evolcore 0.0.35 → 0.0.37

package/dist/cli/restart-monitor.js

@@ -1,4 +1,5 @@
 import fs from 'fs';
+import { ipcQuery } from '../ipc.js';
 import path from 'path';
 import { spawn, execFile } from 'child_process';
 import { promisify } from 'util';
@@ -7,11 +8,12 @@ import { isRestartRecoveryEnabled, loadDaemonConfig } from '../config-store.js';
 import * as platform from '../utils/cross-platform.js';
 import { EventBus } from '../core/event-bus.js';
 import { applyCachedPackageUpgrades, recordPackageUpgrade } from '../utils/package-version-info.js';
+import { acknowledgeUpgradeRestart, readUpgradeJob, updateUpgradeJob } from '../utils/upgrade-jobs.js';
 import { isValidAid } from '../aun/aid/index.js';
 import { msgSend } from '../aun/msg/index.js';
-import { scanInstances, cleanupInstances, writeRestartMonitor, removeRestartMonitor, isRestartMonitorWinner, findOrphanProcesses, killOrphans } from '../utils/instance-registry.js';
-import { MAIN_PACKAGE_NAME, WEB_PACKAGE_NAME } from '../product.js';
-import { shouldSuppressRealRestart } from '../utils/restart-safety.js';
+import { scanInstances, cleanupInstances, writeRestartMonitor, removeRestartMonitor, isRestartMonitorWinner, findOrphanProcesses, killOrphans, updateRestartMonitorPhase } from '../utils/instance-registry.js';
+import { MAIN_PACKAGE_NAME, RESTART_REQUEST_ID_ENV } from '../product.js';
+import { readRestartRequest, updateRestartRequest, removeRestartRequest, shouldSuppressRealRestart } from '../utils/restart-safety.js';
 import { rotateStdoutLog } from '../utils/log-writer.js';
 import { applyDataMigration, finalizeDataMigration, planDataMigration, verifyDataMigration, } from '../core/data-migration.js';
 import { DAEMON_SHUTDOWN_TIMEOUT_MS, requestDaemonShutdown, stopCodexAppServerOrphans } from './daemon-commands.js';
@@ -112,6 +114,8 @@ export async function cmdRestartMonitor() {
         return;
     const p = resolvePaths();
     const restartLog = path.join(p.logs, 'restart.log');
+    const pendingFile = path.join(daemonControlDir(), 'restart-pending.json');
+    const requestId = process.env[RESTART_REQUEST_ID_ENV] || 'unknown';
     const MAX_HEAL_ATTEMPTS = 3;
     const READY_TIMEOUT = 30000; // 30s（AUN sidecar + 外部渠道连接预留）
     const HEAL_TIMEOUT = 30 * 60 * 1000; // 30 分钟，让 claude 自然结束
@@ -120,9 +124,24 @@ export async function cmdRestartMonitor() {
         const _d = new Date();
         const _p = (n) => String(n).padStart(2, '0');
         const ts = `${_d.getFullYear()}-${_p(_d.getMonth() + 1)}-${_p(_d.getDate())} ${_p(_d.getHours())}:${_p(_d.getMinutes())}:${_p(_d.getSeconds())}`;
-        const line = `[${ts}] ${msg}\n`;
+        const line = `[${ts}] [requestId=${requestId}] ${msg}\n`;
         fs.appendFileSync(restartLog, line);
     };
+    const hasCorrelatedRequest = requestId !== 'unknown';
+    const upgradeJobId = hasCorrelatedRequest ? readRestartRequest(pendingFile, requestId)?.upgradeJobId : undefined;
+    const upgradeJob = typeof upgradeJobId === 'string' ? readUpgradeJob(upgradeJobId) : null;
+    if (upgradeJobId && (!upgradeJob || upgradeJob.restartRequestId !== requestId || upgradeJob.released || upgradeJob.deadline <= Date.now())) {
+        if (upgradeJob)
+            updateUpgradeJob(upgradeJob.jobId, { status: 'timeout', success: false, released: true, code: 'UPGRADE_TIMEOUT', reason: 'Update expired before restart preflight' }, '更新已超时，未执行重启');
+        process.exitCode = 2;
+        cleanupPendingFile(pendingFile, log, requestId);
+        return;
+    }
+    if (hasCorrelatedRequest && !readRestartRequest(pendingFile, requestId)) {
+        log('Restart request is missing or has been replaced; refusing to proceed');
+        process.exitCode = 2;
+        return;
+    }
     // 单实例保护：pre-check + post-write 自检（同 main 进程）
     {
         const pre = scanInstances();
@@ -133,8 +152,17 @@ export async function cmdRestartMonitor() {
