memory-eternal 0.10.29 → 0.10.32

package/lib/resident-state.js

@@ -0,0 +1,61 @@
+// 「常驻实例不可替换」这个**长期状态**的记忆（issue #30 第五轮复测：A: 盘那台的稳态）。
+//
+// 现场（报告者实测 5 天 44 次）：
+//   升级 → 端口上那个常驻实例仍是上一代 → 宿主探测到"漂移" → 拉起 --replace 助手
+//   → 助手前置探到 lock.writable = false → **正确地拒绝**（宁可拒绝也不留孤儿）
+//   → 旧实例继续存活 → 下次开机再报一次…… 于是「每一步都按设计工作，合起来永远换不掉」。
+//
+// 所以：**环境限制要当成状态记下来**，而不是每次都当"漂移"重试一遍。记下之后：
+//   · 面板 / align / version-check 能显示「常驻实例不可替换（环境限制）」并给出路；
+//   · 宿主不再每次开机都重试注定失败的替换（首次 warn，之后降级为 boot 级别的一条已知状态）。
+import fs from 'node:fs'
+import path from 'node:path'
+import { watchdogLockPath } from './watchdog.js'
+
+/** 状态文件与看门狗锁同目录（同一个 DSH_HOME）。 */
+export function residentStatePath(env = process.env) {
+  return path.join(path.dirname(watchdogLockPath(env)), 'memory-eternal-state.json')
+}
+
+/** 读状态：不存在/损坏都返回 {}（绝不影响主流程）。 */
+export function readResidentState(env = process.env) {
+  try {
+    const data = JSON.parse(fs.readFileSync(residentStatePath(env), 'utf8'))
+    return data && typeof data === 'object' ? data : {}
+  } catch { return {} }
+}
+
+function writeResidentState(env, data) {
+  const file = residentStatePath(env)
+  const tmp = file + '.tmp'
+  fs.mkdirSync(path.dirname(file), { recursive: true })
+  fs.writeFileSync(tmp, JSON.stringify(data, null, 2), 'utf8')
+  fs.renameSync(tmp, file)
+}
+
+/**
+ * 记下「常驻实例不可替换（环境限制）」。
+ * @returns {{ok:boolean, recordedAt:number, firstTime:boolean}}
+ */
+export function recordReplaceBlocked(env = process.env, { error = '', lockPath = '', port = 0, version = '', served = '' } = {}) {
+  const prev = readResidentState(env).replaceBlocked
+  const firstTime = !prev
+  const at = Date.now()
+  const data = { ...readResidentState(env), replaceBlocked: { at, error: String(error || '').slice(0, 300), lockPath: String(lockPath || ''), port: Number(port) || 0, version: String(version || ''), served: String(served || ''), count: (prev && Number(prev.count) || 0) + 1 } }
+  try { writeResidentState(env, data); return { ok: true, recordedAt: at, firstTime } } catch { return { ok: false, recordedAt: at, firstTime } }
+}
+
+/** 环境修好（锁可写了）→ 清掉这条状态。 */
+export function clearReplaceBlocked(env = process.env) {
+  const cur = readResidentState(env)
+  if (!cur.replaceBlocked) return false
+  delete cur.replaceBlocked
+  try { writeResidentState(env, cur); return true } catch { return false }
+}
+
+/** 供界面/报告用：这条状态是否仍然成立（默认 7 天内算"当前有效"）。 */
+export function replaceBlockedInfo(env = process.env, { freshMs = 7 * 24 * 3600 * 1000 } = {}) {
+  const b = readResidentState(env).replaceBlocked
+  if (!b) return null
+  return { ...b, fresh: Date.now() - Number(b.at || 0) < freshMs }
+}

package/lib/vault.js

@@ -184,7 +184,7 @@ function rowToCard(row) {
     path: row.path,
     kind: row.kind,
     title: row.title,
-    tags: JSON.parse(row.tags || '[]'),
+    tags: normalizeTags(row.tags),
     summary: row.summary || '',
     created: row.created_at,
     updated: row.updated_at,
@@ -199,7 +199,7 @@ function rowToCard(row) {
 
