agentworkshop 0.8.6 → 0.8.8

package/server/services/workshop/daq/daq-controller/sweep-alarms.ts

@@ -1,186 +1,214 @@
-/**
- * DaqControllerSweepAlarms —— 巡检 / 运行时同步 / 告警处理与确认
- * (拆分层,承 DaqControllerIngest;方法体与原文件逐行一致)
- */
-import { DaqControllerIngest } from './ingest'
-import type { DaqRuntimeHost } from '../daq-runtime'
-import { AppError } from '../../../../utils/errors'
-import type { DaqNode } from '../daq-node'
-import { DaqNodeRuntime } from '../daq-runtime'
-import { SWEEP_MAX_CONCURRENCY, log } from './helpers'
-import { getDaqHostPorts } from '../host-ports'
-import { getOps, recordOps } from '../../ops/ops'
-import { getActiveLineRun } from '../../dcw/line-run'
-import { newAlarmId, notifyAlarm } from '../alarm-notify'
-import { resolveDaqDriver } from '../drivers'
-
-export abstract class DaqControllerSweepAlarms extends DaqControllerIngest {
-  /** 网关统一调度:250ms 扫描,到期判定与互斥由各运行时私有节拍自治(单节点慢/停不波及邻居)。
-   *  产线门控:未选定配方(无活动 LineRun)不执行采集 —— 采样与实时下发均由配方驱动。
-   *  推送式协议(MQTT)不走本循环采样:订阅在 syncPushBindings 建立,报文到达即入库;
-   *  该循环只负责订阅对账与陈值看门狗。 */
-  protected sweep(): void {
-    if (!this.pipelineReady || !this.running) return
-    // 订阅对账与看门狗不依赖"是否有活动批次"(未开跑也保持 broker 在线,开跑即恢复入库)
-    this.syncPushBindings()
-    const now = Date.now()
-    this.checkPushLiveness(now)
-    const host = getDaqHostPorts()
-    if (!host || !host.lineRun.hasAnyActiveRun()) return
-    // 在飞闸门:上一拍未消化的并发额度不叠加(避免积压 tick 雪崩)
-    const budget = SWEEP_MAX_CONCURRENCY - this.samplingInFlight
-    if (budget <= 0) return
-    const defs = this.runtimeDefaults()
-    const rts = [...this.runtimes.values()]
-    const total = rts.length
-    if (!total) return
-    let started = 0
-    let examined = 0
-    // 轮转游标 + 到期预判:派发额度有限时,固定从表头遍历会让表尾节点永久饿死
-    // (新入表节点排在最后 → 永远轮不到)。游标保证有限拍内覆盖全表,isDue 保证
-    // 未到期节点不占用额度(否则表头节点每拍吃满额度,即便它们无样本可采)。
-    for (let i = 0; i < total && started < budget; i++) {
-      const rt = rts[(this.sweepCursor + i) % total]!
-      examined = i + 1
-      // 推送式驱动(存在 subscribe)不轮询:否则同一条报文会被采样 tick 重复入库
-      if (typeof resolveDaqDriver(rt.node.driver).subscribe === 'function') continue
-      // 逐产线门控:节点只在其所属产线的活动批次窗口内采集(lineId 空 = 未分配,不采集)
-      if (!host.lineRun.activeRun(rt.node.lineId)) continue
-      if (!rt.isDue(now, defs)) continue
-      this.samplingInFlight += 1
-      started += 1
-      void rt.tick(now).catch(() => { /* 单节点采样异常已在 runtime 内隔离 */ })
-        .finally(() => { this.samplingInFlight -= 1 })
-    }
-    this.sweepCursor = (this.sweepCursor + examined) % total
-  }
-
-  /** 产线停止:该产线全部节点置 offline(开跑后由采样自然恢复) */
-  markLineOffline(lineId: string): void {
-    for (const n of this.repo.all()) {
-      if (n.enabled && n.lineId === lineId) n.state = 'offline'
-    }
-    this.emitController()
-  }
-
-  /** 运行时注册表对账(仓库为权威:增删节点/启动即对齐;返回新建的运行时) */
-  protected syncRuntimes(): void {
-    const live = new Set<string>()
-    for (const node of this.repo.all()) {
-      live.add(node.id)
-      if (!this.runtimes.has(node.id)) {
-        this.runtimes.set(node.id, new DaqNodeRuntime(node, this.host))
-      }
-    }
-    for (const id of [...this.runtimes.keys()]) {
-      if (!live.has(id)) this.runtimes.delete(id)
-    }
-  }
-
-  /** runtime host:网关注入的服务面(采样/入队/管线/告警/全局缺省) */
