,mBAAmB,UAAU;AACjD,OAAK,MAAM,WAAW,YAAY,eAAe,CAC/C,cAAa,KAAK;GAAE;GAAS,UAAU;GAAgB,QAAQ;GAAS,MAAM;GAAW,CAAC;EAI5F;GACE,MAAM,YAAY,KAAK,KAAK;AAC5B,QAAK,MAAM,WAAW,YAAY,eAAe,EAAE;IAEjD,MAAM,OAAO,uBADI,SAAS,QAAQ,CACW;AAC7C,QAAI,CAAC,KAAM;IAEX,MAAM,gBAAgB,GADN,SAAS,gBAAgB,QAAQ,CAChB;IAGjC,MAAM,aAAuB;KAC3B,IAHS,QAAQ,WAAW,eAAe,GAAG,GAAG,EAAE;KAG/C;KAAW,QAAQ;KAAS,MAAM;KACtC,MAAM;KAAe,WAAW;KAAG,SAAS;KAC5C,MAJW,WAAW,KAAK;KAIrB;KAAM;KAAW,WAAW;KACnC;AACD,QAAI;AACF,WAAM,QAAQ,aAAa,CAAC,WAAW,CAAC;YAClC;;;AAMZ,MAAI,eAAe,SAAS,SAAS,EAAE;GACrC,MAAM,mBAAmB,eAAe,MAAM,GAAG,GAAiB;GAClE,MAAM,iBAAiB,KAAK,kBAAkB,YAAY;AAC1D,OAAI,WAAW,eAAe,CAC5B,cAAa,KAAK;IAAE,SAAS;IAAgB,UAAU;IAAkB,QAAQ;IAAU,MAAM;IAAa,CAAC;GAEjH,MAAM,kBAAkB,KAAK,kBAAkB,SAAS;AACxD,QAAK,MAAM,WAAW,YAAY,gBAAgB,EAAE;IAElD,MAAM,OAAO,WADG,SAAS,kBAAkB,QAAQ,CACnB;AAChC,iBAAa,KAAK;KAAE;KAAS,UAAU;KAAkB,QAAQ;KAAU;KAAM,CAAC;;;;AAKxF,OAAM,kBAAkB;CAExB,IAAI,kBAAkB;AAEtB,MAAK,MAAM,EAAE,SAAS,UAAU,QAAQ,UAAU,cAAc;AAC9D,MAAI,mBAAmB,mBAAmB;AACxC,SAAM,kBAAkB;AACxB,qBAAkB;;AAEpB;EAEA,MAAM,UAAU,SAAS,UAAU,QAAQ;AAC3C,MAAI;AAGF,OAFgB,MAAM,qBAAqB,SAAS,WAAW,UAAU,SAAS,QAAQ,KAAK,EAElF;IACX,MAAM,MAAM,MAAM,QAAQ,YAAY,WAAW,QAAQ;AACzD,WAAO;AACP,WAAO,iBAAiB,IAAI;SAE5B,QAAO;UAEH;AAEN,UAAO;;;CAKX,MAAM,4BAAY,IAAI,KAAa;AACnC,MAAK,MAAM,EAAE,SAAS,cAAc,aAClC,WAAU,IAAI,SAAS,UAAU,QAAQ,CAAC;CAG5C,MAAM,eAAe,MAAM,QAAQ,sBAAsB,UAAU;CAEnE,MAAM,aAAuB,EAAE;AAC/B,MAAK,MAAM,KAAK,cAAc;EAC5B,MAAM,WAAW,EAAE,SAAS,UAAU,GAAG,EAAE,MAAM,GAAG,GAAkB,GAAG;AACzE,MAAI,CAAC,UAAU,IAAI,SAAS,CAC1B,YAAW,KAAK,EAAE;;AAItB,KAAI,WAAW,SAAS,EACtB,OAAM,QAAQ,YAAY,WAAW,WAAW;AAGlD,QAAO;;AAOT,MAAM,mBAAmB;;AAIzB,MAAM,8BAA8B;;AAEpC,MAAM,8BAA8B;;;;;;;;;;;;;;;AA8BpC,eAAsB,uBACpB,SACA,YACA,cACA,SACiB;CACjB,MAAM,EAAE,oBAAoB,uBAAuB,MAAM,OAAO;CAEhE,MAAM,YAAY,SAAS,aAAa;CACxC,MAAM,YAAY,SAAS,aAAa;CACxC,MAAM,WAAW,KAAK,KAAK,GAAG;CAE9B,MAAM,OAAO,MAAM,QAAQ,sBAAsB,QAAW,UAAU;AACtE,KAAI,KAAK,WAAW,EAAG,QAAO;CAE9B,MAAM,QAAQ,KAAK;CACnB,IAAI,WAAW;CAGf,MAAM,qCAAqB,IAAI,KAAoD;AACnF,MAAK,MAAM,OAAO,MAAM;EACtB,MAAM,QAAQ,mBAAmB,IAAI,IAAI,WAAW;AACpD,MAAI,MACF,OAAM;MAEN,oBAAmB,IAAI,IAAI,YAAY;GAAE,OAAO;GAAG,YAAY,IAAI;GAAM,CAAC;;CAG9E,MAAM,SAAS,QAAgB,cAAc,IAAI,IAAI,IAAI,WAAW;CACpE,MAAM,iBAAiB,MAAM,KAAK,mBAAmB,SAAS,CAAC,CAC5D,KAAK,CAAC,KAAK,EAAE,OAAO,kBAAkB,KAAK,MAAM,IAAI,CAAC,IAAI,MAAM,gBAAgB,WAAW,GAAG,CAC9F,KAAK,KAAK;AACb,SAAQ,OAAO,MACb,4BAA4B,MAAM,4BAA4B,mBAAmB,KAAK,eAAe,eAAe,IACrH;CAGD,IAAI,mBAAmB;CACvB,IAAI,kBAAkB;AAEtB,MAAK,IAAI,IAAI,GAAG,IAAI,KAAK,QAAQ,KAAK,kBAAkB;AAEtD,MAAI,cAAc,EAAE;AAClB,WAAQ,OAAO,MACb,2CAA2C,SAAS,GAAG,MAAM,gCAC9D;AACD;;AAKF,MAAI,KAAK,KAAK,IAAI,UAAU;AAC1B,WAAQ,OAAO,MACb,0CAA0C,SAAS,GAAG,MAAM,WAAW,UAAU,wCAClF;AACD;;EAGF,MAAM,QAAQ,KAAK,MAAM,GAAG,IAAI,iBAAiB;AAIjD,QAAM,kBAAkB;EAExB,MAAM,OAAO,MAAM,mBAAmB,MAAM,KAAK,MAAM,EAAE,KAAK,CAAC;AAO/D,QAAM,QAAQ,IACZ,MAAM,KAAK,KAAK,MAAM,QAAQ,gBAAgB,IAAI,IAAI,mBAAmB,KAAK,GAAG,CAAC,CAAC,CACpF;AAMD,OAAK,MAAM,EAAE,YAAY,UAAU,OAAO;AACxC,OAAI,eAAe,kBAAkB;AACnC,QAAI,qBAAqB,GACvB,SAAQ,OAAO,MACb,yBAAyB,MAAM,iBAAiB,CAAC,IAAI,gBAAgB,oBACtE;IAEH,MAAM,OAAO,mBAAmB,IAAI,WAAW;AAC/C,YAAQ,OAAO,MACb,0BAA0B,MAAM,WAAW,CAAC,IAAI,MAAM,SAAS,IAAI,uBAAuB,KAAK,KAChG;AACD,uBAAmB;AACnB,sBAAkB;;AAEpB;;AAGF,cAAY,MAAM;EAGlB,MAAM,YAAY,MAAM,MAAM,SAAS;AACvC,UAAQ,OAAO,MACb,yBAAyB,SAAS,GAAG,MAAM,WAAW,MAAM,UAAU,WAAW,CAAC,IAAI,UAAU,KAAK,KACtG;;AAIH,KAAI,qBAAqB,GACvB,SAAQ,OAAO,MACb,yBAAyB,MAAM,iBAAiB,CAAC,IAAI,gBAAgB,oBACtE;AAGH,QAAO;;AAOT,eAAsB,oBACpB,SACA,UACoD;CACpD,MAAM,WAAW,MAAM,SAAS,aAAa,EAAE,QAAQ,UAAU,CAAC;CAElE,MAAM,SAAsB;EAAE,gBAAgB;EAAG,eAAe;EAAG,cAAc;EAAG;AAEpF,MAAK,MAAM,WAAW,UAAU;AAC9B,QAAM,kBAAkB;EACxB,MAAM,IAAI,MAAM,wBAAwB,SAAS,QAAQ,IAAI,QAAQ,WAAW,QAAQ,iBAAiB;AACzG,SAAO,kBAAkB,EAAE;AAC3B,SAAO,iBAAiB,EAAE;AAC1B,SAAO,gBAAgB,EAAE;;AAG3B,QAAO;EAAE,UAAU,SAAS;EAAQ,QAAQ;EAAQ"}

package/dist/auto-route-o0BOsXn7.mjs.map

@@ -1 +0,0 @@
-{"version":3,"file":"auto-route-o0BOsXn7.mjs","names":[],"sources":["../src/session/auto-route.ts"],"sourcesContent":["/**\n * Auto-route: automatic project routing suggestion on session start.\n *\n * Given a working directory (and optional conversation context), determine\n * which registered project the session belongs to.\n *\n * Strategy (in priority order):\n *   1. Path match   — exact or parent-directory match in the project registry\n *   2. Marker walk  — walk up from cwd looking for Notes/PAI.md, resolve slug\n *   3. Topic match  — BM25 keyword search against memory (requires context text)\n *\n * The function is stateless and works directly against the registry/storage\n * backends (no daemon required), making it fast and safe to call during\n * session startup.\n */\n\nimport type { StorageBackend } from \"../storage/interface.js\";\nimport type { RegistryBackend, Project } from \"../storage/registry-interface.js\";\nimport { resolve, dirname } from \"node:path\";\nimport { existsSync } from \"node:fs\";\nimport { readPaiMarker } from \"../registry/pai-marker.js\";\n\n// ---------------------------------------------------------------------------\n// Types\n// ---------------------------------------------------------------------------\n\nexport type AutoRouteMethod = \"path\" | \"marker\" | \"topic\";\n\nexport interface AutoRouteResult {\n  /** Project slug */\n  slug: string;\n  /** Human-readable project name */\n  display_name: string;\n  /** Absolute path to the project root */\n  root_path: string;\n  /** How the project was detected */\n  method: AutoRouteMethod;\n  /** Confidence [0,1]: 1.0 for path/marker matches, BM25 fraction for topic */\n  confidence: number;\n}\n\n// ---------------------------------------------------------------------------\n// Core function\n// ---------------------------------------------------------------------------\n\n/**\n * Determine which project a session should be routed to.\n *\n * @param registryBackend  Registry backend (projects, sessions, ...)\n * @param federation       Memory storage backend (needed only for topic fallback)\n * @param cwd               Working directory to detect from (defaults to process.cwd())\n * @param context           Optional conversation text for topic-based fallback\n * @returns                 Best project match, or null if nothing matched\n */\nexport async function autoRoute(\n  registryBackend: RegistryBackend,\n  federation: StorageBackend,\n  cwd?: string,\n  context?: string\n): Promise<AutoRouteResult | null> {\n  const target = resolve(cwd ?? process.cwd());\n\n  // -------------------------------------------------------------------------\n  // Strategy 1: Path match via registry\n  // -------------------------------------------------------------------------\n\n  const pathMatch = await findProjectByPath(registryBackend, target);\n\n  if (pathMatch) {\n    return {\n      slug: pathMatch.slug,\n      display_name: pathMatch.display_name,\n      root_path: pathMatch.root_path,\n      method: \"path\",\n      confidence: 1.0,\n    };\n  }\n\n  // -------------------------------------------------------------------------\n  // Strategy 2: PAI.md marker file walk\n  //\n  // Walk up from cwd, checking <dir>/Notes/PAI.md at each level.