 /** 将 SQL 行还原为完整 Markdown 文本（frontmatter + body）。 */
 function rowToMarkdown(row) {
-  const tags = JSON.parse(row.tags || '[]')
+  const tags = normalizeTags(row.tags)
   const lines = [
     '---',
     `formatVersion: 1`,
@@ -558,9 +558,43 @@ export async function setCardStatus(root, rel, status, { changedBy = 'system', r
  * `dedupAgainst` 传入候选集合时，去重只跟**这些**卡比（导入备份用：基线取导入前的快照，
  * 让同一批里互相近似的卡不会互吞）。
  */
+/**
+ * 把各种形态的 tags 归一化成字符串数组（issue #41 的根因）。
+ *
+ * 现场：调用方通过 /write 传了**字符串**（`"[\"a\",\"b\"]"`），而 writeCard 直接 `JSON.stringify(tags)`
+ * → 落库变成「一个字符串的 JSON 表示」（双重编码）→ 读出来仍是字符串 → 前端 `tags.slice(...).map` 抛错
+ * → **整页黑屏**（React 根组件挂载失败）。自动沉淀路径本来就有 Array.isArray 检查，只有 /write 等直连入口漏了。
+ * 这里做成唯一入口：数组 / JSON 字符串 / 逗号串 / 双重编码的旧数据，全都收敛成数组。
+ */
+export function normalizeTags(tags) {
+  let v = tags
+  if (typeof v === 'string') {
+    const s = v.trim()
+    if (!s) return []
+    // 最多剥两层：正常 JSON（["a","b"]）+ 双重编码（"[\"a\",\"b\"]" —— #41 的现场，
+    // 它以引号开头，所以只判断 startsWith('[') 会漏掉）。剥不动就按逗号串处理。
+    let cur = s
+    for (let i = 0; i < 2; i++) {
+      if (typeof cur !== 'string') break
+      const t = cur.trim()
+      if (!(t.startsWith('[') || t.startsWith('"'))) break
+      try { cur = JSON.parse(t) } catch { break }
+    }
+    v = typeof cur === 'string' ? cur.split(',') : cur
+  }
+  if (!Array.isArray(v)) return []
+  return v
+    .map((t) => String(t == null ? '' : t).trim())
+    .filter(Boolean)
+    .map((t) => t.slice(0, 30))
+    .slice(0, 20)
+}
+
 export async function writeCard(root, { kind, title, tags = [], body, source = '', status = 'pending', submittedBy = '', severity = 'info', reason = '', deletedAt = '' }, { threshold = 0.62, dedup = true, dedupAgainst = null } = {}) {
   const db = getDb(root)
   const kindRoot = KIND_ROOTS[kind] || KIND_ROOTS.knowledge
+  // 唯一归一化入口：任何调用方（/write、MCP、脚本、自动沉淀）传字符串都不会再双重编码
+  tags = normalizeTags(tags)
 
   // 审核守卫：数据库层强制，调用方传的 status 仅作 fallback
   const enforcedStatus = enforceAudit(root, kind, submittedBy, status)
@@ -634,10 +668,10 @@ export async function mergeCards(root, paths) {
   if (rows.length < 2) return { ok: false, error: '读取失败' }
 
   const first = rows[0]
-  const tags = new Set(JSON.parse(first.tags || '[]'))
+  const tags = new Set(normalizeTags(first.tags))
   let combined = ''
   rows.forEach((row, i) => {
-    const rowTags = JSON.parse(row.tags || '[]')
+    const rowTags = normalizeTags(row.tags)
     rowTags.forEach((t) => tags.add(t))
     combined += (i ? '\n\n---\n\n' : '') + row.body.trim()
   })
@@ -784,7 +818,7 @@ export async function graph(root) {
   const bodyRows = db.prepare('SELECT path, body FROM cards').all()
   const bodyMap = new Map(bodyRows.map((r) => [r.path, r.body]))
   const cards = rows.map((r) => ({
-    path: r.path, kind: r.kind, title: r.title, tags: JSON.parse(r.tags || '[]'),
+    path: r.path, kind: r.kind, title: r.title, tags: normalizeTags(r.tags),
     summary: r.summary || '', updated: r.updated_at, mtime: new Date(r.updated_at).getTime(), status: r.status,
   }))
 