-  protected host: DaqRuntimeHost = {
-    defaults: () => this.runtimeDefaults(),
-    sample: (node, now) => this.sampleNode(node, now),
-    publishSample: env => this.publishSample(env),
-    ingest: (node, env, allowPublish) => this.ingestNode(node, env, allowPublish),
-    broadcastError: (node, message) => this.broadcastDriverError(node, message),
-    onAlarm: (node, value, rule, threshold) => this.handleAlarm(node, value, rule, threshold),
-    onAlarmRecover: (node, value) => this.handleAlarmRecover(node, value),
-    recipeWindowFor: node => this.recipeDaqWindowFor(node),
-  }
-
-  // ---------- S5:报警持久化/外送/确认 ----------
-
-  /** 累计报警次数(R4 指标暴露) */
-  alarmsRaised = 0
-
-  /** alarm 进入沿:落库(同节点同量未确认幂等)→ WS 广播 → webhook 外送;失败绝不影响采集。
-   *  metricKey:帧派生指标告警的键(模板 metrics 规则;缺省 = 模板 key,标量越限) */
-  protected handleAlarm(node: DaqNode, value: number, rule: 'lt-min' | 'gt-max', threshold: number, metricKey?: string): void {
-    this.alarmsRaised++
-    const repo = getOps()?.alarmEvents
-    let id = newAlarmId()
-    const createdAt = new Date().toISOString()
-    const metric = metricKey ?? node.templateKey
-    if (repo) {
-      try {
-        const raised = repo.raise({
-          id, nodeId: node.id, nodeName: node.name, metric,
-          value, rule, threshold, createdAt,
-        })
-        if (!raised) id = '' // 已有同源未确认报警:不重复广播/外送
-      }
-      catch (err) {
-        log.error('[daq-alarm] 报警落库失败(不影响采集):', err instanceof Error ? err.message : err)
-      }
-    }
-    if (!id) return
-    const payload = {
-      id, nodeId: node.id, nodeName: node.name, metric,
-      value, rule, threshold, escalation: 0, createdAt,
-      // S14:告警帧必须带产线归属,否则 scene-events 会当作"线无关"广播给所有 peer
-      lineId: node.lineId || null,
-    }
-    this.broadcast?.('daq.alarm', payload)
-    notifyAlarm(payload)
-    // 运维日志:越限告警入实时事件轨(恢复/详情在告警面板与日志管理);
-    // 告警发生在活动批次内时带上该批次的 product/recipe,产线日志才能按配方检索
-    const zh = rule === 'gt-max' ? '高于上限' : '低于下限'
-    const run = getActiveLineRun(node.lineId)
-    recordOps({
-      actor: 'system',
-      actorName: 'system',
-      actorKind: 'system',
-      action: 'daq.alarm.raise',
-      kind: 'alarm',
-      targetKind: 'daq-node',
-      targetId: node.id,
-      summary: `告警:「${node.name}」${metric} ${zh}阈值 ${threshold}(当前 ${value})`,
-      lineId: node.lineId ?? '',
-      productId: run?.productId ?? '',
-      recipeId: run?.recipeId ?? '',
-      detail: { alarmId: id, metric, value, rule, threshold },
-    })
-  }
-
-  /** alarm 恢复沿:广播恢复(报警保持 open 待人工 ack) */
-  protected handleAlarmRecover(node: DaqNode, value: number): void {
-    this.broadcast?.('daq.alarm.changed', { nodeId: node.id, nodeName: node.name, recovered: true, value, at: new Date().toISOString(), lineId: node.lineId || null })
-  }
-
-  /** 报警确认(HITL 闭环的人为一步;幂等:已确认返回 false) */
-  ackAlarm(id: string, byUserId: string, byName: string): boolean {
-    const repo = getOps()?.alarmEvents
-    if (!repo) throw new AppError(503, 'UNAVAILABLE', '报警持久化未就绪')
-    const ok = repo.ack(id, byUserId, byName, new Date().toISOString())
-    if (ok) {
-      this.broadcast?.('daq.alarm.changed', { id, ackedBy: byName, ackedAt: new Date().toISOString() })