         }
     }
     // 立即登记自己；exit 路径上自动清理 record
-    writeRestartMonitor();
-    process.on('exit', () => {
+    writeRestartMonitor(hasCorrelatedRequest ? requestId : undefined, 'starting');
+    process.on('exit', code => {
+        if (upgradeJob) {
+            const current = readUpgradeJob(upgradeJob.jobId);
+            if (current)
+                updateUpgradeJob(upgradeJob.jobId, {
+                    released: true,
+                    lifecycle: current.lifecycle ?? { status: code === 0 ? 'completed' : 'failed', at: new Date().toISOString(), ...(code !== 0 ? { reason: 'Restart lifecycle failed; inspect restart.log' } : {}) },
+                    ...(['accepted', 'running'].includes(current.status) ? { status: 'failed', success: false, code: 'RESTART_ABORTED', reason: 'Restart monitor exited before entering restart' } : {}),
+                }, ['accepted', 'running'].includes(current.status) ? '重启未能开始，可重新发起' : undefined);
+        }
         removeRestartMonitor();
         // 安装包模式下本进程跑在 stage 副本里。cwd 不在其中（继承自调用方），
         // 所以可以安全自删；留下来只会占磁盘。
@@ -162,21 +190,56 @@ export async function cmdRestartMonitor() {
         const s = scanInstances();
         return s.mains.some(m => m.alive);
     };
-    log('Restart monitor started');
+    const markRequestReady = () => {
+        if (!hasCorrelatedRequest)
+            return;
+        const requesterPid = readRestartRequest(pendingFile, requestId)?.requesterPid;
+        const daemonPid = scanInstances().mains.find(item => item.alive && item.record.pid !== requesterPid)?.record.pid;
+        if (daemonPid && updateRestartRequest(pendingFile, requestId, { status: 'ready', daemonPid })) {
+            if (upgradeJob && !readUpgradeJob(upgradeJob.jobId)?.lifecycle)
+                updateUpgradeJob(upgradeJob.jobId, { lifecycle: { status: 'ready', at: new Date().toISOString() } });
+            log(`Restart request marked ready (daemonPid=${daemonPid})`);
+        }
+        else {
+            log('WARN: could not mark correlated restart request ready');
+        }
+    };
     // Keep this path available for a failed read-only preflight so a stale
     // restart request cannot be replayed by a later daemon startup.
-    const pendingFile = path.join(daemonControlDir(), 'restart-pending.json');
+    log(`Restart monitor started (requestId=${requestId})`);
     // 升级成功时记下版本变化，回执里要告诉 agent 实际跑的是哪个版本。
     let upgradeVersions;
+    if (upgradeJob)
+        updateUpgradeJob(upgradeJob.jobId, { phase: 'preflight', monitorPid: process.pid }, '正在执行原有迁移预检');
     const migrationPreflight = preflightDataMigration(p.root, log);
     if (!migrationPreflight.ok) {
         const message = `重启前用户数据迁移预检失败：${migrationPreflight.reason ?? 'unknown error'}`;
         log(message);
         console.error(`❌ ${message}`);
-        cleanupPendingFile(pendingFile, log);
+        if (upgradeJob)
+            updateUpgradeJob(upgradeJob.jobId, { status: 'failed', success: false, code: 'PREFLIGHT_FAILED', reason: message }, '迁移预检失败，已中断重启');
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exitCode = 2;
         return;
     }
+    if (hasCorrelatedRequest) {
+        if (upgradeJob && (readUpgradeJob(upgradeJob.jobId)?.status !== 'running' || upgradeJob.deadline <= Date.now())) {
+            cleanupPendingFile(pendingFile, log, requestId);
+            process.exitCode = 2;
+            return;
+        }
+        if (!updateRestartRequest(pendingFile, requestId, { status: 'accepted', monitorPid: process.pid })) {
+            log('Restart request changed before acceptance; refusing to stop the daemon');
+            process.exitCode = 2;
+            return;
+        }
+    }
+    updateRestartMonitorPhase(process.pid, 'accepted');
+    log('Restart monitor accepted after migration preflight and monitor election');
+    if (upgradeJob) {