\n  // Once found, resolve the slug against the registry to get full project info.\n  // -------------------------------------------------------------------------\n\n  const markerResult = await findMarkerUpward(registryBackend, target);\n  if (markerResult) {\n    return markerResult;\n  }\n\n  // -------------------------------------------------------------------------\n  // Strategy 3: Topic detection (requires context text)\n  // -------------------------------------------------------------------------\n\n  if (context && context.trim().length > 0) {\n    // Lazy import to avoid bundler pulling in daemon/index.mjs at module load time\n    const { detectTopicShift } = await import(\"../topics/detector.js\");\n    const topicResult = await detectTopicShift(\n      registryBackend,\n      federation,\n      {\n        context,\n        threshold: 0.5, // Lower threshold for initial routing (vs shift detection)\n      }\n    );\n\n    if (topicResult.suggestedProject && topicResult.confidence > 0) {\n      // Look up the full project info from the registry\n      const projectRow = await registryBackend.getProjectBySlug(topicResult.suggestedProject);\n\n      if (projectRow && projectRow.status !== \"archived\") {\n        return {\n          slug: projectRow.slug,\n          display_name: projectRow.display_name,\n          root_path: projectRow.root_path,\n          method: \"topic\",\n          confidence: topicResult.confidence,\n        };\n      }\n    }\n  }\n\n  return null;\n}\n\n// ---------------------------------------------------------------------------\n// Path-match helper\n// ---------------------------------------------------------------------------\n\n/**\n * Exact or parent-directory match against every non-archived project,\n * longest root_path wins. Mirrors src/cli/commands/detect.ts's detectProject()\n * matching logic (that file also enriches with session stats, which autoRoute\n * does not need).\n */\nasync function findProjectByPath(\n  registryBackend: RegistryBackend,\n  target: string\n): Promise<Project | null> {\n  const projects = await registryBackend.listProjectsByPathLengthDesc({ excludeArchived: true });\n\n  for (const p of projects) {\n    const root = resolve(p.root_path);\n    if (target === root) return p;\n    if (target.startsWith(root + \"/\")) return p;\n  }\n\n  return null;\n}\n\n// ---------------------------------------------------------------------------\n// Marker walk helper\n// ---------------------------------------------------------------------------\n\n/**\n * Walk up the directory tree from `startDir`, checking each level for a\n * `Notes/PAI.md` file. If found, read the slug and look up the project.\n *\n * Stops at the filesystem root or after 20 levels (safety guard).\n */\nasync function findMarkerUpward(\n  registryBackend: RegistryBackend,\n  startDir: string\n): Promise<AutoRouteResult | null> {\n  let current = startDir;\n  let depth = 0;\n\n  while (depth < 20) {\n    const markerPath = `${current}/Notes/PAI.md`;\n\n    if (existsSync(markerPath)) {\n      const marker = readPaiMarker(current);\n\n      if (marker && marker.status !== \"archived\") {\n        // Resolve slug to full project info in the registry\n        const projectRow = await registryBackend.getProjectBySlug(marker.slug);\n\n        if (projectRow && projectRow.status !== \"archived\") {\n          return {\n            slug: projectRow.slug,\n            display_name: projectRow.display_name,\n            root_path: projectRow.root_path,\n            method: \"marker\",\n            confidence: 1.0,\n          };\n        }\n      }\n    }\n\n    const parent = dirname(current);\n    if (parent === current) break; // Reached filesystem root\n    current = parent;\n    depth++;\n  }\n\n  return null;\n}\n\n// ---------------------------------------------------------------------------\n// Format helpers\n// ---------------------------------------------------------------------------\n\n/**\n * Format an AutoRouteResult as a human-readable string for CLI output.\n */\nexport function formatAutoRoute(result: AutoRouteResult): string {\n  const lines: string[] = [\n    `slug:         ${result.slug}`,\n    `display_name: ${result.display_name}`,\n    `root_path:    ${result.root_path}`,\n    `method:       ${result.method}`,\n    `confidence:   ${(result.confidence * 100).toFixed(0)}%`,\n  ];\n  return lines.join(\"\\n\");\n}\n\n/**\n * Format an AutoRouteResult as JSON for machine consumption.\n */\nexport function formatAutoRouteJson(result: AutoRouteResult): string {\n  return JSON.stringify(result, null, 2);\n}\n"],"mappings":";;;;;;;;;;;;;;AAsDA,eAAsB,UACpB,iBACA,YACA,KACA,SACiC;CACjC,MAAM,SAAS,QAAQ,OAAO,QAAQ,KAAK,CAAC;CAM5C,MAAM,YAAY,MAAM,kBAAkB,iBAAiB,OAAO;AAElE,KAAI,UACF,QAAO;EACL,MAAM,UAAU;EAChB,cAAc,UAAU;EACxB,WAAW,UAAU;EACrB,QAAQ;EACR,YAAY;EACb;CAUH,MAAM,eAAe,MAAM,iBAAiB,iBAAiB,OAAO;AACpE,KAAI,aACF,QAAO;AAOT,KAAI,WAAW,QAAQ,MAAM,CAAC,SAAS,GAAG;EAExC,MAAM,EAAE,qBAAqB,MAAM,OAAO;EAC1C,MAAM,cAAc,MAAM,iBACxB,iBACA,YACA;GACE;GACA,WAAW;GACZ,CACF;AAED,MAAI,YAAY,oBAAoB,YAAY,aAAa,GAAG;GAE9D,MAAM,aAAa,MAAM,gBAAgB,iBAAiB,YAAY,iBAAiB;AAEvF,OAAI,cAAc,WAAW,WAAW,WACtC,QAAO;IACL,MAAM,WAAW;IACjB,cAAc,WAAW;IACzB,WAAW,WAAW;IACtB,QAAQ;IACR,YAAY,YAAY;IACzB;;;AAKP,QAAO;;;;;;;;AAaT,eAAe,kBACb,iBACA,QACyB;CACzB,MAAM,WAAW,MAAM,gBAAgB,6BAA6B,EAAE,iBAAiB,MAAM,CAAC;AAE9F,MAAK,MAAM,KAAK,UAAU;EACxB,MAAM,OAAO,QAAQ,EAAE,UAAU;AACjC,MAAI,WAAW,KAAM,QAAO;AAC5B,MAAI,OAAO,WAAW,OAAO,IAAI,CAAE,QAAO;;AAG5C,QAAO;;;;;;;;AAaT,eAAe,iBACb,iBACA,UACiC;CACjC,IAAI,UAAU;CACd,IAAI,QAAQ;AAEZ,QAAO,QAAQ,IAAI;AAGjB,MAAI,WAFe,GAAG,QAAQ,eAEJ,EAAE;GAC1B,MAAM,SAAS,cAAc,QAAQ;AAErC,OAAI,UAAU,OAAO,WAAW,YAAY;IAE1C,MAAM,aAAa,MAAM,gBAAgB,iBAAiB,OAAO,KAAK;AAEtE,QAAI,cAAc,WAAW,WAAW,WACtC,QAAO;KACL,MAAM,WAAW;KACjB,cAAc,WAAW;KACzB,WAAW,WAAW;KACtB,QAAQ;KACR,YAAY;KACb;;;EAKP,MAAM,SAAS,QAAQ,QAAQ;AAC/B,MAAI,WAAW,QAAS;AACxB,YAAU;AACV;;AAGF,QAAO;;;;;AAwBT,SAAgB,oBAAoB,QAAiC;AACnE,QAAO,KAAK,UAAU,QAAQ,MAAM,EAAE"}

package/dist/chunker-BH4i-F2b.mjs.map

@@ -1 +0,0 @@
