agentquad 0.3.0

package/src/db.js

@@ -0,0 +1,1503 @@
+import Database from 'better-sqlite3'
+import { randomUUID } from 'node:crypto'
+
+const SCHEMA = `
+CREATE TABLE IF NOT EXISTS todos (
+  id           TEXT PRIMARY KEY,
+  parent_id    TEXT,
+  title        TEXT NOT NULL,
+  description  TEXT NOT NULL DEFAULT '',
+  quadrant     INTEGER NOT NULL CHECK(quadrant IN (1,2,3,4)),
+  status       TEXT NOT NULL DEFAULT 'todo',
+  due_date     INTEGER,
+  work_dir     TEXT,
+  sort_order   REAL NOT NULL,
+  ai_session   TEXT,
+  completed_at INTEGER,
+  created_at   INTEGER NOT NULL,
+  updated_at   INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_todos_quadrant_sort ON todos(quadrant, sort_order);
+CREATE INDEX IF NOT EXISTS idx_todos_status        ON todos(status);
+
+CREATE TABLE IF NOT EXISTS comments (
+  id         TEXT PRIMARY KEY,
+  todo_id    TEXT NOT NULL,
+  content    TEXT NOT NULL,
+  created_at INTEGER NOT NULL,
+  FOREIGN KEY (todo_id) REFERENCES todos(id) ON DELETE CASCADE
+);
+CREATE INDEX IF NOT EXISTS idx_comments_todo ON comments(todo_id, created_at);
+
+CREATE TABLE IF NOT EXISTS ai_session_log (
+  id            TEXT PRIMARY KEY,
+  todo_id       TEXT NOT NULL,
+  tool          TEXT NOT NULL,
+  quadrant      INTEGER NOT NULL,
+  status        TEXT NOT NULL,
+  exit_code     INTEGER,
+  started_at    INTEGER NOT NULL,
+  completed_at  INTEGER NOT NULL,
+  duration_ms   INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_ail_completed_at ON ai_session_log(completed_at);
+CREATE INDEX IF NOT EXISTS idx_ail_tool         ON ai_session_log(tool);
+CREATE INDEX IF NOT EXISTS idx_ail_quadrant     ON ai_session_log(quadrant);
+
+CREATE TABLE IF NOT EXISTS transcript_files (
+  id                INTEGER PRIMARY KEY,
+  tool              TEXT NOT NULL,
+  native_id         TEXT,
+  cwd               TEXT,
+  jsonl_path        TEXT NOT NULL UNIQUE,
+  size              INTEGER NOT NULL,
+  mtime             INTEGER NOT NULL,
+  started_at        INTEGER,
+  ended_at          INTEGER,
+  first_user_prompt TEXT,
+  turn_count        INTEGER NOT NULL DEFAULT 0,
+  input_tokens          INTEGER,
+  output_tokens         INTEGER,
+  cache_read_tokens     INTEGER,
+  cache_creation_tokens INTEGER,
+  primary_model         TEXT,
+  active_ms             INTEGER,
+  bound_todo_id     TEXT,
+  indexed_at        INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_tf_native ON transcript_files(native_id);
+CREATE INDEX IF NOT EXISTS idx_tf_bound  ON transcript_files(bound_todo_id);
+CREATE INDEX IF NOT EXISTS idx_tf_tool_cwd_started ON transcript_files(tool, cwd, started_at);
+
+CREATE TABLE IF NOT EXISTS recurring_rules (
+  id                    TEXT PRIMARY KEY,
+  title                 TEXT NOT NULL,
+  description           TEXT NOT NULL DEFAULT '',
+  quadrant              INTEGER NOT NULL CHECK(quadrant IN (1,2,3,4)),
+  work_dir              TEXT,
+  brainstorm            INTEGER NOT NULL DEFAULT 0,
+  applied_template_ids  TEXT,
+  subtodos              TEXT,
+  frequency             TEXT NOT NULL CHECK(frequency IN ('daily','weekly','monthly')),
+  weekdays              TEXT,
+  month_days            TEXT,
+  active                INTEGER NOT NULL DEFAULT 1,
+  last_generated_date   TEXT,
+  created_at            INTEGER NOT NULL,
+  updated_at            INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_rr_active ON recurring_rules(active);
+
+CREATE TABLE IF NOT EXISTS prompt_templates (
+  id          TEXT PRIMARY KEY,
+  name        TEXT NOT NULL,
+  description TEXT NOT NULL DEFAULT '',
+  content     TEXT NOT NULL,
+  builtin     INTEGER NOT NULL DEFAULT 0,
+  sort_order  REAL NOT NULL DEFAULT 0,
+  created_at  INTEGER NOT NULL,
+  updated_at  INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_pt_sort ON prompt_templates(sort_order);
+
+CREATE TABLE IF NOT EXISTS wiki_runs (
+  id            INTEGER PRIMARY KEY AUTOINCREMENT,
+  started_at    INTEGER NOT NULL,
+  completed_at  INTEGER,
+  todo_count    INTEGER NOT NULL DEFAULT 0,
+  dry_run       INTEGER NOT NULL DEFAULT 0,
+  exit_code     INTEGER,
+  error         TEXT,
+  note          TEXT
+);
+CREATE INDEX IF NOT EXISTS idx_wiki_runs_started ON wiki_runs(started_at DESC);
+
+CREATE TABLE IF NOT EXISTS wiki_todo_coverage (
+  wiki_run_id   INTEGER NOT NULL,
+  todo_id       TEXT NOT NULL,
+  source_path   TEXT,
+  llm_applied   INTEGER NOT NULL DEFAULT 0,
+  PRIMARY KEY (wiki_run_id, todo_id)
+);
+CREATE INDEX IF NOT EXISTS idx_wiki_cov_todo ON wiki_todo_coverage(todo_id, llm_applied);
+
+CREATE TABLE IF NOT EXISTS pending_questions (
+  ticket          TEXT PRIMARY KEY,
+  session_id      TEXT NOT NULL,
+  todo_id         TEXT,
+  question        TEXT NOT NULL,
+  options_json    TEXT NOT NULL,
+  status          TEXT NOT NULL DEFAULT 'pending',
+  answer_text     TEXT,
+  chosen_index    INTEGER,
+  created_at      INTEGER NOT NULL,
+  answered_at     INTEGER,
+  timeout_ms      INTEGER NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_pq_status_created ON pending_questions(status, created_at DESC);
+CREATE INDEX IF NOT EXISTS idx_pq_session ON pending_questions(session_id);
+`
+
+const UNFINISHED = ['todo', 'ai_running', 'ai_pending', 'ai_done']
+
+function normalizeAiSessions(value) {
+  if (!value) return []
+  if (Array.isArray(value)) return value.filter(Boolean)
+  return [value]
+}
+
+function currentAiSession(aiSessions) {
+  if (!aiSessions.length) return null
+  return aiSessions.find(s => s?.status === 'running' || s?.status === 'idle' || s?.status === 'pending_confirm') || aiSessions[0]
+}
+
+function rowToTodo(row) {
+  if (!row) return null
+  const aiSessions = normalizeAiSessions(row.ai_session ? JSON.parse(row.ai_session) : null)
+  return {
+    id: row.id,
+    parentId: row.parent_id ?? null,
+    title: row.title,
+    description: row.description,
+    quadrant: row.quadrant,
+    status: row.status,
+    dueDate: row.due_date,
+    workDir: row.work_dir ?? null,
+    brainstorm: !!row.brainstorm,
+    appliedTemplateIds: row.applied_template_ids ? (() => { try { return JSON.parse(row.applied_template_ids) } catch { return [] } })() : [],
+    sortOrder: row.sort_order,
+    aiSession: currentAiSession(aiSessions),
+    aiSessions,
+    recurringRuleId: row.recurring_rule_id ?? null,
+    instanceDate: row.instance_date ?? null,
+    completedAt: row.completed_at ?? null,
+    stageTag: row.stage_tag ?? null,
+    archivedAt: row.archived_at ?? null,
+    createdAt: row.created_at,
+    updatedAt: row.updated_at,
+  }
+}
+
+function todayStr(now = Date.now()) {
+  const d = new Date(now)
+  const y = d.getFullYear()
+  const m = String(d.getMonth() + 1).padStart(2, '0')
+  const day = String(d.getDate()).padStart(2, '0')
+  return `${y}-${m}-${day}`
+}
+
+function endOfDayMs(now = Date.now()) {
+  const d = new Date(now)
+  d.setHours(23, 59, 59, 999)
+  return d.getTime()
+}
+
+function parseJsonArray(value) {
+  if (!value) return null
+  try {
+    const parsed = JSON.parse(value)
+    return Array.isArray(parsed) ? parsed : null
+  } catch { return null }
+}
+
+function rowToRule(row) {
+  if (!row) return null
+  return {
+    id: row.id,
+    title: row.title,
+    description: row.description,
+    quadrant: row.quadrant,
+    workDir: row.work_dir ?? null,
+    brainstorm: !!row.brainstorm,