-      // 确认是人对产线告警的处置动作:回填报警所属节点/产线/活动批次,避免落成无归属条目
-      const nodeId = repo.nodeIdById(id) ?? ''
-      const node = nodeId ? this.repo.byId(nodeId) : undefined
-      const run = node ? getActiveLineRun(node.lineId) : null
-      recordOps({
-        actor: byUserId,
-        actorName: byName,
-        actorKind: 'user',
-        action: 'daq.alarm.ack',
-        kind: 'alarm',
-        targetKind: 'alarm',
-        targetId: id,
-        lineId: node?.lineId ?? '',
-        productId: run?.productId ?? '',
-        recipeId: run?.recipeId ?? '',
-        summary: `确认报警 ${id}${node ? `(节点「${node.name}」)` : ''}(处理完毕,移出实时告警)`,
-      })
-    }
-    return ok
-  }
-
-  listAlarms(scope: 'open' | 'all', limit = 100) {
-    const repo = getOps()?.alarmEvents
-    if (!repo) return []
-    return scope === 'open' ? repo.listOpen(limit) : repo.list(limit)
-  }
-
-  // ---------- 消费者:队列帧 → 存储/广播/回写 ----------
-}
+/**
+ * DaqControllerSweepAlarms —— 巡检 / 运行时同步 / 告警处理与确认
+ * (拆分层,承 DaqControllerIngest;方法体与原文件逐行一致)
+ */
+import { DaqControllerIngest } from './ingest'
+import type { DaqRuntimeHost } from '../daq-runtime'
+import { AppError } from '../../../../utils/errors'
+import type { DaqNode } from '../daq-node'
+import { DaqNodeRuntime } from '../daq-runtime'
+import { SWEEP_MAX_CONCURRENCY, log } from './helpers'
+import { getDaqHostPorts } from '../host-ports'
+import { getOps, recordOps } from '../../ops/ops'
+import { getActiveLineRun } from '../../dcw/line-run'
+import { newAlarmId, notifyAlarm } from '../alarm-notify'
+import { resolveDaqDriver } from '../drivers'
+
+export abstract class DaqControllerSweepAlarms extends DaqControllerIngest {
+  /** 网关统一调度:250ms 扫描,到期判定与互斥由各运行时私有节拍自治(单节点慢/停不波及邻居)。
+   *  产线门控:未选定配方(无活动 LineRun)不执行采集 —— 采样与实时下发均由配方驱动。
+   *  推送式协议(MQTT)不走本循环采样:订阅在 syncPushBindings 建立,报文到达即入库;
+   *  该循环只负责订阅对账与陈值看门狗。 */
+  protected sweep(): void {
+    if (!this.pipelineReady || !this.running) return
+    // 250ms 定时器回调:任何未捕获异常都会走到进程级 uncaughtException,
+    // 而 stability-guard 只豁免 4 类已知噪声,其余一律 exit(1) —— 一处普通 bug 就能打崩生产。
+    // 逐段兜底(与 runtime/manager/state.ts 同口径):失败只记日志,绝不抛出,
+    // 且保证后面的段仍会执行(此前一拍抛错会跳过同拍剩余逻辑)。
+    this.guardStep('syncPushBindings', () => this.syncPushBindings())
+    const now = Date.now()
+    this.guardStep('checkPushLiveness', () => this.checkPushLiveness(now))
+    const host = getDaqHostPorts()
+    if (!host || !host.lineRun.hasAnyActiveRun()) return
+    // 在飞闸门:上一拍未消化的并发额度不叠加(避免积压 tick 雪崩)
+    const budget = SWEEP_MAX_CONCURRENCY - this.samplingInFlight
+    if (budget <= 0) return
+    let defs: ReturnType<typeof this.runtimeDefaults>
+    let rts: typeof this.runtimes extends Map<string, infer V> ? V[] : never
+    try {
+      defs = this.runtimeDefaults()
+      rts = [...this.runtimes.values()]
+    }
+    catch (err) {
+      log.error('[daq] sweep 准备阶段异常(已隔离):', err)
+      return
+    }
+    const total = rts.length
+    if (!total) return
+    let started = 0
+    let examined = 0
+    // 轮转游标 + 到期预判:派发额度有限时,固定从表头遍历会让表尾节点永久饿死
+    // (新入表节点排在最后 → 永远轮不到)。游标保证有限拍内覆盖全表,isDue 保证
+    // 未到期节点不占用额度(否则表头节点每拍吃满额度,即便它们无样本可采)。
+    for (let i = 0; i < total && started < budget; i++) {
+      const rt = rts[(this.sweepCursor + i) % total]!