+        updateUpgradeJob(upgradeJob.jobId, { status: 'restart-started', success: true, phase: 'restart' }, '预检通过，已进入重启执行；不代表新进程已就绪');
+        await acknowledgeUpgradeRestart(upgradeJob.jobId);
+    }
     // Test daemons live in one temporary HOME per test, so the current HOME's
     // instance registry cannot see them. Their strict environment markers make
     // them safe to reap globally on every real restart. Keep one process-table
@@ -281,7 +344,7 @@ export async function cmdRestartMonitor() {
         if (exited.some(result => !result)) {
             log('❌ Existing daemon process could not be stopped; refusing to spawn a replacement');
             await notifyDaemonOwners(p, '❌ 旧 EvolCore 进程未能退出，已停止重启，避免重复连接 AUN。', log);
-            cleanupPendingFile(pendingFile, log);
+            cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
             process.exit(1);
         }
         // waitForProcessExit above already observed the old daemon exit. A fixed
@@ -295,7 +358,7 @@ export async function cmdRestartMonitor() {
     const remaining = findOrphanProcesses();
     if (scanInstances().mains.some(item => item.alive) || remaining.some(belongsToRuntime)) {
         log('Restart aborted: a runtime process appeared or survived shutdown');
-        cleanupPendingFile(pendingFile, log);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     const otherHome = remaining.filter(o => !o.confirmedTestDaemon && !belongsToRuntime(o));
@@ -314,7 +377,10 @@ export async function cmdRestartMonitor() {
     // 升级会原地替换整包，先留一个可回退的产物副本。备份放 EVOLCORE_HOME 而非包目录：
     // Linux/macOS 的全局目录常常不可写，而 HOME 一定可写。
     const upgradeBackup = await backupDistForUpgrade(p, log);
-    const [upgrade, ecwebUpgrade] = await applyCachedPackageUpgrades(undefined, { log });
+    const [upgrade] = await applyCachedPackageUpgrades(undefined, {
+        log,
+        ...(upgradeJob ? { target: upgradeJob.target, distTag: upgradeJob.distTag, npmConfig: upgradeJob.npmConfig } : {}),
+    });
     let upgradeRolledBack;
     switch (upgrade.status) {
         case 'upgraded':
@@ -355,19 +421,6 @@ export async function cmdRestartMonitor() {
             await notifyDaemonOwners(p, `⚠️ evolcore 升级失败，使用当前版本继续\n${upgrade.error ?? ''}`.trim(), log);
             break;
     }
-    // EC Web independently applies a valid cached result under the install lock.
-    switch (ecwebUpgrade.status) {
-        case 'upgraded':
-            log(`✅ ${WEB_PACKAGE_NAME} upgraded: ${ecwebUpgrade.from} → ${ecwebUpgrade.to}`);
-            await notifyDaemonOwners(p, `📦 ${WEB_PACKAGE_NAME} 已升级 ${ecwebUpgrade.from} → ${ecwebUpgrade.to}`, log);
-            break;
-        case 'no-update':
-            break;
-        case 'failed':
-            log(`⚠ ${WEB_PACKAGE_NAME} upgrade failed (${ecwebUpgrade.from} → ${ecwebUpgrade.to}): ${ecwebUpgrade.error}`);
-            await notifyDaemonOwners(p, `⚠️ ${WEB_PACKAGE_NAME} 升级失败，使用当前版本继续\n${ecwebUpgrade.error ?? ''}`.trim(), log);
-            break;
-    }
     // 编译在 daemon 停止之后、拉起之前：此时没有活进程占着 dist/，
     // 失败还能回滚到上一份可用产物，因此这一步整体是可撤销的。
     const buildOutcome = await buildSourcesTransactionally(log);
@@ -375,7 +428,7 @@ export async function cmdRestartMonitor() {
         const message = `服务未重启：${buildOutcome.reason ?? '构建失败'}`;
         log(`❌ ${message}`);
         await notifyDaemonOwners(p, `❌ ${message}`, log);
-        markRestartOutcome(pendingFile, 'build-failed', buildOutcome.reason, log);
+        markRestartOutcome(pendingFile, 'build-failed', buildOutcome.reason, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     if (buildOutcome.status === 'rolled-back') {
@@ -386,8 +439,8 @@ export async function cmdRestartMonitor() {
     const migration = applyPendingDataMigration(p.root, log);
     if (!migration.ok) {
         const message = `服务未重启：用户数据迁移失败（${migration.reason ?? 'unknown error'}）`;