+    appliedTemplateIds: parseJsonArray(row.applied_template_ids) || [],
+    subtodos: parseJsonArray(row.subtodos) || [],
+    frequency: row.frequency,
+    weekdays: parseJsonArray(row.weekdays) || [],
+    monthDays: parseJsonArray(row.month_days) || [],
+    active: !!row.active,
+    lastGeneratedDate: row.last_generated_date ?? null,
+    createdAt: row.created_at,
+    updatedAt: row.updated_at,
+  }
+}
+
+function ruleShouldProduceOn(rule, dateStr) {
+  const d = new Date(`${dateStr}T12:00:00`)
+  if (rule.frequency === 'daily') return true
+  if (rule.frequency === 'weekly') {
+    const wd = d.getDay()
+    return (rule.weekdays || []).includes(wd)
+  }
+  if (rule.frequency === 'monthly') {
+    const dom = d.getDate()
+    return (rule.monthDays || []).includes(dom)
+  }
+  return false
+}
+
+export function openDb(file = ':memory:') {
+  const db = new Database(file)
+  db.pragma('journal_mode = WAL')
+  db.exec(SCHEMA)
+
+  let ftsAvailable = false
+  try {
+    const existing = db.prepare(`SELECT sql FROM sqlite_master WHERE type='table' AND name='transcript_fts'`).get()
+    const usesTrigram = existing && /tokenize\s*=\s*['"]?trigram/i.test(existing.sql || '')
+    if (existing && !usesTrigram) {
+      db.exec(`DROP TABLE transcript_fts`)
+      // 旧 tokenizer 下的 FTS 已清空；把 transcript_files.size 置 -1 让下一次 scan 视为脏，触发重建
+      try { db.exec(`UPDATE transcript_files SET size = -1`) } catch {}
+    }
+    db.exec(`
+      CREATE VIRTUAL TABLE IF NOT EXISTS transcript_fts USING fts5(
+        content,
+        role UNINDEXED,
+        file_id UNINDEXED,
+        tokenize = "trigram"
+      );
+    `)
+    ftsAvailable = true
+  } catch (e) {
+    ftsAvailable = false
+  }
+
+  const columns = db.prepare(`PRAGMA table_info(todos)`).all()
+  if (!columns.some(col => col.name === 'parent_id')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN parent_id TEXT`)
+  }
+  if (!columns.some(col => col.name === 'work_dir')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN work_dir TEXT`)
+  }
+  if (!columns.some(col => col.name === 'brainstorm')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN brainstorm INTEGER NOT NULL DEFAULT 0`)
+  }
+  if (!columns.some(col => col.name === 'applied_template_ids')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN applied_template_ids TEXT`)
+  }
+  if (!columns.some(col => col.name === 'recurring_rule_id')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN recurring_rule_id TEXT`)
+  }
+  if (!columns.some(col => col.name === 'instance_date')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN instance_date TEXT`)
+  }
+  if (!columns.some(col => col.name === 'completed_at')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN completed_at INTEGER`)
+    // 一次性回填：已完成的旧行用 updated_at 作为近似完成时间
+    db.exec(`UPDATE todos SET completed_at = updated_at WHERE status = 'done' AND completed_at IS NULL`)
+  }
+  if (!columns.some(col => col.name === 'archived_at')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN archived_at INTEGER`)
+  }
+  if (!columns.some(col => col.name === 'stage_tag')) {
+    db.exec(`ALTER TABLE todos ADD COLUMN stage_tag TEXT`)
+  }
+  db.exec(`CREATE INDEX IF NOT EXISTS idx_todos_recurring ON todos(recurring_rule_id, instance_date)`)
+  db.exec(`CREATE INDEX IF NOT EXISTS idx_todos_parent_sort ON todos(parent_id, sort_order)`)
+  db.exec(`CREATE INDEX IF NOT EXISTS idx_todos_quad_parent_sort ON todos(quadrant, parent_id, sort_order)`)
+  db.exec(`CREATE INDEX IF NOT EXISTS idx_todos_completed_at ON todos(completed_at)`)
+  db.exec(`CREATE INDEX IF NOT EXISTS idx_todos_archived_at ON todos(archived_at)`)
+
+  // Phase-out: pipeline feature removed in 2026-05-13 cleanup.
+  // Drop the two tables if they exist (idempotent — no-op for fresh installs).
+  db.exec(`DROP TABLE IF EXISTS pipeline_runs;`)
+  db.exec(`DROP TABLE IF EXISTS pipeline_templates;`)
+
+  const tfCols = db.prepare(`PRAGMA table_info(transcript_files)`).all().map(c => c.name)
+  for (const [name, type] of [
+    ['input_tokens', 'INTEGER'],
+    ['output_tokens', 'INTEGER'],
+    ['cache_read_tokens', 'INTEGER'],
+    ['cache_creation_tokens', 'INTEGER'],
+    ['primary_model', 'TEXT'],
+    ['active_ms', 'INTEGER'],
+  ]) {
+    if (!tfCols.includes(name)) db.exec(`ALTER TABLE transcript_files ADD COLUMN ${name} ${type}`)
+  }
+
+  const stmts = {
+    insert: db.prepare(`
+      INSERT INTO todos (id, parent_id, title, description, quadrant, status, due_date, work_dir, brainstorm, applied_template_ids, sort_order, ai_session, recurring_rule_id, instance_date, completed_at, created_at, updated_at)
+      VALUES (@id, @parent_id, @title, @description, @quadrant, @status, @due_date, @work_dir, @brainstorm, @applied_template_ids, @sort_order, @ai_session, @recurring_rule_id, @instance_date, @completed_at, @created_at, @updated_at)
+    `),
+    getById: db.prepare(`SELECT * FROM todos WHERE id = ?`),
+    listChildrenByParent: db.prepare(`SELECT id FROM todos WHERE parent_id = ? ORDER BY sort_order ASC, created_at ASC`),
+    maxSortInQuadrant: db.prepare(`SELECT MAX(sort_order) AS m FROM todos WHERE quadrant = ? AND parent_id IS NULL`),
+    maxSortInParent: db.prepare(`SELECT MAX(sort_order) AS m FROM todos WHERE parent_id = ?`),
+    deleteById: db.prepare(`DELETE FROM todos WHERE id = ?`),
+  }
+
+  function nextSortOrder(quadrant, parentId = null) {
+    const row = parentId
+      ? stmts.maxSortInParent.get(parentId)
+      : stmts.maxSortInQuadrant.get(quadrant)
+    const m = row?.m
+    return (m == null ? 0 : m) + 1024
+  }
+
+  function resolveParent(parentId) {
+    if (parentId == null) return null
+    const parent = rowToTodo(stmts.getById.get(parentId))
+    if (!parent) throw new Error('parent_not_found')
+    if (parent.parentId) throw new Error('nested_subtodo_not_allowed')
+    return parent
+  }
+
+  function createTodo(data) {
+    const now = Date.now()
+    const parent = resolveParent(data.parentId ?? null)
+    const quadrant = parent ? parent.quadrant : (Number(data.quadrant) || 4)
+    const status = data.status || 'todo'
+    const row = {
+      id: randomUUID(),
+      parent_id: parent?.id ?? null,
+      title: data.title,
+      description: data.description || '',
+      quadrant,
+      status,
+      due_date: data.dueDate ?? null,
+      work_dir: data.workDir ?? null,
+      brainstorm: data.brainstorm ? 1 : 0,
+      applied_template_ids: Array.isArray(data.appliedTemplateIds) ? JSON.stringify(data.appliedTemplateIds) : null,
+      sort_order: data.sortOrder != null ? data.sortOrder : nextSortOrder(quadrant, parent?.id ?? null),
+      ai_session: JSON.stringify(normalizeAiSessions(data.aiSessions ?? data.aiSession)),
+      recurring_rule_id: data.recurringRuleId ?? null,
+      instance_date: data.instanceDate ?? null,
+      completed_at: status === 'done' ? now : null,
+      created_at: now,
+      updated_at: now,
+    }
+    stmts.insert.run(row)
+    return rowToTodo(stmts.getById.get(row.id))
+  }
+
+  function getTodo(id) {
+    return rowToTodo(stmts.getById.get(id))
+  }
+
+  function updateTodo(id, patch) {
+    const existing = rowToTodo(stmts.getById.get(id))
+    if (!existing) return null
+    const fields = []
+    const bind = { id }
+    const map = {
+      title: 'title',
+      description: 'description',