@@ -983,7 +1017,7 @@ export async function stats(root) {
   const weekRow = db.prepare('SELECT COUNT(*) as cnt FROM cards WHERE updated_at >= ?').get(weekAgo)
   const tagRows = db.prepare('SELECT tags FROM cards').all()
   const tagSet = new Set()
-  for (const r of tagRows) { for (const t of JSON.parse(r.tags || '[]')) tagSet.add(t) }
+  for (const r of tagRows) { for (const t of normalizeTags(r.tags)) tagSet.add(t) }
 
   const todayCardsRows = db.prepare('SELECT * FROM cards WHERE updated_at >= ? ORDER BY updated_at DESC LIMIT 30').all(dayAgo)
   const todayCards = todayCardsRows.map((row) => ({ path: row.path, title: row.title, kind: row.kind, updated: row.updated_at, summary: (row.summary || '').slice(0, 80) }))

package/lib/watchdog.js

@@ -108,14 +108,17 @@ export function probeLockWritable(env = process.env) {
  * 尝试取得看门狗锁（按端口分槽）。
  * @returns {{acquired:boolean, file:string, previous:object|null, reaped:number[]}} acquired=false 表示同端口已有活着的看门狗，调用方应直接退出。
  */
-export function acquireWatchdogLock({ env = process.env, port, vault = '', pid = process.pid, pkgVersion = currentPkgVersion() } = {}) {
+export function acquireWatchdogLock({ env = process.env, port, vault = '', pid = process.pid, pkgVersion = currentPkgVersion(), force = false } = {}) {
   const file = watchdogLockPath(env)
   const cur = readWatchdogLock(env)
   const samePort = cur.watchdogs.filter((w) => Number(w.port) === Number(port) && w.pid !== pid)
   // endedAt 的槽是「已结束」的记录（releaseWatchdogLock 留下的痕迹，见 #30）：**绝不能挡住新实例**
   // —— pid 可能已被复用、那个进程还活着，但那条记录明确表示它已经交出这把锁。
   const alive = samePort.find((w) => !w.endedAt && isPidAlive(w.pid))
-  if (alive) return { acquired: false, file, previous: alive, reaped: [] }
+  // force：--replace 助手的语义就是「替换同端口那个」。不允许让位 —— 否则旧实例还活着时
+  // 新助手直接退出，而清场那一步可能根本没跑完（issue #30 实测：清场被十几秒的枚举拖住 →
+  // 助手退在写锁之前 → 锁永远为空 → verify 必然超时、restart 100% 失败）。
+  if (alive && !force) return { acquired: false, file, previous: alive, reaped: [] }
   // 清掉死掉的槽（含其它端口的陈旧记录），保留仍然存活的其它端口
   const reaped = []
   const kept = []
@@ -156,7 +159,9 @@ export function restoreWatchdogLock({ env = process.env, port, watchdogs = [] }
   const want = (Array.isArray(watchdogs) ? watchdogs : []).filter((w) => Number(w.port) === p)
   if (!want.length) return false
   const cur = readWatchdogLock(env)
-  const liveSame = cur.watchdogs.filter((w) => Number(w.port) === p && isPidAlive(w.pid))
+  // 必须排除 endedAt 槽（与 acquireWatchdogLock 同一口径）——报告者实测：PID 复用/正在退出时，
+  // 一条本端口的 endedAt 槽会让这里误判「已有活着的登记」而放弃写回。
+  const liveSame = cur.watchdogs.filter((w) => Number(w.port) === p && !w.endedAt && isPidAlive(w.pid))
   if (liveSame.length) return false
   const others = cur.watchdogs.filter((w) => Number(w.port) !== p && isPidAlive(w.pid))
   writeWatchdogLock(env, [...others, ...want.map((w) => ({ ...w, restoredAt: new Date().toISOString() }))])
@@ -386,9 +391,10 @@ async function listWatchdogProcesses(exec, platform = process.platform, marker =
  * 两条对策：① `try/catch` 包住 spawn，失败就当「枚举不到」；② **用临时文件接 stdout**，完全不碰管道。
  */
 function defaultExec() {
-  return async (cmd, args) => {
+  return async (cmd, args, opts = {}) => {