-        markRestartOutcome(pendingFile, 'migration-failed', migration.reason, log);
-        cleanupPendingFile(pendingFile, log);
+        markRestartOutcome(pendingFile, 'migration-failed', migration.reason, log, hasCorrelatedRequest ? requestId : undefined);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     // ECWeb is owned by the daemon. Starting it here used to race with the
@@ -404,22 +457,23 @@ export async function cmdRestartMonitor() {
             }, log);
         }
         if (upgradeRolledBack) {
-            markRestartOutcome(pendingFile, 'rolled-back', upgradeRolledBack, log);
+            markRestartOutcome(pendingFile, 'rolled-back', upgradeRolledBack, log, hasCorrelatedRequest ? requestId : undefined);
             log(`✓ Service restarted after upgrade rollback`);
         }
         else if (buildOutcome.status === 'rolled-back') {
             // 让发起方的 agent 知道改动没生效，而不是以为重启成功了。
-            markRestartOutcome(pendingFile, 'rolled-back', buildOutcome.reason, log);
+            markRestartOutcome(pendingFile, 'rolled-back', buildOutcome.reason, log, hasCorrelatedRequest ? requestId : undefined);
             log('✓ Service restarted on the previous build');
         }
         else {
             dropStaleBackup(log);
             dropUpgradeBackup(upgradeBackup, log);
             if (upgradeVersions?.to) {
-                markRestartOutcome(pendingFile, 'upgraded', `已升级 ${upgradeVersions.from} → ${upgradeVersions.to}`, log);
+                markRestartOutcome(pendingFile, 'upgraded', `已升级 ${upgradeVersions.from} → ${upgradeVersions.to}`, log, hasCorrelatedRequest ? requestId : undefined);
             }
             log('✓ Service restarted successfully');
         }
+        markRequestReady();
         archiveSelfHealLog(p, log);
         // 发起会话内的“重启成功”通知由新进程自行发送，此处只负责失败/自愈告警。
         await finishSuccessfulRestart();
@@ -438,7 +492,8 @@ export async function cmdRestartMonitor() {
                     ...upgradeVersions, at: new Date().toISOString(), status: 'rolled-back', error: reason,
                 }, log);
                 log(`✓ ${reason}`);
-                markRestartOutcome(pendingFile, 'rolled-back', reason, log);
+                markRestartOutcome(pendingFile, 'rolled-back', reason, log, hasCorrelatedRequest ? requestId : undefined);
+                markRequestReady();
                 await notifyDaemonOwners(p, `⚠️ ${reason}。当前运行旧版本，请检查后重试。`, log);
                 archiveSelfHealLog(p, log);
                 process.exit(0);
@@ -455,7 +510,8 @@ export async function cmdRestartMonitor() {
             if (started) {
                 const reason = '新编译产物启动失败，已回滚到上一个可用版本';
                 log(`✓ ${reason}`);
-                markRestartOutcome(pendingFile, 'rolled-back', reason, log);
+                markRestartOutcome(pendingFile, 'rolled-back', reason, log, hasCorrelatedRequest ? requestId : undefined);
+                markRequestReady();
                 await notifyDaemonOwners(p, `⚠️ ${reason}。改动未生效，请检查代码后重试。`, log);
                 archiveSelfHealLog(p, log);
                 process.exit(0);
@@ -467,21 +523,21 @@ export async function cmdRestartMonitor() {
     if (p.root.startsWith('/tmp/') || process.env.EVOLCORE_TEST === '1') {
         log('❌ Service failed to start (test environment detected, skipping self-heal)');
         await notifyDaemonOwners(p, '❌ 服务启动失败（测试环境，已跳过自动修复）', log);
-        cleanupPendingFile(pendingFile, log);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     if (hasStartupResourceConflict(p)) {
         const message = '服务启动失败：IPC 管道或端口仍被占用，已跳过自动修复，避免重复连接 AUN。';
         log(message);
         await notifyDaemonOwners(p, `❌ ${message}`, log);
-        cleanupPendingFile(pendingFile, log);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     if (!isRestartRecoveryEnabled()) {