+      quadrant: 'quadrant',
+      status: 'status',
+      dueDate: 'due_date',
+      workDir: 'work_dir',
+      brainstorm: 'brainstorm',
+      sortOrder: 'sort_order',
+      stageTag: 'stage_tag',
+    }
+
+    let nextParentId = existing.parentId
+    if (patch.parentId !== undefined) {
+      nextParentId = patch.parentId
+    }
+    const parent = resolveParent(nextParentId)
+    if (parent && parent.id === id) throw new Error('parent_cycle')
+    const nextQuadrant = parent ? parent.quadrant : (patch.quadrant !== undefined ? Number(patch.quadrant) || 4 : existing.quadrant)
+    if (parent && patch.quadrant !== undefined && parent.quadrant !== nextQuadrant) {
+      throw new Error('parent_quadrant_mismatch')
+    }
+
+    for (const [k, col] of Object.entries(map)) {
+      if (patch[k] !== undefined) {
+        fields.push(`${col} = @${col}`)
+        bind[col] = k === 'brainstorm' ? (patch[k] ? 1 : 0) : patch[k]
+      }
+    }
+    if (parent && patch.quadrant === undefined && existing.quadrant !== parent.quadrant) {
+      fields.push(`quadrant = @quadrant`)
+      bind.quadrant = parent.quadrant
+    }
+    if (patch.parentId !== undefined) {
+      fields.push(`parent_id = @parent_id`)
+      bind.parent_id = parent?.id ?? null
+    }
+    if (patch.appliedTemplateIds !== undefined) {
+      fields.push(`applied_template_ids = @applied_template_ids`)
+      bind.applied_template_ids = Array.isArray(patch.appliedTemplateIds) ? JSON.stringify(patch.appliedTemplateIds) : null
+    }
+    if (patch.aiSession !== undefined) {
+      const sessions = patch.aiSession === null ? [] : normalizeAiSessions(patch.aiSession)
+      fields.push(`ai_session = @ai_session`)
+      bind.ai_session = JSON.stringify(sessions)
+    }
+    if (patch.aiSessions !== undefined) {
+      fields.push(`ai_session = @ai_session`)
+      bind.ai_session = JSON.stringify(normalizeAiSessions(patch.aiSessions))
+    }
+    const now = Date.now()
+    if (patch.status !== undefined && patch.status !== existing.status) {
+      if (patch.status === 'done') {
+        fields.push(`completed_at = @completed_at`)
+        bind.completed_at = now
+      } else if (existing.status === 'done') {
+        fields.push(`completed_at = @completed_at`)
+        bind.completed_at = null
+      }
+    }
+    if (!fields.length) return existing
+    fields.push(`updated_at = @updated_at`)
+    bind.updated_at = now
+    const sql = `UPDATE todos SET ${fields.join(', ')} WHERE id = @id`
+    db.prepare(sql).run(bind)
+    if (!existing.parentId && nextQuadrant !== existing.quadrant) {
+      db.prepare(`UPDATE todos SET quadrant = ?, updated_at = ? WHERE parent_id = ?`).run(nextQuadrant, now, id)
+    }
+    return rowToTodo(stmts.getById.get(id))
+  }
+
+  function deleteTodo(id) {
+    const children = stmts.listChildrenByParent.all(id)
+    for (const child of children) {
+      deleteTodo(child.id)
+    }
+    stmts.deleteById.run(id)
+  }
+
+  function listTodos({ quadrant, status, keyword, archived } = {}) {
+    const where = []
+    const params = []
+    if (quadrant != null) {
+      where.push('quadrant = ?')
+      params.push(Number(quadrant))
+    }
+    if (status === 'todo') {
+      where.push(`status IN (${UNFINISHED.map(() => '?').join(',')})`)
+      params.push(...UNFINISHED)
+    } else if (status === 'done') {
+      where.push('status = ?')
+      params.push('done')
+    } else {
+      where.push(`status != 'missed'`)
+    }
+    // archived: undefined|false → 只看未归档；true → 只看已归档；'all' → 都要
+    if (archived === true) {
+      where.push('archived_at IS NOT NULL')
+    } else if (archived === 'all') {
+      // no-op, 全部
+    } else {
+      where.push('archived_at IS NULL')
+    }
+    const whereSql = where.length ? `WHERE ${where.join(' AND ')}` : ''
+    const rows = db.prepare(`
+      SELECT * FROM todos
+      ${whereSql}
+      ORDER BY quadrant ASC, COALESCE(parent_id, id) ASC, CASE WHEN parent_id IS NULL THEN 0 ELSE 1 END ASC, sort_order ASC, created_at ASC
+    `).all(...params)
+    const todos = rows.map(rowToTodo)
+    if (!keyword) return todos
+
+    const needle = keyword.toLowerCase()
+    const byId = new Map(todos.map(todo => [todo.id, todo]))
+    const matched = todos.filter(todo => todo.title.toLowerCase().includes(needle))
+    const includeIds = new Set(matched.map(todo => todo.id))
+    for (const todo of matched) {
+      if (todo.parentId) {
+        includeIds.add(todo.parentId)
+      } else {
+        for (const child of todos) {
+          if (child.parentId === todo.id) includeIds.add(child.id)
+        }
+      }
+    }
+    return todos.filter(todo => includeIds.has(todo.id) || (todo.parentId && includeIds.has(todo.parentId)))
+  }
+
+  function listSubtodosByParent(parentId) {
+    if (!parentId) return []
+    const rows = db.prepare(
+      `SELECT * FROM todos WHERE parent_id = ? ORDER BY sort_order ASC, created_at ASC`
+    ).all(parentId)
+    return rows.map(rowToTodo)
+  }
+
+  function archiveTodo(id) {
+    const existing = stmts.getById.get(id)
+    if (!existing) throw new Error('todo_not_found')
+    if (existing.archived_at != null) return rowToTodo(existing)
+    const now = Date.now()
+    db.prepare(`UPDATE todos SET archived_at = ?, updated_at = ? WHERE id = ?`).run(now, now, id)
+    return rowToTodo(stmts.getById.get(id))
+  }
+
+  function unarchiveTodo(id) {
+    const existing = stmts.getById.get(id)
+    if (!existing) throw new Error('todo_not_found')
+    if (existing.archived_at == null) return rowToTodo(existing)
+    const now = Date.now()
+    db.prepare(`UPDATE todos SET archived_at = NULL, updated_at = ? WHERE id = ?`).run(now, id)
+    return rowToTodo(stmts.getById.get(id))
+  }
+
+  /**
+   * Preview/describe the impact of merging. 不做任何修改。
+   * 返回 { targetId, sources[], movedChildren, movedComments, movedSessions, movedSessionLogs,
+   *        movedCoverage, movedTranscripts, proposedTitle }
+   */
+  function describeMergeTodos({ targetId, sourceIds, titleStrategy = 'keep_target', manualTitle } = {}) {
+    if (!targetId) throw new Error('target_required')
+    if (!Array.isArray(sourceIds) || sourceIds.length === 0) throw new Error('sources_required')
+    const target = stmts.getById.get(targetId)
+    if (!target) throw new Error('target_not_found')
+    const uniqueSources = [...new Set(sourceIds)].filter((s) => s && s !== targetId)
+    if (uniqueSources.length === 0) throw new Error('sources_required')
+    const sources = []
+    for (const sid of uniqueSources) {
+      const row = stmts.getById.get(sid)
+      if (!row) throw new Error(`source_not_found:${sid}`)
+      sources.push(row)
+    }
+    const countOne = (sql, id) => db.prepare(sql).get(id)?.n || 0
+    let movedChildren = 0
+    let movedComments = 0
+    let movedSessions = 0
+    let movedSessionLogs = 0
+    let movedCoverage = 0
+    let movedTranscripts = 0
+    for (const src of sources) {
+      movedChildren += countOne(`SELECT COUNT(*) AS n FROM todos WHERE parent_id = ?`, src.id)
+      movedComments += countOne(`SELECT COUNT(*) AS n FROM comments WHERE todo_id = ?`, src.id)
+      movedSessions += normalizeAiSessions(src.ai_session ? JSON.parse(src.ai_session) : null).length
+      movedSessionLogs += countOne(`SELECT COUNT(*) AS n FROM ai_session_log WHERE todo_id = ?`, src.id)
