memory-eternal 0.10.33 → 0.10.35

package/lib/resident-state.js

@@ -45,6 +45,32 @@ export function recordReplaceBlocked(env = process.env, { error = '', lockPath =
   try { writeResidentState(env, data); return { ok: true, recordedAt: at, firstTime } } catch { return { ok: false, recordedAt: at, firstTime } }
 }
 
+/**
+ * 记下「常驻实例由**宿主**托管」（issue #30 ④ 的落地）：锁不可写时看门狗路径被永久拒绝，
+ * 于是宿主自己停掉旧实例、在原端口拉起新的并保活 —— 这条状态让面板/接口知道「虽然锁里没登记，
+ * 但实例是有人管的」，而不是显示成「未托管 / 孤儿」。
+ */
+export function recordHostManaged(env = process.env, { port = 0, version = '', served = '', reason = '' } = {}) {
+  const at = Date.now()
+  const data = { ...readResidentState(env), hostManaged: { at, port: Number(port) || 0, version: String(version || ''), served: String(served || ''), reason: String(reason || '').slice(0, 300), count: (Number((readResidentState(env).hostManaged || {}).count) || 0) + 1 } }
+  try { writeResidentState(env, data); return { ok: true, at } } catch { return { ok: false, at } }
+}
+
+/** 宿主托管状态（默认 24h 内算有效 —— 宿主进程一重启，这条就该被重新确认）。 */
+export function hostManagedInfo(env = process.env, { freshMs = 24 * 3600 * 1000 } = {}) {
+  const h = readResidentState(env).hostManaged
+  if (!h) return null
+  return { ...h, fresh: Date.now() - Number(h.at || 0) < freshMs }
+}
+
+/** 看门狗重新登记（或用户手工接管）→ 清掉宿主托管状态。 */
+export function clearHostManaged(env = process.env) {
+  const cur = readResidentState(env)
+  if (!cur.hostManaged) return false
+  delete cur.hostManaged
+  try { writeResidentState(env, cur); return true } catch { return false }
+}
+
 /** 环境修好（锁可写了）→ 清掉这条状态。 */
 export function clearReplaceBlocked(env = process.env) {
   const cur = readResidentState(env)