     const { spawn } = await import('node:child_process')
     let outFile = ''
+    let timer = null
     try {
       const os = await import('node:os')
       const path = await import('node:path')
@@ -401,6 +407,10 @@ function defaultExec() {
       } finally {
         try { fs.closeSync(fd) } catch { /* 子进程已继承则关父进程副本 */ }
       }
+      // 超时必须能掐断子进程：issue #30 实测 Windows 上 Get-NetTCPConnection 无条件要 4–7.5 秒
+      // （一次 13.6 秒），而调用方（接管清场）的 8000ms 预算根本管不住它 —— 于是决策基于过期数据。
+      const tm = Number(opts && opts.timeoutMs) || 0
+      if (tm > 0) timer = setTimeout(() => { try { child.kill() } catch { /* 已经在退出 */ } }, tm)
       await new Promise((resolve) => {
         child.on('error', () => resolve(''))
         child.on('close', () => resolve(''))
@@ -409,6 +419,7 @@ function defaultExec() {
     } catch {
       return ''   // 受限环境：宁可「枚举不到」，也不能把调用方（如 --reap）崩掉
     } finally {
+      if (timer) clearTimeout(timer)
       if (outFile) { try { (await import('node:fs')).unlinkSync(outFile) } catch { /* 清理失败无妨 */ } }
     }
   }
@@ -820,6 +831,8 @@ export async function restartResident({
 export async function clearResidentForTakeover({ port = 7999, env = process.env, timeoutMs = 8000, deps = {} } = {}) {
   const p = Number(port) || 7999
   const findListener = deps.findListener || findPortListener
+  // 枚举本身要能被时间预算约束（否则 8s 预算被一次 13.6s 的 PowerShell 查询吃光，决策基于过期数据）
+  const listenerOpts = { timeoutMs: Math.max(500, Math.min(2500, Number(timeoutMs) || 8000)) }
   const out = { stopped: [], webStopped: [], portStopped: [], warnings: [], failed: [], portFree: false }
   const rawKill = deps.kill || ((pid) => { try { process.kill(pid, 'SIGTERM'); return true } catch { return false } })
   // 记下「已经判定收不掉」的 pid：立刻再来一次只会同样失败，白折腾还可能误导用户以为是权限抖动
@@ -841,14 +854,14 @@ export async function clearResidentForTakeover({ port = 7999, env = process.env,
   }
   const wait = deps.sleep || ((ms) => new Promise((r) => setTimeout(r, ms)))
   let left = null
-  try { left = await findListener(p) } catch { left = null }
+  try { left = await findListener(p, listenerOpts) } catch { left = null }
   const deadline = Date.now() + Math.max(0, Number(timeoutMs) || 0)
   // 只对「确认是本插件的 web」下手：别人的进程占着端口时只如实报告，不误杀（同 issue #23 的口径）。
   // 已经判定收不掉的 pid 不再重试 —— 立刻再来一次只会同样失败。
   while (left && left.pid && looksLikeOurWeb(left.command) && !failedKills.has(Number(left.pid)) && Date.now() < deadline) {
     if (!killFn(left.pid)) break
     await wait(200)
-    try { left = await findListener(p) } catch { left = null }
+    try { left = await findListener(p, listenerOpts) } catch { left = null }
   }
   if (left && left.pid) {
     out.portFree = false
@@ -904,19 +917,20 @@ export function parseLsofFields(text) {
  * @param {{exec?: (cmd:string,args:string[])=>Promise<string>}} [opts]
  * @returns {Promise<{pid:number, command:string}|null>}
  */
-export async function findPortListener(port, { exec, platform = process.platform } = {}) {
+export async function findPortListener(port, { exec, platform = process.platform, timeoutMs = 4000 } = {}) {
   const p = Number(port)
   if (!Number.isFinite(p) || p <= 0) return null
   const runExec = exec || defaultExec()