+      movedCoverage += countOne(`SELECT COUNT(*) AS n FROM wiki_todo_coverage WHERE todo_id = ?`, src.id)
+      movedTranscripts += countOne(`SELECT COUNT(*) AS n FROM transcript_files WHERE bound_todo_id = ?`, src.id)
+    }
+    let proposedTitle = target.title
+    if (titleStrategy === 'concat') {
+      proposedTitle = [target.title, ...sources.map((s) => s.title)].join(' + ')
+    } else if (titleStrategy === 'manual') {
+      if (!manualTitle || !String(manualTitle).trim()) throw new Error('manual_title_required')
+      proposedTitle = String(manualTitle).trim()
+    }
+    return {
+      targetId,
+      target: { id: target.id, title: target.title },
+      sources: sources.map((s) => ({ id: s.id, title: s.title })),
+      movedChildren,
+      movedComments,
+      movedSessions,
+      movedSessionLogs,
+      movedCoverage,
+      movedTranscripts,
+      proposedTitle,
+    }
+  }
+
+  /**
+   * 事务化合并：把 sourceIds 的子任务、评论、ai_session JSON、ai_session_log、wiki_coverage、
+   * transcript_files 全部迁移到 targetId，然后删除 source todo。
+   * titleStrategy: 'keep_target' | 'concat' | 'manual'（manual 需 manualTitle）
+   */
+  function mergeTodos({ targetId, sourceIds, titleStrategy = 'keep_target', manualTitle } = {}) {
+    // 先复用 describe 做校验
+    const preview = describeMergeTodos({ targetId, sourceIds, titleStrategy, manualTitle })
+    const now = Date.now()
+    const run = db.transaction(() => {
+      const targetRow = stmts.getById.get(targetId)
+      let mergedSessions = normalizeAiSessions(targetRow.ai_session ? JSON.parse(targetRow.ai_session) : null)
+      for (const src of preview.sources) {
+        db.prepare(`UPDATE todos SET parent_id = ?, updated_at = ? WHERE parent_id = ?`).run(targetId, now, src.id)
+        db.prepare(`UPDATE comments SET todo_id = ? WHERE todo_id = ?`).run(targetId, src.id)
+        db.prepare(`UPDATE ai_session_log SET todo_id = ? WHERE todo_id = ?`).run(targetId, src.id)
+        db.prepare(`UPDATE wiki_todo_coverage SET todo_id = ? WHERE todo_id = ?`).run(targetId, src.id)
+        db.prepare(`UPDATE transcript_files SET bound_todo_id = ? WHERE bound_todo_id = ?`).run(targetId, src.id)
+        // ai_session JSON 合并
+        const srcRow = stmts.getById.get(src.id)
+        const srcSessions = normalizeAiSessions(srcRow.ai_session ? JSON.parse(srcRow.ai_session) : null)
+        mergedSessions = [...mergedSessions, ...srcSessions]
+        stmts.deleteById.run(src.id)
+      }
+      // 写回合并后的 ai_session + 可能的新标题
+      const fields = ['ai_session = ?', 'updated_at = ?']
+      const params = [mergedSessions.length ? JSON.stringify(mergedSessions) : null, now]
+      if (preview.proposedTitle !== targetRow.title) {
+        fields.push('title = ?')
+        params.push(preview.proposedTitle)
+      }
+      params.push(targetId)
+      db.prepare(`UPDATE todos SET ${fields.join(', ')} WHERE id = ?`).run(...params)
+    })
+    run()
+    return {
+      ...preview,
+      ok: true,
+      resultingTodo: rowToTodo(stmts.getById.get(targetId)),
+    }
+  }
+
+  /**
+   * 对一组 todo id 批量 patch。Patch 字段限白名单（quadrant/status/archived/dueDate）。
+   * archived: true → 设置 archived_at=now；false → 清空；不传则不改。
+   */
+  function bulkUpdateTodos({ ids, patch } = {}) {
+    if (!Array.isArray(ids) || ids.length === 0) throw new Error('ids_required')
+    if (!patch || typeof patch !== 'object') throw new Error('patch_required')
+    const allowed = ['quadrant', 'status', 'archived', 'dueDate']
+    const keys = Object.keys(patch).filter((k) => allowed.includes(k))
+    if (keys.length === 0) throw new Error('patch_empty')
+    const now = Date.now()
+    const changed = []
+    const run = db.transaction(() => {
+      for (const id of ids) {
+        const existing = stmts.getById.get(id)
+        if (!existing) continue
+        const sets = []
+        const params = []
+        if ('quadrant' in patch) {
+          sets.push('quadrant = ?')
+          params.push(Number(patch.quadrant))
+        }
+        if ('status' in patch) {
+          sets.push('status = ?')
+          params.push(String(patch.status))
+          if (patch.status === 'done') {
+            sets.push('completed_at = ?')
+            params.push(now)
+          }
+        }
+        if ('dueDate' in patch) {
+          sets.push('due_date = ?')
+          params.push(patch.dueDate == null ? null : Number(patch.dueDate))
+        }
+        if ('archived' in patch) {
+          sets.push('archived_at = ?')
+          params.push(patch.archived ? now : null)
+        }
+        sets.push('updated_at = ?')
+        params.push(now)
+        params.push(id)
+        db.prepare(`UPDATE todos SET ${sets.join(', ')} WHERE id = ?`).run(...params)
+        changed.push(id)
+      }
+    })
+    run()
+    return { changedIds: changed, count: changed.length }
+  }
+
+  function listCompletedTodos({ since, until }) {
+    const rows = db.prepare(`
+      SELECT * FROM todos
+      WHERE status = 'done'
+        AND completed_at IS NOT NULL
+        AND completed_at >= ?
+        AND completed_at < ?
+      ORDER BY completed_at DESC
+    `).all(Number(since), Number(until))
+    return rows.map(rowToTodo)
+  }
+
+  function countMissedInRange({ since, until }) {
+    // 循环任务过期：status='missed'，在 sweepRecurring 里用 updated_at 标记时间
+    const row = db.prepare(`
+      SELECT COUNT(*) AS n FROM todos
+      WHERE status = 'missed'
+        AND updated_at >= ?
+        AND updated_at < ?
+    `).get(Number(since), Number(until))
+    return row?.n || 0
+  }
+
+  const commentStmts = {
+    insert: db.prepare(`INSERT INTO comments (id, todo_id, content, created_at) VALUES (@id, @todo_id, @content, @created_at)`),
+    listByTodo: db.prepare(`SELECT * FROM comments WHERE todo_id = ? ORDER BY created_at ASC`),
+    deleteById: db.prepare(`DELETE FROM comments WHERE id = ?`),
+    getById: db.prepare(`SELECT * FROM comments WHERE id = ?`),
+  }
+
+  function addComment(todoId, content) {
+    const row = {
+      id: randomUUID(),
+      todo_id: todoId,
+      content,
+      created_at: Date.now(),
+    }
+    commentStmts.insert.run(row)
+    return { id: row.id, todoId: row.todo_id, content: row.content, createdAt: row.created_at }
+  }
+
+  function listComments(todoId) {
+    return commentStmts.listByTodo.all(todoId).map(r => ({
+      id: r.id,
+      todoId: r.todo_id,
+      content: r.content,
+      createdAt: r.created_at,
+    }))
+  }
+
+  function deleteComment(id) {
+    commentStmts.deleteById.run(id)
+  }
+
+  function getComment(id) {
+    const r = commentStmts.getById.get(id)
+    if (!r) return null
+    return { id: r.id, todoId: r.todo_id, content: r.content, createdAt: r.created_at }
+  }
+
+  const aiLogStmts = {
+    insert: db.prepare(`
+      INSERT OR REPLACE INTO ai_session_log
+        (id, todo_id, tool, quadrant, status, exit_code, started_at, completed_at, duration_ms)
+      VALUES
+        (@id, @todo_id, @tool, @quadrant, @status, @exit_code, @started_at, @completed_at, @duration_ms)
+    `),
+    listSince: db.prepare(`SELECT * FROM ai_session_log WHERE completed_at >= ? AND completed_at < ? ORDER BY completed_at DESC`),
+    listInWindow: db.prepare(`
+      SELECT id, todo_id, tool, started_at, completed_at, duration_ms
+      FROM ai_session_log
+      WHERE tool = ? AND started_at BETWEEN ? AND ?