         const message = '服务启动失败，restartRecovery.enabled 已关闭，已跳过 Claude 自动修复。';
         log(`⚠️ ${message}`);
         await notifyDaemonOwners(p, `⚠️ ${message}`, log);
-        cleanupPendingFile(pendingFile, log);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     // 启动失败，进入 self-heal 循环
@@ -493,17 +549,19 @@ export async function cmdRestartMonitor() {
         const message = '找不到 claude CLI，已停止自动修复；请检查后台进程 PATH。';
         log(message);
         await notifyDaemonOwners(p, `❌ ${message}`, log);
-        cleanupPendingFile(pendingFile, log);
+        cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
         process.exit(1);
     }
     for (let attempt = 1; attempt <= MAX_HEAL_ATTEMPTS; attempt++) {
         // 前置检查：服务可能已被上一轮 claude 修复并启动
         if (isServiceAlive()) {
             log(`✓ Service already running before attempt ${attempt}, skipping`);
+            markRequestReady();
             await sendHealSummary(p, attempt - 1, log);
             eventBus.publish({ type: 'self-heal:completed', success: true, attempts: attempt - 1 });
             archiveSelfHealLog(p, log);
-            cleanupPendingFile(pendingFile, log);
+            if (!hasCorrelatedRequest)
+                cleanupPendingFile(pendingFile, log);
             process.exit(0);
         }
         log(`Self-heal attempt ${attempt}/${MAX_HEAL_ATTEMPTS}`);
@@ -513,10 +571,12 @@ export async function cmdRestartMonitor() {
         // 后置检查：不管 invokeClaude 返回什么，都检查服务实际状态
         if (isServiceAlive()) {
             log(`✓ Service is running after attempt ${attempt}`);
+            markRequestReady();
             await sendHealSummary(p, attempt, log);
             eventBus.publish({ type: 'self-heal:completed', success: true, attempts: attempt });
             archiveSelfHealLog(p, log);
-            cleanupPendingFile(pendingFile, log);
+            if (!hasCorrelatedRequest)
+                cleanupPendingFile(pendingFile, log);
             process.exit(0);
         }
         if (!healed) {
@@ -527,10 +587,12 @@ export async function cmdRestartMonitor() {
         started = await spawnAndWaitReady(p, log, READY_TIMEOUT);
         if (started) {
             log(`✓ Self-heal succeeded on attempt ${attempt}`);
+            markRequestReady();
             await sendHealSummary(p, attempt, log);
             eventBus.publish({ type: 'self-heal:completed', success: true, attempts: attempt });
             archiveSelfHealLog(p, log);
-            cleanupPendingFile(pendingFile, log);
+            if (!hasCorrelatedRequest)
+                cleanupPendingFile(pendingFile, log);
             process.exit(0);
         }
         log(`Attempt ${attempt}: still failing after fix`);
@@ -538,16 +600,18 @@ export async function cmdRestartMonitor() {
     // 全部失败 — 最后再检查一次
     if (isServiceAlive()) {
         log('✓ Service recovered during final check');
+        markRequestReady();
         await sendHealSummary(p, MAX_HEAL_ATTEMPTS, log);
         eventBus.publish({ type: 'self-heal:completed', success: true, attempts: MAX_HEAL_ATTEMPTS });
         archiveSelfHealLog(p, log);
-        cleanupPendingFile(pendingFile, log);
+        if (!hasCorrelatedRequest)
+            cleanupPendingFile(pendingFile, log);
         process.exit(0);
     }
     log(`❌ All ${MAX_HEAL_ATTEMPTS} self-heal attempts failed`);
     eventBus.publish({ type: 'self-heal:completed', success: false, attempts: MAX_HEAL_ATTEMPTS });
     await notifyDaemonOwners(p, `❌ ${MAX_HEAL_ATTEMPTS} 次自动修复均失败，需要人工介入。\n修复记录：${p.selfHealLog}`, log);
-    cleanupPendingFile(pendingFile, log);
+    cleanupPendingFile(pendingFile, log, hasCorrelatedRequest ? requestId : undefined);
     process.exit(1);
 }
 /**
@@ -803,11 +867,15 @@ function runBuildCommand(log) {
  * code change took effect. Older daemons ignore unknown fields, so writing this
  * stays safe when the rolled-back artifact predates the field.