package/lib/watchdog.js

@@ -154,6 +154,53 @@ export function releaseWatchdogLock({ env = process.env, pid = process.pid, keep
  * 只补本端口、且当前没有**活着的**同端口登记时才写，避免覆盖新实例的登记。
  * @returns {boolean} 是否真的写回
  */
+/**
+ * 「接管助手」的单飞互斥（issue #30 第六轮实测）：宿主 tick 与 CLI 会各自拉起一个 `--replace`，
+ * 两个助手谁都不写锁、还互相收对方的 web —— 现场更难复盘。用一个 claim 文件做互斥：
+ * 助手开工前独占创建；已存在且**持有者还活着 + 没过期**就退出（不抢）。
+ *
+ * 失败开放（fail-open）：受限环境写不了 claim 时**照常进行**（不能让权限问题把接管彻底堵死），
+ * 但要留一条可查的理由。
+ */
+export function replaceClaimPath(env = process.env) {
+  return path.join(path.dirname(watchdogLockPath(env)), 'memory-eternal-replace.claim')
+}
+
+/** 读当前 claim（不存在/损坏返回 null）。 */
+export function readReplaceClaim(env = process.env) {
+  try {
+    const d = JSON.parse(fs.readFileSync(replaceClaimPath(env), 'utf8'))
+    if (!d || typeof d !== 'object' || !d.pid) return null
+    return d
+  } catch { return null }
+}
+
+/** 抢 claim：抢到 true；有人在跑则是 false（附持有者信息）。 */
+export function claimReplaceSlot({ env = process.env, port = 0, pid = process.pid, staleMs = 60000, now = Date.now() } = {}) {
+  const file = replaceClaimPath(env)
+  const cur = readReplaceClaim(env)
+  if (cur) {
+    const fresh = now - Number(cur.at || 0) < Math.max(1000, Number(staleMs) || 60000)
+    if (fresh && Number(cur.pid) !== Number(pid) && isPidAlive(cur.pid)) {
+      return { claimed: false, holder: cur, reason: '已有接管助手在跑（pid ' + cur.pid + '，' + Math.round((now - Number(cur.at || now)) / 1000) + 's 前）' }
+    }
+  }
+  try {
+    fs.mkdirSync(path.dirname(file), { recursive: true })
+    fs.writeFileSync(file, JSON.stringify({ pid: Number(pid), port: Number(port) || 0, at: now }))
+    return { claimed: true, holder: cur || null, reason: cur ? '接管了陈旧 claim' : '抢到 claim' }
+  } catch (e) {
+    return { claimed: true, holder: cur || null, reason: 'claim 写不进去（' + String((e && e.message) || e) + '）→ 失败开放，继续接管' }
+  }
+}
+
+/** 释放 claim（只清自己那条；别人的不动）。 */
+export function releaseReplaceClaim({ env = process.env, pid = process.pid } = {}) {
+  const cur = readReplaceClaim(env)
+  if (cur && Number(cur.pid) !== Number(pid)) return false
+  try { fs.unlinkSync(replaceClaimPath(env)); return true } catch { return false }
+}
+
 export function restoreWatchdogLock({ env = process.env, port, watchdogs = [] } = {}) {
   const p = Number(port) || 0
   const want = (Array.isArray(watchdogs) ? watchdogs : []).filter((w) => Number(w.port) === p)
@@ -782,6 +829,19 @@ export async function restartResident({
       }
     }
   }
+  // 单飞互斥（issue #30 第六轮实测）：接管还在进行中时**不要再拉第二个助手** ——
+  // 宿主 tick 与 CLI 同时来的时候，两个助手会互相收对方的 web，现场反而更难复盘。
+  if (!deps.ignoreClaim) {
+    try {
+      const inflight = readReplaceClaim(env)
+      if (inflight && Number(inflight.pid) !== process.pid && isPidAlive(inflight.pid) && Date.now() - Number(inflight.at || 0) < 60000) {
+        return {
+          ok: false, stage: 'in-flight', expect, served: '', waitedMs: 0, helperPid: 0, before, lockRestored: false, occupant: null,
+          error: '已有接管正在进行中（pid ' + inflight.pid + '，' + Math.round((Date.now() - Number(inflight.at || Date.now())) / 1000) + 's 前启动）→ 本次不再重复拉起助手；稍后可再试 dsh-memory restart --port ' + p,
+        }
+      }
+    } catch { /* 读不到 claim 就当没有 */ }
+  }
   let helperPid = 0
   try {
     const child = (deps.spawnHelper || spawnReplaceHelper)({ port: p, vaultRoot, interval, maxRestart })
@@ -1302,6 +1362,15 @@ if (process.argv[1] && /[\\/]watchdog\.js$/.test(process.argv[1])) {
     try { fs.writeSync(2, `[watchdog] ${msg}\n`) } catch { log(msg) }
     process.exitCode = 0
   } else if (argv.includes('--replace')) {
+    // 单飞互斥（issue #30 第六轮实测：宿主 tick 与 CLI 各拉一个助手互相踩）：
+    // 抢不到 claim 就**直接退出**，把接管让给已经在跑的那个；抢到就在退出时释放。
+    const slotClaim = claimReplaceSlot({ port })
+    if (!slotClaim.claimed) {
+      log('跳过本次接管：' + slotClaim.reason + '（避免两个助手互相收对方的 web）')
+      process.exitCode = 0
+    } else {
+    if (slotClaim.holder) log(slotClaim.reason)
+    process.on('exit', () => { try { releaseReplaceClaim() } catch { /* 无妨 */ } })
     // 版本漂移自愈：先收掉旧常驻实例，再作为新一代看门狗接管（见 clearResidentForTakeover 的说明）。
     const cleared = await clearResidentForTakeover({ port })
     for (const w of cleared.warnings) log('⚠ ' + w)
@@ -1323,6 +1392,7 @@ if (process.argv[1] && /[\\/]watchdog\.js$/.test(process.argv[1])) {
     }
     await noteReplace('boot', summary)
     bootWatchdog()
+    }
   } else {
     bootWatchdog()
   }