-{"version":3,"file":"chunker-BH4i-F2b.mjs","names":[],"sources":["../src/utils/hash.ts","../src/memory/chunker.ts"],"sourcesContent":["/**\n * Shared hashing utilities. Centralises all SHA-256 usage so every module\n * obtains digests through the same function rather than inlining createHash.\n */\n\nimport { createHash } from \"node:crypto\";\n\n/**\n * Compute a SHA-256 hex digest of the given string.\n * Aliased as sha256File for compatibility with existing call-sites that use\n * that name to hash file contents.\n */\nexport function sha256(content: string): string {\n  return createHash(\"sha256\").update(content).digest(\"hex\");\n}\n\n/** Alias kept for backwards compatibility with memory/indexer call-sites. */\nexport const sha256File = sha256;\n","/**\n * Markdown text chunker for the PAI memory engine.\n *\n * Splits markdown files into overlapping text segments suitable for BM25\n * full-text indexing.  Respects heading boundaries where possible, falling\n * back to paragraph and sentence splitting when sections are large.\n */\n\nimport { sha256 } from \"../utils/hash.js\";\n\n/** Bump whenever chunk text or boundaries change; forces one re-chunk of every file. */\nexport const CHUNKER_VERSION = 2;\n\n/** File-level change-detection hash: content plus chunker version. */\nexport function fileContentHash(content: string, version: number = CHUNKER_VERSION): string {\n  return sha256(`chunker v${version}\\n${content}`);\n}\n\nexport interface Chunk {\n  text: string;\n  startLine: number;  // 1-indexed\n  endLine: number;    // 1-indexed, inclusive\n  hash: string;       // SHA-256 of text\n  /** Ancestor heading titles, outermost first, including the chunk's own section heading. */\n  headingPath: string[];\n}\n\nexport interface ChunkOptions {\n  /** Approximate maximum tokens per chunk. Default 400. */\n  maxTokens?: number;\n  /** Overlap in tokens from the previous chunk. Default 80. */\n  overlap?: number;\n}\n\nconst DEFAULT_MAX_TOKENS = 400;\nconst DEFAULT_OVERLAP = 80;\n\n/**\n * Approximate token count using a words * 1.3 heuristic.\n * Matches the OpenClaw estimate approach.\n */\nexport function estimateTokens(text: string): number {\n  const wordCount = text.split(/\\s+/).filter(Boolean).length;\n  return Math.ceil(wordCount * 1.3);\n}\n\n// sha256 imported from utils/hash.ts\n\n// ---------------------------------------------------------------------------\n// Heading parser (the only one — chunker and memory_outline both use it)\n// ---------------------------------------------------------------------------\n\nexport interface Heading {\n  level: number;  // 1-6\n  title: string;\n  line: number;   // 1-indexed\n}\n\n/**\n * Parse ATX headings (levels 1-6) from lines. Lines inside ``` or ~~~ fences\n * are never headings.\n */\nexport function parseHeadings(lines: string[]): Heading[] {\n  const headings: Heading[] = [];\n  let fence: { ch: string; len: number } | null = null;\n\n  for (let i = 0; i < lines.length; i++) {\n    const text = lines[i] ?? \"\";\n    const f = /^ {0,3}(`{3,}|~{3,})/.exec(text);\n    if (f) {\n      const marker = f[1]!;\n      if (!fence) fence = { ch: marker[0]!, len: marker.length };\n      else if (marker[0] === fence.ch && marker.length >= fence.len && !text.slice(f[0].length).trim()) fence = null;\n      continue;\n    }\n    if (fence) continue;\n    const h = /^ {0,3}(#{1,6})\\s+(.*?)(?:\\s+#+)?\\s*$/.exec(text);\n    if (h && h[2]) headings.push({ level: h[1]!.length, title: h[2], line: i + 1 });\n  }\n  return headings;\n}\n\n/** Heading path (outermost first) in effect at each heading, via a level stack. */\nfunction headingPaths(headings: Heading[]): Map<number, string[]> {\n  const paths = new Map<number, string[]>();\n  const stack: Heading[] = [];\n  for (const h of headings) {\n    while (stack.length > 0 && stack[stack.length - 1]!.level >= h.level) stack.pop();\n    stack.push(h);\n    paths.set(h.line, stack.map((x) => x.title));\n  }\n  return paths;\n}\n\nexport interface OutlineNode {\n  level: number;\n  title: string;\n  startLine: number;\n  endLine: number;\n  tokens: number;\n  children: OutlineNode[];\n}\n\n/**\n * Heading tree of a markdown file. A section ends on the line before the next\n * heading of the same or higher level (or at EOF) and includes its children.\n */\nexport function buildOutline(content: string): OutlineNode[] {\n  const lines = content.split(\"\\n\");\n  const last = lines[lines.length - 1] === \"\" ? lines.length - 1 : lines.length;\n  const headings = parseHeadings(lines);\n  const roots: OutlineNode[] = [];\n  const stack: OutlineNode[] = [];\n\n  headings.forEach((h, i) => {\n    const next = headings.slice(i + 1).find((n) => n.level <= h.level);\n    const endLine = Math.max(h.line, next ? next.line - 1 : last);\n    const node: OutlineNode = {\n      level: h.level,\n      title: h.title,\n      startLine: h.line,\n      endLine,\n      tokens: estimateTokens(lines.slice(h.line - 1, endLine).join(\"\\n\")),\n      children: [],\n    };\n    while (stack.length > 0 && stack[stack.length - 1]!.level >= h.level) stack.pop();\n    (stack.length > 0 ? stack[stack.length - 1]!.children : roots).push(node);\n    stack.push(node);\n  });\n  return roots;\n}\n\n// ---------------------------------------------------------------------------\n// Internal section / paragraph / sentence splitters\n// ---------------------------------------------------------------------------\n\n/**\n * A contiguous block of lines associated with an approximate token count.\n */\ninterface LineBlock {\n  lines: Array<{ text: string; lineNo: number }>;\n  tokens: number;\n  headingPath: string[];\n}\n\n/**\n * Split content into sections delimited by ## or ### headings.\n * Each section starts at its heading line (or at line 1 for a preamble).\n */\nfunction splitBySections(\n  lines: Array<{ text: string; lineNo: number }>,\n): LineBlock[] {\n  const sections: LineBlock[] = [];\n  let current: Array<{ text: string; lineNo: number }> = [];\n  let currentPath: string[] = [];\n\n  const headings = parseHeadings(lines.map((l) => l.text));\n  const paths = headingPaths(headings);\n  const splitAt = new Set(headings.filter((h) => h.level <= 3).map((h) => h.line));\n\n  const flush = () => {\n    const text = current.map((l) => l.text).join(\"\\n\");\n    sections.push({ lines: current, tokens: estimateTokens(text), headingPath: currentPath });\n    current = [];\n  };\n\n  for (const line of lines) {\n    if (splitAt.has(line.lineNo)) {\n      if (current.length > 0) flush();\n      currentPath = paths.get(line.lineNo) ?? [];\n    }\n    current.push(line);\n  }\n\n  if (current.length > 0) flush();\n\n  return sections;\n}\n\n/**\n * Split a LineBlock by double-newline paragraph boundaries.