+    `),
+  }
+
+  function insertSessionLog(row) {
+    aiLogStmts.insert.run({
+      id: row.id,
+      todo_id: row.todoId,
+      tool: row.tool,
+      quadrant: Number(row.quadrant) || 0,
+      status: row.status,
+      exit_code: row.exitCode ?? null,
+      started_at: row.startedAt,
+      completed_at: row.completedAt,
+      duration_ms: Math.max(0, row.completedAt - row.startedAt),
+    })
+  }
+
+  function querySessionStats({ since, until = Date.now() } = {}) {
+    const rows = aiLogStmts.listSince.all(since, until)
+    const stats = {
+      total: rows.length,
+      byStatus: { done: 0, failed: 0, stopped: 0 },
+      byTool: { claude: 0, codex: 0, cursor: 0 },
+      byQuadrant: { 1: 0, 2: 0, 3: 0, 4: 0 },
+      totalDurationMs: 0,
+      avgDurationMs: 0,
+      timeline: [],
+    }
+    if (!rows.length) return stats
+    const buckets = new Map()
+    const bucketSize = (until - since) > 7 * 86400_000 ? 86400_000 : 3600_000
+    for (const r of rows) {
+      stats.byStatus[r.status] = (stats.byStatus[r.status] || 0) + 1
+      stats.byTool[r.tool] = (stats.byTool[r.tool] || 0) + 1
+      stats.byQuadrant[r.quadrant] = (stats.byQuadrant[r.quadrant] || 0) + 1
+      stats.totalDurationMs += r.duration_ms
+      const bucket = Math.floor(r.completed_at / bucketSize) * bucketSize
+      buckets.set(bucket, (buckets.get(bucket) || 0) + 1)
+    }
+    stats.avgDurationMs = Math.round(stats.totalDurationMs / rows.length)
+    stats.timeline = [...buckets.entries()]
+      .sort((a, b) => a[0] - b[0])
+      .map(([t, count]) => ({ t, count }))
+    return stats
+  }
+
+  const tfStmts = {
+    getByPath: db.prepare(`SELECT * FROM transcript_files WHERE jsonl_path = ?`),
+    listAllPaths: db.prepare(`SELECT id, jsonl_path, size, mtime FROM transcript_files`),
+    upsert: db.prepare(`
+      INSERT INTO transcript_files (tool, native_id, cwd, jsonl_path, size, mtime, started_at, ended_at, first_user_prompt, turn_count, bound_todo_id, indexed_at, input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens, primary_model, active_ms)
+      VALUES (@tool, @native_id, @cwd, @jsonl_path, @size, @mtime, @started_at, @ended_at, @first_user_prompt, @turn_count, @bound_todo_id, @indexed_at, @input_tokens, @output_tokens, @cache_read_tokens, @cache_creation_tokens, @primary_model, @active_ms)
+      ON CONFLICT(jsonl_path) DO UPDATE SET
+        tool=excluded.tool,
+        native_id=excluded.native_id,
+        cwd=excluded.cwd,
+        size=excluded.size,
+        mtime=excluded.mtime,
+        started_at=excluded.started_at,
+        ended_at=excluded.ended_at,
+        first_user_prompt=excluded.first_user_prompt,
+        turn_count=excluded.turn_count,
+        indexed_at=excluded.indexed_at,
+        input_tokens=excluded.input_tokens,
+        output_tokens=excluded.output_tokens,
+        cache_read_tokens=excluded.cache_read_tokens,
+        cache_creation_tokens=excluded.cache_creation_tokens,
+        primary_model=excluded.primary_model,
+        active_ms=excluded.active_ms
+    `),
+    deleteByPath: db.prepare(`DELETE FROM transcript_files WHERE jsonl_path = ?`),
+    getById: db.prepare(`SELECT * FROM transcript_files WHERE id = ?`),
+    setBound: db.prepare(`UPDATE transcript_files SET bound_todo_id = ? WHERE id = ?`),
+    findByNative: db.prepare(`SELECT * FROM transcript_files WHERE native_id = ? AND tool = ?`),
+    countUnbound: db.prepare(`SELECT COUNT(*) AS n FROM transcript_files WHERE bound_todo_id IS NULL`),
+    listUnboundForMatching: db.prepare(`SELECT * FROM transcript_files WHERE bound_todo_id IS NULL`),
+  }
+  const ftsStmts = ftsAvailable ? {
+    deleteByFile: db.prepare(`DELETE FROM transcript_fts WHERE file_id = ?`),
+    insert: db.prepare(`INSERT INTO transcript_fts (content, role, file_id) VALUES (?, ?, ?)`),
+  } : null
+
+  function upsertTranscriptFile(row) {
+    tfStmts.upsert.run({
+      tool: row.tool,
+      native_id: row.nativeId ?? null,
+      cwd: row.cwd ?? null,
+      jsonl_path: row.jsonlPath,
+      size: row.size,
+      mtime: row.mtime,
+      started_at: row.startedAt ?? null,
+      ended_at: row.endedAt ?? null,
+      first_user_prompt: row.firstUserPrompt ?? null,
+      turn_count: row.turnCount ?? 0,
+      bound_todo_id: row.boundTodoId ?? null,
+      indexed_at: Date.now(),
+      input_tokens: row.inputTokens ?? null,
+      output_tokens: row.outputTokens ?? null,
+      cache_read_tokens: row.cacheReadTokens ?? null,
+      cache_creation_tokens: row.cacheCreationTokens ?? null,
+      primary_model: row.primaryModel ?? null,
+      active_ms: row.activeMs ?? null,
+    })
+    return tfStmts.getByPath.get(row.jsonlPath)
+  }
+
+  function deleteTranscriptFile(jsonlPath) {
+    const existing = tfStmts.getByPath.get(jsonlPath)
+    if (!existing) return
+    if (ftsStmts) ftsStmts.deleteByFile.run(existing.id)
+    tfStmts.deleteByPath.run(jsonlPath)
+  }
+
+  function writeFtsTurns(fileId, turns) {
+    if (!ftsStmts) return
+    const tx = db.transaction(() => {
+      ftsStmts.deleteByFile.run(fileId)
+      for (const t of turns) {
+        if (!t?.content) continue
+        ftsStmts.insert.run(String(t.content), String(t.role || ''), fileId)
+      }
+    })
+    tx()
+  }
+
+  function searchTranscripts({ q, tool, cwd, since, unboundOnly, limit = 50, offset = 0 } = {}) {
+    const where = []
+    const params = []
+    if (tool) { where.push('tf.tool = ?'); params.push(tool) }
+    if (cwd) { where.push('tf.cwd = ?'); params.push(cwd) }
+    if (since) { where.push('tf.started_at >= ?'); params.push(since) }
+    if (unboundOnly) where.push('tf.bound_todo_id IS NULL')
+
+    if (q && ftsAvailable) {
+      // trigram tokenizer 要求 ≥3 字才能走 MATCH；<3 字用 LIKE 兜底扫 FTS 的 content 列
+      if (q.length < 3) {
+        const like = `%${q.replace(/[\\%_]/g, s => '\\' + s)}%`
+        where.push(`tf.id IN (SELECT file_id FROM transcript_fts WHERE content LIKE ? ESCAPE '\\')`)
+        params.push(like)
+        const whereSql = where.length ? `WHERE ${where.join(' AND ')}` : ''
+        const total = db.prepare(`SELECT COUNT(*) AS n FROM transcript_files tf ${whereSql}`).get(...params).n
+        const rows = db.prepare(`
+          SELECT tf.*, (
+            SELECT SUBSTR(content, MAX(1, INSTR(content, ?) - 16), 64)
+            FROM transcript_fts WHERE file_id = tf.id AND content LIKE ? ESCAPE '\\' LIMIT 1
+          ) AS snippet
+          FROM transcript_files tf
+          ${whereSql}
+          ORDER BY tf.started_at DESC
+          LIMIT ? OFFSET ?
+        `).all(q, like, ...params, limit, offset)
+        return { total, items: rows }
+      }
+      const ftsQuery = q.replace(/"/g, '""')
+      where.push('tf.id IN (SELECT file_id FROM transcript_fts WHERE transcript_fts MATCH ?)')