package/lib/web.js

@@ -281,10 +281,50 @@ async function waitForOurWeb(port, timeoutMs = 6000, intervalMs = 500) {
  * 端口被非本项目进程占用时向后漂移（+1..+10）。
  * @returns {Promise<{url:string, port:number, spawned:boolean}>}
  */
-export async function ensureWebServer({ port = DEFAULT_WEB_PORT, vaultRoot = defaultVaultDir(), spawnEnv = {}, totalTimeoutMs = 15000, takeoverWaitMs = 6000 } = {}) {
+/**
+ * 强制接管端口上的**自家旧 web**：收掉它、等端口空出来（issue #30 第四点：锁不可写的机器上
+ * 看门狗路径被永久拒绝，只能由**宿主**自己把常驻实例换掉 —— 宿主能写、也能 kill 自己人的进程）。
+ *
+ * 安全闸门：只对 `looksLikeOurWeb(command)` 为真的占用者下手；别人的进程一个都不许动。
+ *
+ * @returns {Promise<{took:boolean, pid:number, reason:string}>}
+ */
+export async function takeoverOwnWeb(port, { findListener, kill, sleep, looksLike, timeoutMs = 4000 } = {}) {
+  const p = Number(port) || DEFAULT_WEB_PORT
+  const find = findListener || (await import('./watchdog.js')).findPortListener
+  const isOurs = looksLike || looksLikeOurWeb
+  const wait = sleep || ((ms) => new Promise((r) => setTimeout(r, ms)))
+  const killFn = kill || ((pid) => { try { process.kill(pid, 'SIGTERM'); return true } catch { return false } })
+  let occ = null
+  try { occ = await find(p) } catch { occ = null }
+  if (!occ || !occ.pid) return { took: false, pid: 0, reason: '端口上没有占用者' }
+  if (!isOurs(occ.command)) return { took: false, pid: Number(occ.pid), reason: '占用者不是本插件的 web（' + (occ.command || '命令行未知') + '）→ 不动它' }
+  let alive = true
+  try { alive = await isOurs(p) } catch { alive = true }
+  void alive
+  if (!killFn(occ.pid)) return { took: false, pid: Number(occ.pid), reason: '终止失败（可能无权限）' }
+  const deadline = Date.now() + Math.max(0, Number(timeoutMs) || 0)
+  while (Date.now() < deadline) {
+    await wait(200)
+    let left = null
+    try { left = await find(p) } catch { left = null }
+    if (!left || !left.pid) return { took: true, pid: Number(occ.pid), reason: '已收掉自家旧 web' }
+  }
+  return { took: false, pid: Number(occ.pid), reason: '已发出终止但端口仍被占用' }
+}
+
+export async function ensureWebServer({ port = DEFAULT_WEB_PORT, vaultRoot = defaultVaultDir(), spawnEnv = {}, totalTimeoutMs = 15000, takeoverWaitMs = 6000, forceTakeover = false, deps = {} } = {}) {
   // 探测可用端口：先找「已是我们的服务」的端口，再找空闲端口
   let target = port
   let found = null
+  // 强制接管（issue #30 ④）：宿主自管的机器上，先把自家旧实例收掉，再在原端口起新的 ——
+  // 不这么做就会走到下面的「让位到 port+1」，于是既换不掉旧的、又多出一个实例。
+  if (forceTakeover) {
+    try {
+      const r = await takeoverOwnWeb(port, { findListener: deps.findListener, kill: deps.kill, sleep: deps.sleep, timeoutMs: Math.max(1500, Math.min(5000, totalTimeoutMs)) })
+      process.stderr.write(`[memory-eternal] 强制接管端口 ${port}：${r.reason}${r.pid ? '（pid ' + r.pid + '）' : ''}\n`)
+    } catch (e) { process.stderr.write(`[memory-eternal] 强制接管失败：${String((e && e.message) || e)}\n`) }
+  }
   for (let p = port; p < port + 10; p++) {
     const alive = await probeWebServer(p)
     if (alive) { found = { url: `http://127.0.0.1:${p}`, port: p, spawned: false }; break }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "memory-eternal",
-  "version": "0.10.33",
+  "version": "0.10.35",
   "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 && node tests/tags-normalize.test.mjs && node tests/resident-state.test.mjs && node tests/llm-route.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 && node tests/llm-route.test.mjs && node tests/host-managed.test.mjs && node tests/replace-claim.test.mjs"
   },
   "dependencies": {
     "@deepseek-ai/schemastery": "^3.18.1"