+  const execOpts = { timeoutMs: Math.max(200, Number(timeoutMs) || 4000) }
   try {
     if (platform === 'win32') {
-      const out = await runExec('powershell', ['-NoProfile', '-Command', `Get-NetTCPConnection -LocalPort ${p} -State Listen -ErrorAction SilentlyContinue | Select-Object -ExpandProperty OwningProcess`])
+      const out = await runExec('powershell', ['-NoProfile', '-Command', `Get-NetTCPConnection -LocalPort ${p} -State Listen -ErrorAction SilentlyContinue | Select-Object -ExpandProperty OwningProcess`], execOpts)
       const pid = Number(String(out || '').split(/\r?\n/).map((s) => s.trim()).find((s) => /^\d+$/.test(s))) || 0
       if (!pid) return null
-      const cmd = await runExec('powershell', ['-NoProfile', '-Command', `(Get-CimInstance Win32_Process -Filter "ProcessId=${pid}" | Select-Object -ExpandProperty CommandLine)`])
+      const cmd = await runExec('powershell', ['-NoProfile', '-Command', `(Get-CimInstance Win32_Process -Filter "ProcessId=${pid}" | Select-Object -ExpandProperty CommandLine)`], execOpts)
       return { pid, command: String(cmd || '').trim() }
     }
-    const fields = await runExec('lsof', ['-nP', `-iTCP:${p}`, '-sTCP:LISTEN', '-Fpcn'])
+    const fields = await runExec('lsof', ['-nP', `-iTCP:${p}`, '-sTCP:LISTEN', '-Fpcn'], execOpts)
     const list = parseLsofFields(fields)
     if (!list.length) return null
     const pid = list[0].pid
@@ -1287,6 +1301,12 @@ if (process.argv[1] && /[\\/]watchdog\.js$/.test(process.argv[1])) {
     log(summary)
     if (cleared.failed.length) log('⚠ 未能停止：pid ' + cleared.failed.join(', '))
     if (!cleared.portFree) {
+      // issue #30：**清场失败也必须留下登记**。老实现直接 exit(1)，于是锁里什么都没有 ——
+      // 而 restartResident 的自检要求「锁里出现新看门狗」，必然超时，用户看到的永远是「（无响应）」。
+      // 登记一条（pid 已退出时会自动落成 endedAt 记录，不挡后续），现场从此可复盘。
+      const claim = acquireWatchdogLock({ port, vault: vaultRoot || '', pkgVersion: currentPkgVersion(), force: true })
+      if (claim.acquired) log('已登记本次替换尝试（端口仍被占用，看门狗未启动）：pid ' + process.pid)
+      else log('⚠ 未能登记本次替换（锁不可写？）：' + (claim.file || ''))
       const why = `${summary} —— 端口没空出来，已中止接管（不退让到 port+1，避免又留一个孤儿 web）`
       await noteReplace('fail', why)
       process.exit(1)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "memory-eternal",
-  "version": "0.10.29",
+  "version": "0.10.32",
   "description": "记忆核心（Memory Eternal）：自研的 DeepSeek Harness 记忆插件，不移植任何既有记忆框架——对话结束后自动沉淀知识卡到本地 Markdown Vault（自研去重、自研 CJK 检索、可 git 管理），设置页提供图形化知识库（统计 / 搜索 / 知识图谱 + 侧边栏一键弹窗），Agent 通过 memory_recall 工具按需召回历史上下文。零人工干预。",
   "license": "MIT",
   "repository": {
@@ -64,7 +64,7 @@
     "prepublishOnly": "node build.mjs",
     "release": "node scripts/release.mjs",
     "release:dry": "node scripts/release.mjs --dry-run",