+      params.push(`"${ftsQuery}"`)
+      const whereSql = where.length ? `WHERE ${where.join(' AND ')}` : ''
+      const total = db.prepare(`SELECT COUNT(*) AS n FROM transcript_files tf ${whereSql}`).get(...params).n
+      const rows = db.prepare(`
+        SELECT tf.*, (
+          SELECT snippet(transcript_fts, 0, '<mark>', '</mark>', '…', 16)
+          FROM transcript_fts WHERE file_id = tf.id AND transcript_fts MATCH ? LIMIT 1
+        ) AS snippet
+        FROM transcript_files tf
+        ${whereSql}
+        ORDER BY tf.started_at DESC NULLS LAST
+        LIMIT ? OFFSET ?
+      `).all(`"${ftsQuery}"`, ...params, limit, offset)
+      return { total, items: rows }
+    }
+
+    if (q && !ftsAvailable) {
+      where.push('LOWER(tf.first_user_prompt) LIKE ?')
+      params.push(`%${q.toLowerCase()}%`)
+    }
+    const whereSql = where.length ? `WHERE ${where.join(' AND ')}` : ''
+    const total = db.prepare(`SELECT COUNT(*) AS n FROM transcript_files tf ${whereSql}`).get(...params).n
+    const rows = db.prepare(`
+      SELECT tf.*, NULL AS snippet
+      FROM transcript_files tf
+      ${whereSql}
+      ORDER BY tf.started_at DESC
+      LIMIT ? OFFSET ?
+    `).all(...params, limit, offset)
+    return { total, items: rows }
+  }
+
+  const ptStmts = {
+    list: db.prepare(`SELECT * FROM prompt_templates ORDER BY builtin DESC, sort_order ASC, created_at ASC`),
+    get: db.prepare(`SELECT * FROM prompt_templates WHERE id = ?`),
+    insert: db.prepare(`INSERT INTO prompt_templates (id, name, description, content, builtin, sort_order, created_at, updated_at) VALUES (@id, @name, @description, @content, @builtin, @sort_order, @created_at, @updated_at)`),
+    update: db.prepare(`UPDATE prompt_templates SET name = @name, description = @description, content = @content, sort_order = @sort_order, updated_at = @updated_at WHERE id = @id`),
+    delete: db.prepare(`DELETE FROM prompt_templates WHERE id = ?`),
+    countAll: db.prepare(`SELECT COUNT(*) AS n FROM prompt_templates`),
+  }
+
+  function rowToTemplate(r) {
+    if (!r) return null
+    return {
+      id: r.id,
+      name: r.name,
+      description: r.description,
+      content: r.content,
+      builtin: !!r.builtin,
+      sortOrder: r.sort_order,
+      createdAt: r.created_at,
+      updatedAt: r.updated_at,
+    }
+  }
+
+  function listTemplates() { return ptStmts.list.all().map(rowToTemplate) }
+  function getTemplate(id) { return rowToTemplate(ptStmts.get.get(id)) }
+  function createTemplate(data) {
+    const now = Date.now()
+    const row = {
+      id: randomUUID(),
+      name: data.name || '未命名模板',
+      description: data.description || '',
+      content: data.content || '',
+      builtin: data.builtin ? 1 : 0,
+      sort_order: Number.isFinite(data.sortOrder) ? data.sortOrder : now,
+      created_at: now,
+      updated_at: now,
+    }
+    ptStmts.insert.run(row)
+    return rowToTemplate(ptStmts.get.get(row.id))
+  }
+  function updateTemplate(id, patch) {
+    const existing = ptStmts.get.get(id)
+    if (!existing) return null
+    if (existing.builtin) {
+      throw new Error('builtin_template_readonly')
+    }
+    ptStmts.update.run({
+      id,
+      name: patch.name ?? existing.name,
+      description: patch.description ?? existing.description,
+      content: patch.content ?? existing.content,
+      sort_order: Number.isFinite(patch.sortOrder) ? patch.sortOrder : existing.sort_order,
+      updated_at: Date.now(),
+    })
+    return rowToTemplate(ptStmts.get.get(id))
+  }
+  function deleteTemplate(id) {
+    const existing = ptStmts.get.get(id)
+    if (!existing) return
+    if (existing.builtin) throw new Error('builtin_template_readonly')
+    ptStmts.delete.run(id)
+  }
+
+  // ─── pending_questions ──────────────────────────────────────────
+  // 配合 ask_user MCP 工具：AI 在 PTY 里发起问题时，写一行 pending；
+  // 用户在 OpenClaw（微信）回复后，通过 submit_user_reply 回填。
+  const pqStmts = {
+    insert: db.prepare(`
+      INSERT INTO pending_questions (
+        ticket, session_id, todo_id, question, options_json,
+        status, created_at, timeout_ms
+      ) VALUES (
+        @ticket, @session_id, @todo_id, @question, @options_json,
+        'pending', @created_at, @timeout_ms
+      )
+    `),
+    getByTicket: db.prepare(`SELECT * FROM pending_questions WHERE ticket = ?`),
+    listPending: db.prepare(`
+      SELECT * FROM pending_questions
+      WHERE status = 'pending'
+      ORDER BY created_at DESC
+    `),
+    listLatestPending: db.prepare(`
+      SELECT * FROM pending_questions
+      WHERE status = 'pending'
+      ORDER BY created_at DESC
+      LIMIT 1
+    `),
+    setAnswered: db.prepare(`
+      UPDATE pending_questions
+      SET status = 'answered', answer_text = ?, chosen_index = ?, answered_at = ?
+      WHERE ticket = ? AND status = 'pending'
+    `),
+    setStatus: db.prepare(`
+      UPDATE pending_questions
+      SET status = ?, answered_at = ?
+      WHERE ticket = ? AND status = 'pending'
+    `),
+    sweepExpired: db.prepare(`
+      UPDATE pending_questions
+      SET status = 'timeout', answered_at = ?
+      WHERE status = 'pending' AND (created_at + timeout_ms) < ?
+    `),
+  }
+
+  function rowToPending(r) {
+    if (!r) return null
+    let options = []
+    try { options = JSON.parse(r.options_json) } catch {}
+    return {
+      ticket: r.ticket,
+      sessionId: r.session_id,
+      todoId: r.todo_id,
+      question: r.question,
+      options,
+      status: r.status,
+      answerText: r.answer_text,
+      chosenIndex: r.chosen_index,
+      createdAt: r.created_at,
+      answeredAt: r.answered_at,
+      timeoutMs: r.timeout_ms,
+    }
+  }
+
+  function createPendingQuestion({ ticket, sessionId, todoId, question, options, timeoutMs }) {
+    if (!ticket) throw new Error('ticket_required')
+    if (!sessionId) throw new Error('session_id_required')
+    if (!question) throw new Error('question_required')
+    if (!Array.isArray(options) || options.length === 0) throw new Error('options_required')
+    pqStmts.insert.run({
+      ticket,
+      session_id: sessionId,
+      todo_id: todoId || null,
+      question,
+      options_json: JSON.stringify(options),
+      created_at: Date.now(),
+      timeout_ms: Number(timeoutMs) || 600000,
+    })
+    return rowToPending(pqStmts.getByTicket.get(ticket))
+  }
+
+  function getPendingQuestion(ticket) {
+    return rowToPending(pqStmts.getByTicket.get(ticket))
+  }
+
+  function listPendingQuestions() {
+    return pqStmts.listPending.all().map(rowToPending)
+  }
+
+  function getLatestPendingQuestion() {
+    return rowToPending(pqStmts.listLatestPending.get())
+  }
+
+  function answerPendingQuestion(ticket, { answerText, chosenIndex }) {
+    const r = pqStmts.setAnswered.run(
+      answerText ?? null,
+      Number.isInteger(chosenIndex) ? chosenIndex : null,
+      Date.now(),
+      ticket,
+    )
+    if (r.changes === 0) return null
+    return rowToPending(pqStmts.getByTicket.get(ticket))
+  }
+
+  function setPendingStatus(ticket, status) {
+    if (!['timeout', 'cancelled'].includes(status)) throw new Error('invalid_status')
+    pqStmts.setStatus.run(status, Date.now(), ticket)
+    return rowToPending(pqStmts.getByTicket.get(ticket))
+  }
+
+  function sweepExpiredPendingQuestions() {
+    const r = pqStmts.sweepExpired.run(Date.now(), Date.now())
+    return r.changes
+  }
+
+  function seedBuiltinTemplatesIfEmpty() {
+    if (ptStmts.countAll.get().n > 0) return
+    const now = Date.now()
+    const seeds = [
+      {
+        name: 'Brainstorm（脑爆）',
+        description: '先脑爆方向，不急着动手',
+        content: '请先不要直接动手实现。先针对下面的任务 brainstorm：\n- 列出 2-3 种可选方案，说明优缺点\n- 指出风险点与需要用户拍板的关键决策\n- 明确验收标准\n\n在我确认方案后再进入实现。',
+      },
+      {
+        name: 'Bug 修复',
+        description: '复现 → 定位 → 最小用例 → 修复 → 回归',
+        content: '按 bug 修复流程处理下面的问题：\n1. 先复现（给出复现步骤和实际 vs 预期）\n2. 定位根因（不要过早修改代码）\n3. 写一个能复现该 bug 的最小用例（如果有测试框架）\n4. 修复根因，不是修现象\n5. 回归：跑相关测试；考虑同类 bug 是否还存在',
+      },
+      {
+        name: '重构',
+        description: '先读懂 → 列出影响面 → 小步重构',
+        content: '按照小步重构原则处理下面的任务：\n1. 先通读相关代码，复述你的理解\n2. 列出此次重构的影响面（调用方 / 测试 / 类型）\n3. 每一步只改一件事，保持可运行\n4. 每步后跑一次测试（如果有）\n5. 不要顺手加功能、不要引入新抽象，除非当前任务要求',
+      },
+      {
+        name: '写测试',
+        description: 'TDD：红 → 绿 → 重构',
+        content: '用 TDD 的方式处理下面的任务：\n1. 先列出测试矩阵（输入 × 场景）\n2. 先写一个最简失败用例（红）\n3. 用最小改动让它通过（绿）\n4. 重构（保持绿）\n5. 重复 2-4 直到覆盖矩阵\n不 mock 真实依赖（除非跨网络/支付等）。',
+      },
+      {
+        name: '代码评审',
+        description: '只评审，不改代码',
+        content: '请只做代码评审，不要修改代码。按下面的维度给出具体反馈：\n- 可读性：命名、结构、注释\n- 正确性：边界、错误处理、并发\n- 安全性：注入、鉴权、敏感数据\n- 性能：明显的 N+1 / 无谓复制\n- 简洁性：是否有过度设计 / 可删除的冗余\n每条反馈给出文件:行号 + 建议。',
+      },
+    ]
+    seeds.forEach((s, i) => {
+      ptStmts.insert.run({
+        id: randomUUID(),
+        name: s.name,
+        description: s.description,
+        content: s.content,
+        builtin: 1,
+        sort_order: i,
+        created_at: now,
+        updated_at: now,
+      })
+    })
+  }
+  seedBuiltinTemplatesIfEmpty()
+
+  const wikiStmts = {
+    insertRun: db.prepare(`
+      INSERT INTO wiki_runs (started_at, todo_count, dry_run)
+      VALUES (?, ?, ?)