\n */\nfunction splitByParagraphs(block: LineBlock): LineBlock[] {\n  const paragraphs: LineBlock[] = [];\n  let current: Array<{ text: string; lineNo: number }> = [];\n\n  for (const line of block.lines) {\n    if (line.text.trim() === \"\" && current.length > 0) {\n      // Empty line — potential paragraph boundary\n      const text = current.map((l) => l.text).join(\"\\n\");\n      paragraphs.push({ lines: [...current], tokens: estimateTokens(text), headingPath: block.headingPath });\n      current = [];\n    } else {\n      current.push(line);\n    }\n  }\n\n  if (current.length > 0) {\n    const text = current.map((l) => l.text).join(\"\\n\");\n    paragraphs.push({ lines: current, tokens: estimateTokens(text), headingPath: block.headingPath });\n  }\n\n  return paragraphs.length > 0 ? paragraphs : [block];\n}\n\n/**\n * Split a LineBlock by sentence boundaries (. ! ?) when even paragraphs are\n * too large.  Works character-by-character within joined lines.\n */\nfunction splitBySentences(block: LineBlock, maxTokens: number): LineBlock[] {\n  const fullText = block.lines.map((l) => l.text).join(\" \");\n  // Very rough sentence split — split on '. ', '! ', '? ' followed by uppercase\n  const sentenceRe = /(?<=[.!?])\\s+(?=[A-Z\"'])/g;\n  const sentences = fullText.split(sentenceRe);\n\n  const result: LineBlock[] = [];\n  let accText = \"\";\n  // We can't recover exact line numbers inside a single oversized paragraph,\n  // so we approximate using the block's start/end lines distributed evenly.\n  const startLine = block.lines[0]?.lineNo ?? 1;\n  const endLine = block.lines[block.lines.length - 1]?.lineNo ?? startLine;\n  const totalLines = endLine - startLine + 1;\n  const linesPerSentence = Math.max(1, Math.floor(totalLines / Math.max(1, sentences.length)));\n\n  let sentenceIdx = 0;\n  let approxLine = startLine;\n\n  const flush = () => {\n    if (!accText.trim()) return;\n    const endApprox = Math.min(approxLine + linesPerSentence - 1, endLine);\n    result.push({\n      lines: [{ text: accText.trim(), lineNo: approxLine }],\n      tokens: estimateTokens(accText),\n      headingPath: block.headingPath,\n    });\n    approxLine = endApprox + 1;\n    accText = \"\";\n  };\n\n  for (const sentence of sentences) {\n    sentenceIdx++;\n    const candidateText = accText ? accText + \" \" + sentence : sentence;\n    if (estimateTokens(candidateText) > maxTokens && accText) {\n      flush();\n      accText = sentence;\n    } else {\n      accText = candidateText;\n    }\n  }\n  void sentenceIdx; // used only for iteration count\n  flush();\n\n  return result.length > 0 ? result : [block];\n}\n\n// ---------------------------------------------------------------------------\n// Overlap helper\n// ---------------------------------------------------------------------------\n\n/**\n * Extract the last `overlapTokens` worth of text from a list of previously\n * emitted chunks to prepend to the next chunk.\n */\nfunction buildOverlapPrefix(\n  lastChunk: { text: string; startLine: number; endLine: number } | undefined,\n  overlapTokens: number,\n): Array<{ text: string; lineNo: number }> {\n  if (overlapTokens <= 0 || !lastChunk) return [];\n\n  const lines = lastChunk.text.split(\"\\n\");\n  const kept: string[] = [];\n  let acc = 0;\n\n  for (let i = lines.length - 1; i >= 0; i--) {\n    const lineTokens = estimateTokens(lines[i] ?? \"\");\n    acc += lineTokens;\n    kept.unshift(lines[i] ?? \"\");\n    if (acc >= overlapTokens) break;\n  }\n\n  // Distribute overlap lines across the lastChunk's line range\n  const startLine = lastChunk.endLine - kept.length + 1;\n  return kept.map((text, idx) => ({ text, lineNo: Math.max(lastChunk.startLine, startLine + idx) }));\n}\n\n// ---------------------------------------------------------------------------\n// Public API\n// ---------------------------------------------------------------------------\n\n/**\n * Chunk a markdown file into overlapping segments for BM25 indexing.\n *\n * Strategy:\n *  1. Split by headings (##, ###) as natural boundaries.\n *  2. If a section exceeds maxTokens, split by paragraphs.\n *  3. If a paragraph still exceeds maxTokens, split by sentences.\n *  4. Apply overlap: each chunk includes the last `overlap` tokens from the\n *     previous chunk.\n */\n/**\n * Strip `<private>...</private>` blocks from content before indexing.\n * Content within these tags is excluded from memory — never stored or searched.\n */\nexport function stripPrivateTags(content: string): string {\n  return content.replace(/<private>[\\s\\S]*?<\\/private>/gi, \"\");\n}\n\n/** One breadcrumb line (\"[A > B]\\n\") for a non-empty heading path, else \"\". */\nfunction breadcrumb(path: string[]): string {\n  return path.length > 0 ? `[${path.join(\" > \")}]\\n` : \"\";\n}\n\nexport function chunkMarkdown(content: string, opts?: ChunkOptions): Chunk[] {\n  const maxTokens = opts?.maxTokens ?? DEFAULT_MAX_TOKENS;\n  const overlapTokens = opts?.overlap ?? DEFAULT_OVERLAP;\n\n  // Strip private content before indexing\n  content = stripPrivateTags(content);\n\n  if (!content.trim()) return [];\n\n  const rawLines = content.split(\"\\n\");\n  const lines: Array<{ text: string; lineNo: number }> = rawLines.map((text, idx) => ({\n    text,\n    lineNo: idx + 1, // 1-indexed\n  }));\n\n  // Step 1: section split\n  const sections = splitBySections(lines);\n\n  // Step 2 & 3: further split oversized sections. The breadcrumb line is\n  // prepended to every chunk, so it counts against the token budget.\n  const finalBlocks: LineBlock[] = [];\n  for (const section of sections) {\n    const budget = Math.max(1, maxTokens - estimateTokens(breadcrumb(section.headingPath)));\n    if (section.tokens <= budget) {\n      finalBlocks.push(section);\n      continue;\n    }\n    // Too big — split by paragraphs\n    const paras = splitByParagraphs(section);\n    for (const para of paras) {\n      if (para.tokens <= budget) {\n        finalBlocks.push(para);\n        continue;\n      }\n      // Still too big — split by sentences\n      const sentences = splitBySentences(para, budget);\n      finalBlocks.push(...sentences);\n    }\n  }\n\n  // Step 4: build final chunks with overlap\n  const chunks: Chunk[] = [];\n  let prev: { text: string; startLine: number; endLine: number } | undefined;\n\n  for (const block of finalBlocks) {\n    if (block.lines.length === 0) continue;\n\n    // Build overlap prefix from the previous chunk's raw (unprefixed) text\n    const overlapLines = buildOverlapPrefix(prev, overlapTokens);\n\n    // Combine overlap + block lines\n    const allLines = [...overlapLines, ...block.lines];\n    const raw = allLines.map((l) => l.text).join(\"\\n\").trim();\n\n    if (!raw) continue;\n\n    const startLine = block.lines[0]?.lineNo ?? 