-    "test": "node build.mjs && node tests/vault.test.mjs && node tests/capture.test.mjs && node tests/capture-route.test.mjs && node tests/stall.test.mjs && node tests/vault-path.test.mjs && node tests/vault-resolve.test.mjs && node tests/minhash.test.mjs && node tests/settings-save.test.mjs && node tests/config-coverage.test.mjs && node tests/config-sync.test.mjs && node tests/config-merge.test.mjs && node tests/sse.test.mjs && node tests/api.test.mjs && node tests/import-roundtrip.test.mjs && node tests/graph-perf.test.mjs && node tests/settings-compat.test.mjs && node tests/node-bin.test.mjs && node tests/watchdog-lock.test.mjs && node tests/markdown.test.mjs && node tests/feedback.test.mjs && node tests/graph-lod.test.mjs && node tests/ssr.test.mjs && node tests/web-assets.test.mjs && node tests/manifests.test.mjs && node tests/two-store.test.mjs && node tests/noise-gate.test.mjs && node tests/e2e-recall.test.mjs && node tests/issue-fixes.test.mjs && node tests/issue-fixes-2.test.mjs && node tests/issue-fixes-3.test.mjs && node tests/watchdog-orphans.test.mjs && node tests/watchdog-restart.test.mjs && node tests/readme-i18n.test.mjs && node tests/issue-fixes-4.test.mjs && node tests/host-drain.test.mjs && node tests/e2e-host-drain.mjs && node tests/issue-fixes-5.test.mjs && node tests/issue-fixes-6.test.mjs && node tests/capture-json-repair.test.mjs && node tests/csrf-origin.test.mjs && node tests/watchdog-orphan-web.test.mjs && node tests/config-validate.test.mjs && node tests/issue-fixes-8.test.mjs && node tests/issue-fixes-9.test.mjs && node tests/release-verify.test.mjs && node tests/refine.test.mjs && node tests/db-restore.test.mjs && node tests/align.test.mjs"
+    "test": "node build.mjs && node tests/vault.test.mjs && node tests/capture.test.mjs && node tests/capture-route.test.mjs && node tests/stall.test.mjs && node tests/vault-path.test.mjs && node tests/vault-resolve.test.mjs && node tests/minhash.test.mjs && node tests/settings-save.test.mjs && node tests/config-coverage.test.mjs && node tests/config-sync.test.mjs && node tests/config-merge.test.mjs && node tests/sse.test.mjs && node tests/api.test.mjs && node tests/import-roundtrip.test.mjs && node tests/graph-perf.test.mjs && node tests/settings-compat.test.mjs && node tests/node-bin.test.mjs && node tests/watchdog-lock.test.mjs && node tests/markdown.test.mjs && node tests/feedback.test.mjs && node tests/graph-lod.test.mjs && node tests/ssr.test.mjs && node tests/web-assets.test.mjs && node tests/manifests.test.mjs && node tests/two-store.test.mjs && node tests/noise-gate.test.mjs && node tests/e2e-recall.test.mjs && node tests/issue-fixes.test.mjs && node tests/issue-fixes-2.test.mjs && node tests/issue-fixes-3.test.mjs && node tests/watchdog-orphans.test.mjs && node tests/watchdog-restart.test.mjs && node tests/readme-i18n.test.mjs && node tests/issue-fixes-4.test.mjs && node tests/host-drain.test.mjs && node tests/e2e-host-drain.mjs && node tests/issue-fixes-5.test.mjs && node tests/issue-fixes-6.test.mjs && node tests/capture-json-repair.test.mjs && node tests/csrf-origin.test.mjs && node tests/watchdog-orphan-web.test.mjs && node tests/config-validate.test.mjs && node tests/issue-fixes-8.test.mjs && node tests/issue-fixes-9.test.mjs && node tests/release-verify.test.mjs && node tests/refine.test.mjs && node tests/db-restore.test.mjs && node tests/align.test.mjs && node tests/tags-normalize.test.mjs && node tests/resident-state.test.mjs"
   },
   "dependencies": {
     "@deepseek-ai/schemastery": "^3.18.1"