+    `),
+    completeRun: db.prepare(`
+      UPDATE wiki_runs SET completed_at = ?, exit_code = ?, note = ?
+      WHERE id = ?
+    `),
+    failRun: db.prepare(`
+      UPDATE wiki_runs SET completed_at = ?, exit_code = ?, error = ?
+      WHERE id = ?
+    `),
+    listRuns: db.prepare(`
+      SELECT * FROM wiki_runs ORDER BY started_at DESC LIMIT ?
+    `),
+    orphanRuns: db.prepare(`
+      SELECT * FROM wiki_runs WHERE completed_at IS NULL
+    `),
+    upsertCoverage: db.prepare(`
+      INSERT INTO wiki_todo_coverage (wiki_run_id, todo_id, source_path, llm_applied)
+      VALUES (?, ?, ?, ?)
+      ON CONFLICT(wiki_run_id, todo_id) DO UPDATE SET
+        source_path = excluded.source_path,
+        llm_applied = excluded.llm_applied
+    `),
+    markApplied: db.prepare(`
+      UPDATE wiki_todo_coverage SET llm_applied = 1 WHERE wiki_run_id = ?
+    `),
+    coverageForTodo: db.prepare(`
+      SELECT * FROM wiki_todo_coverage WHERE todo_id = ? ORDER BY wiki_run_id DESC
+    `),
+    unappliedDoneTodos: db.prepare(`
+      SELECT t.* FROM todos t
+      WHERE t.status = 'done'
+        AND NOT EXISTS (
+          SELECT 1 FROM wiki_todo_coverage c
+          WHERE c.todo_id = t.id AND c.llm_applied = 1
+        )
+      ORDER BY t.updated_at DESC
+    `),
+  }
+
+  function createWikiRun({ todoCount = 0, dryRun = 0 } = {}) {
+    const now = Date.now()
+    const info = wikiStmts.insertRun.run(now, Number(todoCount) || 0, dryRun ? 1 : 0)
+    return { id: info.lastInsertRowid, started_at: now, completed_at: null }
+  }
+  function completeWikiRun(id, { exitCode = 0, note = '' } = {}) {
+    wikiStmts.completeRun.run(Date.now(), exitCode, note || '', id)
+  }
+  function failWikiRun(id, errorMsg) {
+    wikiStmts.failRun.run(Date.now(), -1, String(errorMsg || 'unknown'), id)
+  }
+  function listWikiRuns({ limit = 20 } = {}) {
+    return wikiStmts.listRuns.all(Math.max(1, Math.min(200, limit)))
+  }
+  function findOrphanWikiRuns() {
+    return wikiStmts.orphanRuns.all()
+  }
+  function upsertWikiCoverage(runId, todoId, sourcePath, llmApplied) {
+    wikiStmts.upsertCoverage.run(runId, todoId, sourcePath || null, llmApplied ? 1 : 0)
+  }
+  function markCoverageApplied(runId) {
+    wikiStmts.markApplied.run(runId)
+  }
+  function listCoverageForTodo(todoId) {
+    return wikiStmts.coverageForTodo.all(todoId)
+  }
+  function listUnappliedDoneTodos() {
+    return wikiStmts.unappliedDoneTodos.all().map(rowToTodo)
+  }
+
+  const ruleStmts = {
+    insert: db.prepare(`
+      INSERT INTO recurring_rules
+        (id, title, description, quadrant, work_dir, brainstorm, applied_template_ids, subtodos, frequency, weekdays, month_days, active, last_generated_date, created_at, updated_at)
+      VALUES
+        (@id, @title, @description, @quadrant, @work_dir, @brainstorm, @applied_template_ids, @subtodos, @frequency, @weekdays, @month_days, @active, @last_generated_date, @created_at, @updated_at)
+    `),
+    get: db.prepare(`SELECT * FROM recurring_rules WHERE id = ?`),
+    list: db.prepare(`SELECT * FROM recurring_rules ORDER BY created_at DESC`),
+    listActive: db.prepare(`SELECT * FROM recurring_rules WHERE active = 1`),
+    update: db.prepare(`
+      UPDATE recurring_rules SET
+        title = @title,
+        description = @description,
+        quadrant = @quadrant,
+        work_dir = @work_dir,
+        brainstorm = @brainstorm,
+        applied_template_ids = @applied_template_ids,
+        subtodos = @subtodos,
+        frequency = @frequency,
+        weekdays = @weekdays,
+        month_days = @month_days,
+        updated_at = @updated_at
+      WHERE id = @id
+    `),
+    setActive: db.prepare(`UPDATE recurring_rules SET active = ?, updated_at = ? WHERE id = ?`),
+    setLastGenerated: db.prepare(`UPDATE recurring_rules SET last_generated_date = ?, updated_at = ? WHERE id = ?`),
+    delete: db.prepare(`DELETE FROM recurring_rules WHERE id = ?`),
+    unlinkInstances: db.prepare(`UPDATE todos SET recurring_rule_id = NULL WHERE recurring_rule_id = ?`),
+  }
+
+  function normalizeRuleInput(data) {
+    const frequency = data.frequency
+    if (!['daily', 'weekly', 'monthly'].includes(frequency)) {
+      throw new Error('invalid_frequency')
+    }
+    let weekdays = null
+    let monthDays = null
+    if (frequency === 'weekly') {
+      const arr = Array.isArray(data.weekdays) ? data.weekdays.filter(n => Number.isInteger(n) && n >= 0 && n <= 6) : []
+      if (!arr.length) throw new Error('weekdays_required')
+      weekdays = [...new Set(arr)].sort((a, b) => a - b)
+    }
+    if (frequency === 'monthly') {
+      const arr = Array.isArray(data.monthDays) ? data.monthDays.filter(n => Number.isInteger(n) && n >= 1 && n <= 31) : []
+      if (!arr.length) throw new Error('month_days_required')
+      monthDays = [...new Set(arr)].sort((a, b) => a - b)
+    }
+    const q = Number(data.quadrant)
+    if (![1, 2, 3, 4].includes(q)) throw new Error('invalid_quadrant')
+    return { frequency, weekdays, monthDays, quadrant: q }
+  }
+
+  function instantiateRule(rule, now) {
+    const today = todayStr(now)
+    const due = endOfDayMs(now)
+    const parent = createTodo({
+      title: rule.title,
+      description: rule.description,
+      quadrant: rule.quadrant,
+      status: 'todo',
+      dueDate: due,
+      workDir: rule.workDir ?? null,
+      brainstorm: !!rule.brainstorm,
+      appliedTemplateIds: rule.appliedTemplateIds || [],
+      recurringRuleId: rule.id,
+      instanceDate: today,
+    })
+    for (const st of (rule.subtodos || [])) {
+      if (!st || !st.title) continue
+      createTodo({
+        title: st.title,
+        description: st.description || '',
+        status: 'todo',
+        dueDate: due,
+        parentId: parent.id,
+        recurringRuleId: rule.id,
+        instanceDate: today,
+      })
+    }
+    return parent
+  }
+
+  function createRecurringRule(data) {
+    const { frequency, weekdays, monthDays, quadrant } = normalizeRuleInput(data)
+    if (!data.title || typeof data.title !== 'string') throw new Error('title_required')
+    const now = Date.now()
+    const today = todayStr(now)
+    const rule = {
+      id: randomUUID(),
+      title: data.title.trim(),