1;\n    const endLine = block.lines[block.lines.length - 1]?.lineNo ?? startLine;\n    const text = breadcrumb(block.headingPath) + raw;\n\n    prev = { text: raw, startLine, endLine };\n    chunks.push({\n      text,\n      startLine,\n      endLine,\n      hash: sha256(text),\n      headingPath: block.headingPath,\n    });\n  }\n\n  return chunks;\n}\n"],"mappings":";;;;;;;;;;;;AAYA,SAAgB,OAAO,SAAyB;AAC9C,QAAO,WAAW,SAAS,CAAC,OAAO,QAAQ,CAAC,OAAO,MAAM;;;AAI3D,MAAa,aAAa;;;;;;;;;;;;ACN1B,MAAa,kBAAkB;;AAG/B,SAAgB,gBAAgB,SAAiB,UAAkB,iBAAyB;AAC1F,QAAO,OAAO,YAAY,QAAQ,IAAI,UAAU;;AAmBlD,MAAM,qBAAqB;AAC3B,MAAM,kBAAkB;;;;;AAMxB,SAAgB,eAAe,MAAsB;CACnD,MAAM,YAAY,KAAK,MAAM,MAAM,CAAC,OAAO,QAAQ,CAAC;AACpD,QAAO,KAAK,KAAK,YAAY,IAAI;;;;;;AAmBnC,SAAgB,cAAc,OAA4B;CACxD,MAAM,WAAsB,EAAE;CAC9B,IAAI,QAA4C;AAEhD,MAAK,IAAI,IAAI,GAAG,IAAI,MAAM,QAAQ,KAAK;EACrC,MAAM,OAAO,MAAM,MAAM;EACzB,MAAM,IAAI,uBAAuB,KAAK,KAAK;AAC3C,MAAI,GAAG;GACL,MAAM,SAAS,EAAE;AACjB,OAAI,CAAC,MAAO,SAAQ;IAAE,IAAI,OAAO;IAAK,KAAK,OAAO;IAAQ;YACjD,OAAO,OAAO,MAAM,MAAM,OAAO,UAAU,MAAM,OAAO,CAAC,KAAK,MAAM,EAAE,GAAG,OAAO,CAAC,MAAM,CAAE,SAAQ;AAC1G;;AAEF,MAAI,MAAO;EACX,MAAM,IAAI,wCAAwC,KAAK,KAAK;AAC5D,MAAI,KAAK,EAAE,GAAI,UAAS,KAAK;GAAE,OAAO,EAAE,GAAI;GAAQ,OAAO,EAAE;GAAI,MAAM,IAAI;GAAG,CAAC;;AAEjF,QAAO;;;AAIT,SAAS,aAAa,UAA4C;CAChE,MAAM,wBAAQ,IAAI,KAAuB;CACzC,MAAM,QAAmB,EAAE;AAC3B,MAAK,MAAM,KAAK,UAAU;AACxB,SAAO,MAAM,SAAS,KAAK,MAAM,MAAM,SAAS,GAAI,SAAS,EAAE,MAAO,OAAM,KAAK;AACjF,QAAM,KAAK,EAAE;AACb,QAAM,IAAI,EAAE,MAAM,MAAM,KAAK,MAAM,EAAE,MAAM,CAAC;;AAE9C,QAAO;;;;;;AAgBT,SAAgB,aAAa,SAAgC;CAC3D,MAAM,QAAQ,QAAQ,MAAM,KAAK;CACjC,MAAM,OAAO,MAAM,MAAM,SAAS,OAAO,KAAK,MAAM,SAAS,IAAI,MAAM;CACvE,MAAM,WAAW,cAAc,MAAM;CACrC,MAAM,QAAuB,EAAE;CAC/B,MAAM,QAAuB,EAAE;AAE/B,UAAS,SAAS,GAAG,MAAM;EACzB,MAAM,OAAO,SAAS,MAAM,IAAI,EAAE,CAAC,MAAM,MAAM,EAAE,SAAS,EAAE,MAAM;EAClE,MAAM,UAAU,KAAK,IAAI,EAAE,MAAM,OAAO,KAAK,OAAO,IAAI,KAAK;EAC7D,MAAM,OAAoB;GACxB,OAAO,EAAE;GACT,OAAO,EAAE;GACT,WAAW,EAAE;GACb;GACA,QAAQ,eAAe,MAAM,MAAM,EAAE,OAAO,GAAG,QAAQ,CAAC,KAAK,KAAK,CAAC;GACnE,UAAU,EAAE;GACb;AACD,SAAO,MAAM,SAAS,KAAK,MAAM,MAAM,SAAS,GAAI,SAAS,EAAE,MAAO,OAAM,KAAK;AACjF,GAAC,MAAM,SAAS,IAAI,MAAM,MAAM,SAAS,GAAI,WAAW,OAAO,KAAK,KAAK;AACzE,QAAM,KAAK,KAAK;GAChB;AACF,QAAO;;;;;;AAoBT,SAAS,gBACP,OACa;CACb,MAAM,WAAwB,EAAE;CAChC,IAAI,UAAmD,EAAE;CACzD,IAAI,cAAwB,EAAE;CAE9B,MAAM,WAAW,cAAc,MAAM,KAAK,MAAM,EAAE,KAAK,CAAC;CACxD,MAAM,QAAQ,aAAa,SAAS;CACpC,MAAM,UAAU,IAAI,IAAI,SAAS,QAAQ,MAAM,EAAE,SAAS,EAAE,CAAC,KAAK,MAAM,EAAE,KAAK,CAAC;CAEhF,MAAM,cAAc;EAClB,MAAM,OAAO,QAAQ,KAAK,MAAM,EAAE,KAAK,CAAC,KAAK,KAAK;AAClD,WAAS,KAAK;GAAE,OAAO;GAAS,QAAQ,eAAe,KAAK;GAAE,aAAa;GAAa,CAAC;AACzF,YAAU,EAAE;;AAGd,MAAK,MAAM,QAAQ,OAAO;AACxB,MAAI,QAAQ,IAAI,KAAK,OAAO,EAAE;AAC5B,OAAI,QAAQ,SAAS,EAAG,QAAO;AAC/B,iBAAc,MAAM,IAAI,KAAK,OAAO,IAAI,EAAE;;AAE5C,UAAQ,KAAK,KAAK;;AAGpB,KAAI,QAAQ,SAAS,EAAG,QAAO;AAE/B,QAAO;;;;;AAMT,SAAS,kBAAkB,OAA+B;CACxD,MAAM,aAA0B,EAAE;CAClC,IAAI,UAAmD,EAAE;AAEzD,MAAK,MAAM,QAAQ,MAAM,MACvB,KAAI,KAAK,KAAK,MAAM,KAAK,MAAM,QAAQ,SAAS,GAAG;EAEjD,MAAM,OAAO,QAAQ,KAAK,MAAM,EAAE,KAAK,CAAC,KAAK,KAAK;AAClD,aAAW,KAAK;GAAE,OAAO,CAAC,GAAG,QAAQ;GAAE,QAAQ,eAAe,KAAK;GAAE,aAAa,MAAM;GAAa,CAAC;AACtG,YAAU,EAAE;OAEZ,SAAQ,KAAK,KAAK;AAItB,KAAI,QAAQ,SAAS,GAAG;EACtB,MAAM,OAAO,QAAQ,KAAK,MAAM,EAAE,KAAK,CAAC,KAAK,KAAK;AAClD,aAAW,KAAK;GAAE,OAAO;GAAS,QAAQ,eAAe,KAAK;GAAE,aAAa,MAAM;GAAa,CAAC;;AAGnG,QAAO,WAAW,SAAS,IAAI,aAAa,CAAC,MAAM;;;;;;AAOrD,SAAS,iBAAiB,OAAkB,WAAgC;CAI1E,MAAM,YAHW,MAAM,MAAM,KAAK,MAAM,EAAE,KAAK,CAAC,KAAK,IAAI,CAG9B,MADR,4BACyB;CAE5C,MAAM,SAAsB,EAAE;CAC9B,IAAI,UAAU;CAGd,MAAM,YAAY,MAAM,MAAM,IAAI,UAAU;CAC5C,MAAM,UAAU,MAAM,MAAM,MAAM,MAAM,SAAS,IAAI,UAAU;CAC/D,MAAM,aAAa,UAAU,YAAY;CACzC,MAAM,mBAAmB,KAAK,IAAI,GAAG,KAAK,MAAM,aAAa,KAAK,IAAI,GAAG,UAAU,OAAO,CAAC,CAAC;CAE5F,IAAI,cAAc;CAClB,IAAI,aAAa;CAEjB,MAAM,cAAc;AAClB,MAAI,CAAC,QAAQ,MAAM,CAAE;EACrB,MAAM,YAAY,KAAK,IAAI,aAAa,mBAAmB,GAAG,QAAQ;AACtE,SAAO,KAAK;GACV,OAAO,CAAC;IAAE,MAAM,QAAQ,MAAM;IAAE,QAAQ;IAAY,CAAC;GACrD,QAAQ,eAAe,QAAQ;GAC/B,aAAa,MAAM;GACpB,CAAC;AACF,eAAa,YAAY;AACzB,YAAU;;AAGZ,MAAK,MAAM,YAAY,WAAW;AAChC;EACA,MAAM,gBAAgB,UAAU,UAAU,MAAM,WAAW;AAC3D,MAAI,eAAe,cAAc,GAAG,aAAa,SAAS;AACxD,UAAO;AACP,aAAU;QAEV,WAAU;;AAId,QAAO;AAEP,QAAO,OAAO,SAAS,IAAI,SAAS,CAAC,MAAM;;;;;;AAW7C,SAAS,mBACP,WACA,eACyC;AACzC,KAAI,iBAAiB,KAAK,CAAC,UAAW,QAAO,EAAE;CAE/C,MAAM,QAAQ,UAAU,KAAK,MAAM,KAAK;CACxC,MAAM,OAAiB,EAAE;CACzB,IAAI,MAAM;AAEV,MAAK,IAAI,IAAI,MAAM,SAAS,GAAG,KAAK,GAAG,KAAK;EAC1C,MAAM,aAAa,eAAe,MAAM,MAAM,GAAG;AACjD,SAAO;AACP,OAAK,QAAQ,MAAM,MAAM,GAAG;AAC5B,MAAI,OAAO,cAAe;;CAI5B,MAAM,YAAY,UAAU,UAAU,KAAK,SAAS;AACpD,QAAO,KAAK,KAAK,MAAM,SAAS;EAAE;EAAM,QAAQ,KAAK,IAAI,UAAU,WAAW,YAAY,IAAI;EAAE,EAAE;;;;;;;;;;;;;;;;AAqBpG,SAAgB,iBAAiB,SAAyB;AACxD,QAAO,QAAQ,QAAQ,kCAAkC,GAAG;;;AAI9D,SAAS,WAAW,MAAwB;AAC1C,QAAO,KAAK,SAAS,IAAI,IAAI,KAAK,KAAK,MAAM,CAAC,OAAO;;AAGvD,SAAgB,cAAc,SAAiB,MAA8B;CAC3E,MAAM,YAAY,MAAM,aAAa;CACrC,MAAM,gBAAgB,MAAM,WAAW;AAGvC,WAAU,iBAAiB,QAAQ;AAEnC,KAAI,CAAC,QAAQ,MAAM,CAAE,QAAO,EAAE;CAS9B,MAAM,WAAW,gBAPA,QAAQ,MAAM,KAAK,CAC4B,KAAK,MAAM,SAAS;EAClF;EACA,QAAQ,MAAM;EACf,EAAE,CAGoC;CAIvC,MAAM,cAA2B,EAAE;AACnC,MAAK,MAAM,WAAW,UAAU;EAC9B,MAAM,SAAS,KAAK,IAAI,GAAG,YAAY,eAAe,WAAW,QAAQ,YAAY,CAAC,CAAC;AACvF,MAAI,QAAQ,UAAU,QAAQ;AAC5B,eAAY,KAAK,QAAQ;AACzB;;EAGF,MAAM,QAAQ,kBAAkB,QAAQ;AACxC,OAAK,MAAM,QAAQ,OAAO;AACxB,OAAI,KAAK,UAAU,QAAQ;AACzB,gBAAY,KAAK,KAAK;AACtB;;GAGF,MAAM,YAAY,iBAAiB,MAAM,OAAO;AAChD,eAAY,KAAK,GAAG,UAAU;;;CAKlC,MAAM,SAAkB,EAAE;CAC1B,IAAI;AAEJ,MAAK,MAAM,SAAS,aAAa;AAC/B,MAAI,MAAM,MAAM,WAAW,EAAG;EAO9B,MAAM,MADW,CAAC,GAHG,mBAAmB,MAAM,cAAc,EAGzB,GAAG,MAAM,MAAM,CAC7B,KAAK,MAAM,EAAE,KAAK,CAAC,KAAK,KAAK,CAAC,MAAM;AAEzD,MAAI,CAAC,IAAK;EAEV,MAAM,YAAY,MAAM,MAAM,IAAI,UAAU;EAC5C,MAAM,UAAU,MAAM,MAAM,MAAM,MAAM,SAAS,IAAI,UAAU;EAC/D,MAAM,OAAO,WAAW,MAAM,YAAY,GAAG;AAE7C,SAAO;GAAE,MAAM;GAAK;GAAW;GAAS;AACxC,SAAO,KAAK;GACV;GACA;GACA;GACA,MAAM,OAAO,KAAK;GAClB,aAAa,MAAM;GACpB,CAAC;;AAGJ,QAAO"}

package/dist/clusters-BCtD3fbe.mjs

@@ -1,169 +0,0 @@
-import { t as STOP_WORDS } from "./stop-words-BdQuaE9K.mjs";
-
-//#region src/graph/clusters.ts
-/**
-* Aggregate per-path observation type counts into cluster-level counts,
-* then pick the dominant type.