  */
-function markRestartOutcome(pendingFile, outcome, reason, log) {
+function markRestartOutcome(pendingFile, outcome, reason, log, requestId) {
     try {
         if (!fs.existsSync(pendingFile))
             return;
         const info = JSON.parse(fs.readFileSync(pendingFile, 'utf-8'));
+        if (requestId && info.requestId !== requestId)
+            return;
+        if (typeof info.upgradeJobId === 'string')
+            updateUpgradeJob(info.upgradeJobId, { lifecycle: { status: outcome, reason, at: new Date().toISOString() } });
         info.outcome = outcome;
         if (reason)
             info.outcomeReason = reason;
@@ -920,9 +988,25 @@ function dropUpgradeBackup(backupDir, log) {
         log('WARN: could not remove upgrade backup');
     }
 }
-function cleanupPendingFile(filePath, log) {
+function cleanupPendingFile(filePath, log, requestId) {
     try {
         if (fs.existsSync(filePath)) {
+            if (requestId) {
+                const request = readRestartRequest(filePath, requestId);
+                if (!request)
+                    return;
+                if (request.status === 'ready') {
+                    log('Preserved restart-pending.json until the new daemon sends its completion notice');
+                    return;
+                }
+                removeRestartRequest(filePath, requestId);
+                try {
+                    fs.rmSync(`${filePath}_`, { force: true });
+                }
+                catch { }
+                log('Cleaned up correlated restart-pending.json');
+                return;
+            }
             fs.unlinkSync(filePath);
             log('Cleaned up restart-pending.json');
         }
@@ -961,6 +1045,7 @@ async function spawnAndWaitReady(p, log, timeout) {
             ...process.env,
             EVOLCORE_HOME: p.root,
             EVOLCORE_LAUNCHED_BY: 'restart-monitor',
+            [RESTART_REQUEST_ID_ENV]: '',
             LOG_LEVEL: process.env.LOG_LEVEL || 'INFO',
             MESSAGE_LOG: process.env.MESSAGE_LOG || 'true',
             EVENT_LOG: process.env.EVENT_LOG || 'true',
@@ -1099,50 +1184,57 @@ function archiveSelfHealLog(p, log) {
     log(`Archived self-heal log to ${archivePath}`);
 }
 async function notifyDaemonOwners(p, message, log) {
-    const cfg = loadDaemonConfig();
-    const daemonAid = typeof cfg.aid === 'string' ? cfg.aid.trim() : '';
-    if (!daemonAid) {
-        log('Daemon owner notification skipped: daemon.json.aid not configured');
-        return;
-    }
-    if (!isValidAid(daemonAid)) {
-        log(`Daemon owner notification skipped: invalid daemon.json.aid ${daemonAid}`);
-        return;
-    }
-    const owners = Array.from(new Set((Array.isArray(cfg.owners) ? cfg.owners : [])
-        .map(owner => typeof owner === 'string' ? owner.trim() : '')
-        .filter(Boolean)));
-    if (owners.length === 0) {
-        log('Daemon owner notification skipped: daemon.json.owners not configured');
-        return;
-    }
-    const validOwners = owners.filter(owner => {
-        const valid = isValidAid(owner);
-        if (!valid)
-            log(`Daemon owner notification skipped invalid owner AID: ${owner}`);
-        return valid;
-    });
-    if (validOwners.length === 0)
-        return;
-    const text = formatDaemonOwnerNotice(p, message);
-    for (const owner of validOwners) {
-        try {
-            const result = await withTimeout(msgSend({
-                from: daemonAid,
-                to: owner,
-                body: { mode: 'text', text },
-                slotId: 'restart-monitor',
-            }), OWNER_NOTIFICATION_TIMEOUT_MS);
-            if (!result.ok) {
-                log(`Daemon owner notification failed for ${owner}: ${result.error}`);
-                continue;
-            }
-            log(`Daemon owner notification sent to ${owner}: ${result.message_id}`);
+    try {
+        const cfg = loadDaemonConfig();
+        const daemonAid = typeof cfg.aid === 'string' ? cfg.aid.trim() : '';
+        if (!daemonAid) {
+            log('Daemon owner notification skipped: daemon.json.aid not configured');
+            return;
         }
-        catch (error) {
-            log(`Daemon owner notification failed for ${owner}: ${error.message?.slice(0, 200) || error}`);
+        if (!isValidAid(daemonAid)) {