+      description: data.description || '',
+      quadrant,
+      work_dir: data.workDir || null,
+      brainstorm: data.brainstorm ? 1 : 0,
+      applied_template_ids: JSON.stringify(Array.isArray(data.appliedTemplateIds) ? data.appliedTemplateIds : []),
+      subtodos: JSON.stringify(Array.isArray(data.subtodos) ? data.subtodos.map(s => ({
+        title: String(s.title || '').trim(),
+        description: String(s.description || ''),
+      })).filter(s => s.title) : []),
+      frequency,
+      weekdays: weekdays ? JSON.stringify(weekdays) : null,
+      month_days: monthDays ? JSON.stringify(monthDays) : null,
+      active: 1,
+      last_generated_date: today,
+      created_at: now,
+      updated_at: now,
+    }
+    ruleStmts.insert.run(rule)
+    const ruleObj = rowToRule(ruleStmts.get.get(rule.id))
+    let firstInstance = null
+    if (ruleShouldProduceOn(ruleObj, today)) {
+      firstInstance = instantiateRule(ruleObj, now)
+    }
+    return { rule: ruleObj, firstInstance }
+  }
+
+  function updateRecurringRule(id, patch) {
+    const existing = rowToRule(ruleStmts.get.get(id))
+    if (!existing) return null
+    const merged = {
+      title: patch.title ?? existing.title,
+      description: patch.description ?? existing.description,
+      quadrant: patch.quadrant ?? existing.quadrant,
+      workDir: patch.workDir !== undefined ? patch.workDir : existing.workDir,
+      brainstorm: patch.brainstorm !== undefined ? !!patch.brainstorm : existing.brainstorm,
+      appliedTemplateIds: patch.appliedTemplateIds !== undefined ? patch.appliedTemplateIds : existing.appliedTemplateIds,
+      subtodos: patch.subtodos !== undefined ? patch.subtodos : existing.subtodos,
+      frequency: patch.frequency ?? existing.frequency,
+      weekdays: patch.frequency === 'weekly' || (patch.frequency === undefined && existing.frequency === 'weekly')
+        ? (patch.weekdays ?? existing.weekdays)
+        : undefined,
+      monthDays: patch.frequency === 'monthly' || (patch.frequency === undefined && existing.frequency === 'monthly')
+        ? (patch.monthDays ?? existing.monthDays)
+        : undefined,
+    }
+    const { frequency, weekdays, monthDays, quadrant } = normalizeRuleInput(merged)
+    ruleStmts.update.run({
+      id,
+      title: merged.title,
+      description: merged.description,
+      quadrant,
+      work_dir: merged.workDir || null,
+      brainstorm: merged.brainstorm ? 1 : 0,
+      applied_template_ids: JSON.stringify(Array.isArray(merged.appliedTemplateIds) ? merged.appliedTemplateIds : []),
+      subtodos: JSON.stringify(Array.isArray(merged.subtodos) ? merged.subtodos.map(s => ({
+        title: String(s.title || '').trim(),
+        description: String(s.description || ''),
+      })).filter(s => s.title) : []),
+      frequency,
+      weekdays: weekdays ? JSON.stringify(weekdays) : null,
+      month_days: monthDays ? JSON.stringify(monthDays) : null,
+      updated_at: Date.now(),
+    })
+    return rowToRule(ruleStmts.get.get(id))
+  }
+
+  function getRecurringRule(id) {
+    return rowToRule(ruleStmts.get.get(id))
+  }
+
+  function setRecurringRuleActive(id, active) {
+    ruleStmts.setActive.run(active ? 1 : 0, Date.now(), id)
+    return rowToRule(ruleStmts.get.get(id))
+  }
+
+  function deleteRecurringRule(id) {
+    ruleStmts.unlinkInstances.run(id)
+    ruleStmts.delete.run(id)
+  }
+
+  function sweepRecurring(now = Date.now()) {
+    const today = todayStr(now)
+    const startOfToday = new Date(now)
+    startOfToday.setHours(0, 0, 0, 0)
+    const startOfTodayMs = startOfToday.getTime()
+
+    db.prepare(`
+      UPDATE todos
+      SET status = 'missed', updated_at = ?
+      WHERE recurring_rule_id IS NOT NULL
+        AND instance_date IS NOT NULL
+        AND instance_date < ?
+        AND status IN ('todo', 'ai_done')
+    `).run(now, today)
+
+    const rules = ruleStmts.listActive.all().map(rowToRule)
+    const tx = db.transaction(() => {
+      for (const rule of rules) {
+        if (rule.lastGeneratedDate === today) continue
+        if (ruleShouldProduceOn(rule, today)) {
+          instantiateRule(rule, now)
+        }
+        ruleStmts.setLastGenerated.run(today, now, rule.id)
+      }
+    })
+    tx()
+    return { today, startOfTodayMs }
+  }
+
+  return {
+    raw: db,
+    listTemplates,
+    getTemplate,
+    createTemplate,
+    updateTemplate,
+    deleteTemplate,
+    createTodo,
+    getTodo,
+    updateTodo,
+    deleteTodo,
+    listTodos,
+    listSubtodosByParent,
+    listCompletedTodos,
+    countMissedInRange,
+    archiveTodo,
+    unarchiveTodo,
+    describeMergeTodos,
+    mergeTodos,
+    bulkUpdateTodos,
+    nextSortOrder,
+    addComment,
+    listComments,
+    deleteComment,
+    getComment,
+    insertSessionLog,
+    querySessionStats,
+    listSessionLogsInWindow: (tool, startedAt, windowMs) => {
+      const lo = startedAt - windowMs
+      const hi = startedAt + windowMs
+      return aiLogStmts.listInWindow.all(tool, lo, hi)
+    },
+    ftsAvailable,
+    transcriptFilesStmts: tfStmts,
+    upsertTranscriptFile,
+    deleteTranscriptFile,
+    writeFtsTurns,
+    searchTranscripts,
+    getTranscriptFile: (id) => tfStmts.getById.get(id),
+    listTranscriptFilesMeta: () => tfStmts.listAllPaths.all(),
+    listUnboundTranscriptFiles: () => tfStmts.listUnboundForMatching.all(),
+    findTranscriptByNative: (nativeId, tool) => tfStmts.findByNative.get(nativeId, tool),
+    setTranscriptBound: (id, todoId) => tfStmts.setBound.run(todoId, id),
+    countUnboundTranscripts: () => tfStmts.countUnbound.get().n,
+    createRecurringRule,
+    updateRecurringRule,
+    getRecurringRule,
+    setRecurringRuleActive,
+    deleteRecurringRule,
+    sweepRecurring,
+    createWikiRun,
+    completeWikiRun,
+    failWikiRun,
+    listWikiRuns,
+    findOrphanWikiRuns,
+    upsertWikiCoverage,
+    markCoverageApplied,
+    listCoverageForTodo,
+    listUnappliedDoneTodos,
+    // pending_questions (open-claw 桥接)
+    createPendingQuestion,
+    getPendingQuestion,
+    listPendingQuestions,
+    getLatestPendingQuestion,
+    answerPendingQuestion,
+    setPendingStatus,
+    sweepExpiredPendingQuestions,
+    close: () => db.close(),
+  }
+}