-*/
-function aggregateObservationTypes(paths, byPath) {
-	const counts = {};
-	for (const path of paths) {
-		const pathCounts = byPath.get(path);
-		if (!pathCounts) continue;
-		for (const [type, n] of Object.entries(pathCounts)) counts[type] = (counts[type] ?? 0) + n;
-	}
-	let dominant = "unknown";
-	let maxCount = 0;
-	for (const [type, n] of Object.entries(counts)) if (n > maxCount) {
-		maxCount = n;
-		dominant = type;
-	}
-	return {
-		dominant,
-		counts
-	};
-}
-const SKIP_PREFIXES = [
-	"Attachments/",
-	"🗓️ Daily Notes/",
-	"Copilot/copilot-conversations/",
-	"Z - Zettelkasten/Tweets/"
-];
-/**
-* Cluster vault notes by wikilink connectivity when embeddings aren't available.
-* Uses BFS to find connected components in the link graph, then picks the
-* largest components as clusters. Labels are derived from the most common
-* title words in each component.
-*/
-async function clusterByLinks(backend, lookbackDays, minSize, maxClusters) {
-	const now = Date.now();
-	const from = now - lookbackDays * 864e5;
-	const recentNotes = (await backend.getRecentVaultFiles(from)).filter((f) => f.vaultPath.endsWith(".md"));
-	const noteMap = /* @__PURE__ */ new Map();
-	for (const n of recentNotes) noteMap.set(n.vaultPath, {
-		title: n.title,
-		indexed_at: n.indexedAt
-	});
-	const adj = /* @__PURE__ */ new Map();
-	for (const path of noteMap.keys()) if (!adj.has(path)) adj.set(path, /* @__PURE__ */ new Set());
-	const linkGraph = await backend.getVaultLinkGraph();
-	for (const { source_path, target_path } of linkGraph) if (noteMap.has(source_path) && noteMap.has(target_path)) {
-		adj.get(source_path).add(target_path);
-		adj.get(target_path).add(source_path);
-	}
-	const degrees = [...adj.entries()].map(([p, s]) => ({
-		path: p,
-		degree: s.size
-	}));
-	degrees.sort((a, b) => b.degree - a.degree);
-	const hubThreshold = Math.max(10, degrees[Math.floor(degrees.length * .05)]?.degree ?? 10);
-	const hubNodes = /* @__PURE__ */ new Set();
-	for (const { path, degree } of degrees) if (degree >= hubThreshold) hubNodes.add(path);
-	else break;
-	for (const hub of hubNodes) adj.delete(hub);
-	for (const [, neighbors] of adj) for (const hub of hubNodes) neighbors.delete(hub);
-	const visited = /* @__PURE__ */ new Set();
-	const components = [];
-	for (const path of noteMap.keys()) {
-		if (visited.has(path) || hubNodes.has(path)) continue;
-		if (SKIP_PREFIXES.some((p) => path.startsWith(p))) {
-			visited.add(path);
-			continue;
-		}
-		const component = [];
-		const queue = [path];
-		visited.add(path);
-		while (queue.length > 0) {
-			const current = queue.shift();
-			component.push(current);
-			const neighbors = adj.get(current);
-			if (!neighbors) continue;
-			for (const neighbor of neighbors) if (!visited.has(neighbor) && !SKIP_PREFIXES.some((p) => neighbor.startsWith(p))) {
-				visited.add(neighbor);
-				queue.push(neighbor);
-			}
-		}
-		if (component.length >= minSize) components.push(component);
-	}
-	components.sort((a, b) => b.length - a.length);
-	const topComponents = components.slice(0, maxClusters);
-	function generateLinkLabel(paths) {
-		const wordCounts = /* @__PURE__ */ new Map();
-		for (const p of paths) {
-			const title = noteMap.get(p)?.title;
-			if (!title) continue;
-			const words = title.toLowerCase().replace(/[^a-z0-9äöüàéèêëçñß\s]/g, " ").split(/\s+/).filter((w) => w.length > 2 && !STOP_WORDS.has(w));
-			for (const word of words) wordCounts.set(word, (wordCounts.get(word) ?? 0) + 1);
-		}
-		return [...wordCounts.entries()].sort((a, b) => b[1] - a[1]).slice(0, 3).map(([w]) => w).join(" / ") || "Linked Notes";
-	}
-	return {
-		themes: topComponents.map((component, idx) => {
-			const notes = component.map((p) => ({
-				path: p,
-				title: noteMap.get(p)?.title ?? null
-			}));
-			const avgRecency = component.reduce((sum, p) => sum + (noteMap.get(p)?.indexed_at ?? 0), 0) / component.length;
-			const uniqueFolders = new Set(component.map((p) => p.split("/")[0]));
-			return {
-				id: idx,
-				label: generateLinkLabel(component),
-				notes,
-				size: component.length,
-				folderDiversity: uniqueFolders.size / component.length,
-				avgRecency,
-				linkedRatio: 1,
-				suggestIndexNote: component.length >= 10
-			};
-		}),
-		totalNotesAnalyzed: recentNotes.length,
-		timeWindow: {
-			from,
-			to: now
-		}
-	};
-}
-async function handleGraphClusters(backend, params) {
-	const minSize = params.min_size ?? 3;
-	const maxClusters = params.max_clusters ?? 20;
-	const lookbackDays = params.lookback_days ?? 90;
-	if (!(params.project_id ?? 0)) throw new Error("graph_clusters: project_id is required (pass the vault project's numeric ID)");
-	const themeResult = await clusterByLinks(backend, lookbackDays, minSize, maxClusters);
-	const allPaths = themeResult.themes.flatMap((t) => t.notes.map((n) => n.path));
-	const observationsByPath = await backend.getObservationTypesForPaths(allPaths, params.project_id);
-	const fileRows = await backend.getVaultFilesByPaths(allPaths);
-	const indexedAtMap = new Map(fileRows.map((f) => [f.vaultPath, f.indexedAt]));
-	const clusters = themeResult.themes.map((theme) => {
-		const notePaths = theme.notes.map((n) => n.path);
-		const notesWithTimestamps = theme.notes.map((n) => ({
-			vault_path: n.path,
-			title: n.title ?? n.path.split("/").pop() ?? n.path,
-			indexed_at: indexedAtMap.get(n.path) ?? 0
-		}));
-		const avgRecency = theme.avgRecency;
-		const { dominant, counts } = aggregateObservationTypes(notePaths, observationsByPath);
-		return {
-			id: theme.id,
-			label: theme.label,
-			size: theme.size,
-			folder_diversity: theme.folderDiversity,
-			avg_recency: avgRecency,
-			linked_ratio: theme.linkedRatio,
-			dominant_observation_type: dominant,
-			observation_type_counts: counts,
-			suggest_index_note: theme.suggestIndexNote,
-			has_idea_note: false,
-			notes: notesWithTimestamps
-		};
-	});
-	clusters.sort((a, b) => b.size - a.size);
-	return {
-		clusters: clusters.slice(0, maxClusters),
-		total_notes_analyzed: themeResult.totalNotesAnalyzed,
-		time_window: themeResult.timeWindow
-	};
-}
-
-//#endregion
-export { handleGraphClusters };
-//# sourceMappingURL=clusters-BCtD3fbe.mjs.map

package/dist/clusters-BCtD3fbe.mjs.map

@@ -1 +0,0 @@