+            log(`Daemon owner notification skipped: invalid daemon.json.aid ${daemonAid}`);
+            return;
+        }
+        const response = await ipcQuery(p.socket, { type: 'aid.ownership', argv: ['aid', 'bind', 'status'] }, OWNER_NOTIFICATION_TIMEOUT_MS).catch(() => null);
+        const state = response?.ok ? response.state : undefined;
+        const owners = state?.subjectAid === daemonAid && state?.consensusStatus === 'consensus'
+            && typeof state.ownerAid === 'string' && isValidAid(state.ownerAid) ? [state.ownerAid] : [];
+        if (owners.length === 0) {
+            log('Daemon owner notification skipped: fresh protocol consensus unavailable');
+            return;
+        }
+        const validOwners = owners.filter(owner => {
+            const valid = isValidAid(owner);
+            if (!valid)
+                log(`Daemon owner notification skipped invalid owner AID: ${owner}`);
+            return valid;
+        });
+        if (validOwners.length === 0)
+            return;
+        const text = formatDaemonOwnerNotice(p, message);
+        for (const owner of validOwners) {
+            try {
+                const result = await withTimeout(msgSend({
+                    from: daemonAid,
+                    to: owner,
+                    body: { mode: 'text', text },
+                    slotId: 'restart-monitor',
+                }), OWNER_NOTIFICATION_TIMEOUT_MS);
+                if (!result.ok) {
+                    log(`Daemon owner notification failed for ${owner}: ${result.error}`);
+                    continue;
+                }
+                log(`Daemon owner notification sent to ${owner}: ${result.message_id}`);
+            }
+            catch (error) {
+                log(`Daemon owner notification failed for ${owner}: ${error.message?.slice(0, 200) || error}`);
+            }
         }
     }
+    catch (error) {
+        const reason = error instanceof Error ? error.message : String(error);
+        log(`WARN: Daemon owner notification failed: ${reason.slice(0, 200)}`);
+    }
 }
 function formatDaemonOwnerNotice(p, message) {
     return [

package/dist/cli/session-command.js

@@ -149,7 +149,7 @@ async function cmdFind(args) {
         console.log(`ec 会话 id: ${f.sessionId}`);
         console.log(`  agentSessionId: ${f.agentSessionId ?? '(空)'}`);
         console.log(`  baseagent: ${f.baseagent ?? '(未知)'}`);
-        console.log(`  threadId: ${f.threadId ? f.threadId : '(主会话)'}`);
+        console.log(`  threadId: ${f.threadId || '(主会话)'}`);
         if (f.name)
             console.log(`  名称: ${f.name}`);
         if (f.chatType)

package/dist/cli/stats.js

@@ -41,8 +41,8 @@ Usage: ec stats [options]
   --groups [--limit N]    群聊列表（同上）
   --summary               指定时间范围总消耗汇总（token/USD/CNY）
   --peer-detail <id>      指定对端 AID 或 peer_key，按天返回消耗明细
-  --task-calls <taskId>   一个 task 的逐次大模型调用明细
-  --session-calls <id>    一个会话的逐次大模型调用明细
+  --task-calls <taskId>   一个 task 的模型用量明细（Codex 为回合/子 agent 汇总）
+  --session-calls <id>    一个会话的模型用量明细（Codex 为回合/子 agent 汇总）
 
 输出格式:
   --format json           JSON 输出
@@ -417,7 +417,7 @@ export async function handleStats(args) {
         console.log();
         return;
     }
-    // 逐次大模型调用明细
+    // 模型用量明细
     if ((flags['task-calls'] && typeof flags['task-calls'] === 'string') ||
         (flags['session-calls'] && typeof flags['session-calls'] === 'string')) {
         const byTask = typeof flags['task-calls'] === 'string';
@@ -431,13 +431,15 @@ export async function handleStats(args) {
             console.log('No model calls found.');
             return;
         }
-        console.log(`\n${BOLD}📊 大模型调用明细 — ${key}${RESET}\n`);
-        const headers = ['#', 'Model', 'Input', 'Output', 'CacheR', 'CacheW', 'Task', 'Deg'];
+        console.log(`\n${BOLD}📊 模型用量明细 — ${key}${RESET}\n`);
+        const headers = ['#', 'Model', 'Source', 'Input', 'Output', 'CacheR', 'CacheW', 'Provider Thread', 'Task', 'Deg'];
         const tableRows = rows.map(r => [
             String(r.call_index),
             r.model.split('-').slice(0, 3).join('-'),
+            r.source,
             fmtTokens(r.input_tokens), fmtTokens(r.output_tokens),
             fmtTokens(r.cache_read_tokens), fmtTokens(r.cache_creation_tokens),
+            r.provider_thread_id ?? '',
             r.task_id, r.degraded ? 'Y' : '',
         ]);
         printTable(headers, tableRows);