-{"version":3,"file":"clusters-BCtD3fbe.mjs","names":[],"sources":["../src/graph/clusters.ts"],"sourcesContent":["/**\n * clusters.ts — graph_clusters endpoint handler\n *\n * Reuses the zettelThemes() agglomerative clustering algorithm and enriches\n * each cluster with observation-type statistics, avg_recency from member\n * timestamps, and helper flags for the Obsidian knowledge plugin.\n */\n\nimport type { StorageBackend } from \"../storage/interface.js\";\nimport { STOP_WORDS } from \"../utils/stop-words.js\";\n\n// ---------------------------------------------------------------------------\n// Public param / result types\n// ---------------------------------------------------------------------------\n\nexport interface GraphClustersParams {\n  project_id?: number;\n  min_size?: number;\n  max_clusters?: number;\n  lookback_days?: number;\n  similarity_threshold?: number;\n}\n\nexport interface ClusterNode {\n  id: number;\n  label: string;\n  size: number;\n  folder_diversity: number;\n  avg_recency: number;\n  linked_ratio: number;\n  dominant_observation_type: string;\n  observation_type_counts: Record<string, number>;\n  suggest_index_note: boolean;\n  has_idea_note: boolean;\n  notes: Array<{ vault_path: string; title: string; indexed_at: number }>;\n}\n\nexport interface GraphClustersResult {\n  clusters: ClusterNode[];\n  total_notes_analyzed: number;\n  time_window: { from: number; to: number };\n}\n\n/**\n * Aggregate per-path observation type counts into cluster-level counts,\n * then pick the dominant type.\n */\nfunction aggregateObservationTypes(\n  paths: string[],\n  byPath: Map<string, Record<string, number>>\n): { dominant: string; counts: Record<string, number> } {\n  const counts: Record<string, number> = {};\n  for (const path of paths) {\n    const pathCounts = byPath.get(path);\n    if (!pathCounts) continue;\n    for (const [type, n] of Object.entries(pathCounts)) {\n      counts[type] = (counts[type] ?? 0) + n;\n    }\n  }\n\n  let dominant = \"unknown\";\n  let maxCount = 0;\n  for (const [type, n] of Object.entries(counts)) {\n    if (n > maxCount) {\n      maxCount = n;\n      dominant = type;\n    }\n  }\n\n  return { dominant, counts };\n}\n\n// ---------------------------------------------------------------------------\n// Link-based fallback clustering (wikilink connected components)\n// ---------------------------------------------------------------------------\n\nconst SKIP_PREFIXES = [\n  \"Attachments/\", \"🗓️ Daily Notes/\", \"Copilot/copilot-conversations/\",\n  \"Z - Zettelkasten/Tweets/\",\n];\n\n/**\n * Cluster vault notes by wikilink connectivity when embeddings aren't available.\n * Uses BFS to find connected components in the link graph, then picks the\n * largest components as clusters. Labels are derived from the most common\n * title words in each component.\n */\nasync function clusterByLinks(\n  backend: StorageBackend,\n  lookbackDays: number,\n  minSize: number,\n  maxClusters: number,\n): Promise<{ themes: Array<{ id: number; label: string; notes: Array<{ path: string; title: string | null }>; size: number; folderDiversity: number; avgRecency: number; linkedRatio: number; suggestIndexNote: boolean }>; totalNotesAnalyzed: number; timeWindow: { from: number; to: number } }> {\n  const now = Date.now();\n  const from = now - lookbackDays * 86400000;\n\n  // Get recent notes\n  const recentFiles = await backend.getRecentVaultFiles(from);\n  const recentNotes = recentFiles.filter(f => f.vaultPath.endsWith(\".md\"));\n\n  const noteMap = new Map<string, { title: string | null; indexed_at: number }>();\n  for (const n of recentNotes) {\n    noteMap.set(n.vaultPath, { title: n.title, indexed_at: n.indexedAt });\n  }\n\n  // Build adjacency list from vault_links (only for recent notes)\n  const adj = new Map<string, Set<string>>();\n  for (const path of noteMap.keys()) {\n    if (!adj.has(path)) adj.set(path, new Set());\n  }\n\n  const linkGraph = await backend.getVaultLinkGraph();\n\n  for (const { source_path, target_path } of linkGraph) {\n    if (noteMap.has(source_path) && noteMap.has(target_path)) {\n      adj.get(source_path)!.add(target_path);\n      adj.get(target_path)!.add(source_path);\n    }\n  }\n\n  // Remove hub nodes before BFS\n  const degrees = [...adj.entries()].map(([p, s]) => ({ path: p, degree: s.size }));\n  degrees.sort((a, b) => b.degree - a.degree);\n  const hubThreshold = Math.max(10, degrees[Math.floor(degrees.length * 0.05)]?.degree ?? 10);\n  const hubNodes = new Set<string>();\n  for (const { path, degree } of degrees) {\n    if (degree >= hubThreshold) hubNodes.add(path);\n    else break;\n  }\n\n  for (const hub of hubNodes) {\n    adj.delete(hub);\n  }\n  for (const [, neighbors] of adj) {\n    for (const hub of hubNodes) {\n      neighbors.delete(hub);\n    }\n  }\n\n  // BFS connected components\n  const visited = new Set<string>();\n  const components: string[][] = [];\n\n  for (const path of noteMap.keys()) {\n    if (visited.has(path) || hubNodes.has(path)) continue;\n    if (SKIP_PREFIXES.some(p => path.startsWith(p))) { visited.add(path); continue; }\n    const component: string[] = [];\n    const queue = [path];\n    visited.add(path);\n\n    while (queue.length > 0) {\n      const current = queue.shift()!;\n      component.push(current);\n      const neighbors = adj.get(current);\n      if (!neighbors) continue;\n      for (const neighbor of neighbors) {\n        if (!visited.has(neighbor) && !SKIP_PREFIXES.some(p => neighbor.startsWith(p))) {\n          visited.add(neighbor);\n          queue.push(neighbor);\n        }\n      }\n    }\n\n    if (component.length >= minSize) {\n      components.push(component);\n    }\n  }\n\n  components.sort((a, b) => b.length - a.length);\n  const topComponents = components.slice(0, maxClusters);\n\n  // STOP_WORDS imported from utils/stop-words.ts (module-level import)\n\n  function generateLinkLabel(paths: string[]): string {\n    const wordCounts = new Map<string, number>();\n    for (const p of paths) {\n      const title = noteMap.get(p)?.title;\n      if (!title) continue;\n      const words = title.toLowerCase().replace(/[^a-z0-9äöüàéèêëçñß\\s]/g, \" \").split(/\\s+/)\n        .filter(w => w.length > 2 && !STOP_WORDS.has(w));\n      for (const word of words) {\n        wordCounts.set(word, (wordCounts.get(word) ?? 0) + 1);\n      }\n    }\n    const sorted = [...wordCounts.entries()].sort((a, b) => b[1] - a[1]);\n    return sorted.slice(0, 3).map(([w]) => w).join(\" / \") || \"Linked Notes\";\n  }\n\n  const themes = topComponents.map((component, idx) => {\n    const notes = component.map(p => ({\n      path: p,\n      title: noteMap.get(p)?.title ?? null,\n    }));\n    const avgRecency = component.reduce((sum, p) => sum + (noteMap.get(p)?.indexed_at ?? 0), 0) / component.length;\n    const uniqueFolders = new Set(component.map(p => p.split(\"/\")[0]));\n\n    return {\n      id: idx,\n      label: generateLinkLabel(component),\n      notes,\n      size: component.length,\n      folderDiversity: uniqueFolders.size / component.length,\n      avgRecency,\n      linkedRatio: 1.0,\n      suggestIndexNote: component.length >= 10,\n    };\n  });\n\n  return {\n    themes,\n    totalNotesAnalyzed: recentNotes.length,\n    timeWindow: { from, to: now },\n  };\n}\n\n// ---------------------------------------------------------------------------\n// Main handler\n// ---------------------------------------------------------------------------\n\nexport async function handleGraphClusters(\n  backend: StorageBackend,\n  params: GraphClustersParams\n): Promise<GraphClustersResult> {\n  const minSize = params.min_size ?? 3;\n  const maxClusters = params.max_clusters ?? 20;\n  const lookbackDays = params.lookback_days ?? 90;\n\n  const vaultProjectId = params.project_id ?? 0;\n\n  if (!vaultProjectId) {\n    throw new Error(\n      \"graph_clusters: project_id is required (pass the vault project's numeric ID)\"\n    );\n  }\n\n  const themeResult = await clusterByLinks(backend, lookbackDays, minSize, maxClusters);\n\n  const allPaths = themeResult.themes.flatMap((t) => t.notes.map((n) => n.path));\n\n  const observationsByPath = await backend.getObservationTypesForPaths(allPaths, params.project_id);\n\n  // Fetch indexed_at timestamps for all notes in bulk\n  const fileRows = await backend.getVaultFilesByPaths(allPaths);\n  const indexedAtMap = new Map<string, number>(fileRows.map(f => [f.vaultPath, f.indexedAt]));\n\n  const clusters: ClusterNode[] = themeResult.themes.map((theme) => {\n    const notePaths = theme.notes.map((n) => n.path);\n\n    const notesWithTimestamps = theme.notes.map((n) => ({\n      vault_path: n.path,\n      title: n.title ?? n.path.split(\"/\").pop() ?? n.path,\n      indexed_at: indexedAtMap.get(n.path) ?? 0,\n    }));\n\n    const avgRecency = theme.avgRecency;\n\n    const { dominant, counts } = aggregateObservationTypes(\n      notePaths,\n      observationsByPath\n    );\n\n    return {\n      id: theme.id,\n      label: theme.label,\n      size: theme.size,\n      folder_diversity: theme.folderDiversity,\n      avg_recency: avgRecency,\n      linked_ratio: theme.linkedRatio,\n      dominant_observation_type: dominant,\n      observation_type_counts: counts,\n      suggest_index_note: theme.suggestIndexNote,\n      has_idea_note: false,\n      notes: notesWithTimestamps,\n    };\n  });\n\n  clusters.sort((a, b) => b.size - a.size);\n\n  return {\n    clusters: clusters.slice(0, maxClusters),\n    total_notes_analyzed: themeResult.totalNotesAnalyzed,\n    time_window: themeResult.timeWindow,\n  };\n}\n"],"mappings":";;;;;;;AA+CA,SAAS,0BACP,OACA,QACsD;CACtD,MAAM,SAAiC,EAAE;AACzC,MAAK,MAAM,QAAQ,OAAO;EACxB,MAAM,aAAa,OAAO,IAAI,KAAK;AACnC,MAAI,CAAC,WAAY;AACjB,OAAK,MAAM,CAAC,MAAM,MAAM,OAAO,QAAQ,WAAW,CAChD,QAAO,SAAS,OAAO,SAAS,KAAK;;CAIzC,IAAI,WAAW;CACf,IAAI,WAAW;AACf,MAAK,MAAM,CAAC,MAAM,MAAM,OAAO,QAAQ,OAAO,CAC5C,KAAI,IAAI,UAAU;AAChB,aAAW;AACX,aAAW;;AAIf,QAAO;EAAE;EAAU;EAAQ;;AAO7B,MAAM,gBAAgB;CACpB;CAAgB;CAAoB;CACpC;CACD;;;;;;;AAQD,eAAe,eACb,SACA,cACA,SACA,aACkS;CAClS,MAAM,MAAM,KAAK,KAAK;CACtB,MAAM,OAAO,MAAM,eAAe;CAIlC,MAAM,eADc,MAAM,QAAQ,oBAAoB,KAAK,EAC3B,QAAO,MAAK,EAAE,UAAU,SAAS,MAAM,CAAC;CAExE,MAAM,0BAAU,IAAI,KAA2D;AAC/E,MAAK,MAAM,KAAK,YACd,SAAQ,IAAI,EAAE,WAAW;EAAE,OAAO,EAAE;EAAO,YAAY,EAAE;EAAW,CAAC;CAIvE,MAAM,sBAAM,IAAI,KAA0B;AAC1C,MAAK,MAAM,QAAQ,QAAQ,MAAM,CAC/B,KAAI,CAAC,IAAI,IAAI,KAAK,CAAE,KAAI,IAAI,sBAAM,IAAI,KAAK,CAAC;CAG9C,MAAM,YAAY,MAAM,QAAQ,mBAAmB;AAEnD,MAAK,MAAM,EAAE,aAAa,iBAAiB,UACzC,KAAI,QAAQ,IAAI,YAAY,IAAI,QAAQ,IAAI,YAAY,EAAE;AACxD,MAAI,IAAI,YAAY,CAAE,IAAI,YAAY;AACtC,MAAI,IAAI,YAAY,CAAE,IAAI,YAAY;;CAK1C,MAAM,UAAU,CAAC,GAAG,IAAI,SAAS,CAAC,CAAC,KAAK,CAAC,GAAG,QAAQ;EAAE,MAAM;EAAG,QAAQ,EAAE;EAAM,EAAE;AACjF,SAAQ,MAAM,GAAG,MAAM,EAAE,SAAS,EAAE,OAAO;CAC3C,MAAM,eAAe,KAAK,IAAI,IAAI,QAAQ,KAAK,MAAM,QAAQ,SAAS,IAAK,GAAG,UAAU,GAAG;CAC3F,MAAM,2BAAW,IAAI,KAAa;AAClC,MAAK,MAAM,EAAE,MAAM,YAAY,QAC7B,KAAI,UAAU,aAAc,UAAS,IAAI,KAAK;KACzC;AAGP,MAAK,MAAM,OAAO,SAChB,KAAI,OAAO,IAAI;AAEjB,MAAK,MAAM,GAAG,cAAc,IAC1B,MAAK,MAAM,OAAO,SAChB,WAAU,OAAO,IAAI;CAKzB,MAAM,0BAAU,IAAI,KAAa;CACjC,MAAM,aAAyB,EAAE;AAEjC,MAAK,MAAM,QAAQ,QAAQ,MAAM,EAAE;AACjC,MAAI,QAAQ,IAAI,KAAK,IAAI,SAAS,IAAI,KAAK,CAAE;AAC7C,MAAI,cAAc,MAAK,MAAK,KAAK,WAAW,EAAE,CAAC,EAAE;AAAE,WAAQ,IAAI,KAAK;AAAE;;EACtE,MAAM,YAAsB,EAAE;EAC9B,MAAM,QAAQ,CAAC,KAAK;AACpB,UAAQ,IAAI,KAAK;AAEjB,SAAO,MAAM,SAAS,GAAG;GACvB,MAAM,UAAU,MAAM,OAAO;AAC7B,aAAU,KAAK,QAAQ;GACvB,MAAM,YAAY,IAAI,IAAI,QAAQ;AAClC,OAAI,CAAC,UAAW;AAChB,QAAK,MAAM,YAAY,UACrB,KAAI,CAAC,QAAQ,IAAI,SAAS,IAAI,CAAC,cAAc,MAAK,MAAK,SAAS,WAAW,EAAE,CAAC,EAAE;AAC9E,YAAQ,IAAI,SAAS;AACrB,UAAM,KAAK,SAAS;;;AAK1B,MAAI,UAAU,UAAU,QACtB,YAAW,KAAK,UAAU;;AAI9B,YAAW,MAAM,GAAG,MAAM,EAAE,SAAS,EAAE,OAAO;CAC9C,MAAM,gBAAgB,WAAW,MAAM,GAAG,YAAY;CAItD,SAAS,kBAAkB,OAAyB;EAClD,MAAM,6BAAa,IAAI,KAAqB;AAC5C,OAAK,MAAM,KAAK,OAAO;GACrB,MAAM,QAAQ,QAAQ,IAAI,EAAE,EAAE;AAC9B,OAAI,CAAC,MAAO;GACZ,MAAM,QAAQ,MAAM,aAAa,CAAC,QAAQ,2BAA2B,IAAI,CAAC,MAAM,MAAM,CACnF,QAAO,MAAK,EAAE,SAAS,KAAK,CAAC,WAAW,IAAI,EAAE,CAAC;AAClD,QAAK,MAAM,QAAQ,MACjB,YAAW,IAAI,OAAO,WAAW,IAAI,KAAK,IAAI,KAAK,EAAE;;AAIzD,SADe,CAAC,GAAG,WAAW,SAAS,CAAC,CAAC,MAAM,GAAG,MAAM,EAAE,KAAK,EAAE,GAAG,CACtD,MAAM,GAAG,EAAE,CAAC,KAAK,CAAC,OAAO,EAAE,CAAC,KAAK,MAAM,IAAI;;AAuB3D,QAAO;EACL,QArBa,cAAc,KAAK,WAAW,QAAQ;GACnD,MAAM,QAAQ,UAAU,KAAI,OAAM;IAChC,MAAM;IACN,OAAO,QAAQ,IAAI,EAAE,EAAE,SAAS;IACjC,EAAE;GACH,MAAM,aAAa,UAAU,QAAQ,KAAK,MAAM,OAAO,QAAQ,IAAI,EAAE,EAAE,cAAc,IAAI,EAAE,GAAG,UAAU;GACxG,MAAM,gBAAgB,IAAI,IAAI,UAAU,KAAI,MAAK,EAAE,MAAM,IAAI,CAAC,GAAG,CAAC;AAElE,UAAO;IACL,IAAI;IACJ,OAAO,kBAAkB,UAAU;IACnC;IACA,MAAM,UAAU;IAChB,iBAAiB,cAAc,OAAO,UAAU;IAChD;IACA,aAAa;IACb,kBAAkB,UAAU,UAAU;IACvC;IACD;EAIA,oBAAoB,YAAY;EAChC,YAAY;GAAE;GAAM,IAAI;GAAK;EAC9B;;AAOH,eAAsB,oBACpB,SACA,QAC8B;CAC9B,MAAM,UAAU,OAAO,YAAY;CACnC,MAAM,cAAc,OAAO,gBAAgB;CAC3C,MAAM,eAAe,OAAO,iBAAiB;AAI7C,KAAI,EAFmB,OAAO,cAAc,GAG1C,OAAM,IAAI,MACR,+EACD;CAGH,MAAM,cAAc,MAAM,eAAe,SAAS,cAAc,SAAS,YAAY;CAErF,MAAM,WAAW,YAAY,OAAO,SAAS,MAAM,EAAE,MAAM,KAAK,MAAM,EAAE,KAAK,CAAC;CAE9E,MAAM,qBAAqB,MAAM,QAAQ,4BAA4B,UAAU,OAAO,WAAW;CAGjG,MAAM,WAAW,MAAM,QAAQ,qBAAqB,SAAS;CAC7D,MAAM,eAAe,IAAI,IAAoB,SAAS,KAAI,MAAK,CAAC,EAAE,WAAW,EAAE,UAAU,CAAC,CAAC;CAE3F,MAAM,WAA0B,YAAY,OAAO,KAAK,UAAU;EAChE,MAAM,YAAY,MAAM,MAAM,KAAK,MAAM,EAAE,KAAK;EAEhD,MAAM,sBAAsB,MAAM,MAAM,KAAK,OAAO;GAClD,YAAY,EAAE;GACd,OAAO,EAAE,SAAS,EAAE,KAAK,MAAM,IAAI,CAAC,KAAK,IAAI,EAAE;GAC/C,YAAY,aAAa,IAAI,EAAE,KAAK,IAAI;GACzC,EAAE;EAEH,MAAM,aAAa,MAAM;EAEzB,MAAM,EAAE,UAAU,WAAW,0BAC3B,WACA,mBACD;AAED,SAAO;GACL,IAAI,MAAM;GACV,OAAO,MAAM;GACb,MAAM,MAAM;GACZ,kBAAkB,MAAM;GACxB,aAAa;GACb,cAAc,MAAM;GACpB,2BAA2B;GAC3B,yBAAyB;GACzB,oBAAoB,MAAM;GAC1B,eAAe;GACf,OAAO;GACR;GACD;AAEF,UAAS,MAAM,GAAG,MAAM,EAAE,OAAO,EAAE,KAAK;AAExC,QAAO;EACL,UAAU,SAAS,MAAM,GAAG,YAAY;EACxC,sBAAsB,YAAY;EAClC,aAAa,YAAY;EAC1B"}