+      examined = i + 1
+      try {
+        // 推送式驱动(存在 subscribe)不轮询:否则同一条报文会被采样 tick 重复入库
+        if (typeof resolveDaqDriver(rt.node.driver).subscribe === 'function') continue
+        // 逐产线门控:节点只在其所属产线的活动批次窗口内采集(lineId 空 = 未分配,不采集)
+        if (!host.lineRun.activeRun(rt.node.lineId)) continue
+        if (!rt.isDue(now, defs)) continue
+      }
+      catch (err) {
+        // 单节点判定失败不阻断同拍其余节点
+        log.error(`[daq] 节点 ${rt.node.id} 到期判定异常(已隔离):`, err)
+        continue
+      }
+      this.samplingInFlight += 1
+      started += 1
+      void rt.tick(now).catch(() => { /* 单节点采样异常已在 runtime 内隔离 */ })
+        .finally(() => { this.samplingInFlight -= 1 })
+    }
+    this.sweepCursor = (this.sweepCursor + examined) % total
+  }
+
+  /** 定时器回调内的安全包裹:异常只记日志,绝不冒泡到进程级 */
+  private guardStep(label: string, fn: () => void): void {
+    try {
+      fn()
+    }
+    catch (err) {
+      log.error(`[daq] sweep ${label} 异常(已隔离):`, err)
+    }
+  }
+
+  /** 产线停止:该产线全部节点置 offline(开跑后由采样自然恢复) */
+  markLineOffline(lineId: string): void {
+    for (const n of this.repo.all()) {
+      if (n.enabled && n.lineId === lineId) n.state = 'offline'
+    }
+    this.emitController()
+  }
+
+  /** 运行时注册表对账(仓库为权威:增删节点/启动即对齐;返回新建的运行时) */
+  protected syncRuntimes(): void {
+    const live = new Set<string>()
+    for (const node of this.repo.all()) {
+      live.add(node.id)
+      if (!this.runtimes.has(node.id)) {
+        this.runtimes.set(node.id, new DaqNodeRuntime(node, this.host))
+      }
+    }
+    for (const id of [...this.runtimes.keys()]) {
+      if (!live.has(id)) this.runtimes.delete(id)
+    }
+  }
+
+  /** runtime host:网关注入的服务面(采样/入队/管线/告警/全局缺省) */
+  protected host: DaqRuntimeHost = {
+    defaults: () => this.runtimeDefaults(),
+    sample: (node, now) => this.sampleNode(node, now),
+    publishSample: env => this.publishSample(env),
+    ingest: (node, env, allowPublish) => this.ingestNode(node, env, allowPublish),
+    broadcastError: (node, message) => this.broadcastDriverError(node, message),
+    onAlarm: (node, value, rule, threshold) => this.handleAlarm(node, value, rule, threshold),
+    onAlarmRecover: (node, value) => this.handleAlarmRecover(node, value),
+    recipeWindowFor: node => this.recipeDaqWindowFor(node),
+  }
+
+  // ---------- S5:报警持久化/外送/确认 ----------
+
+  /** 累计报警次数(R4 指标暴露) */
+  alarmsRaised = 0
+
+  /** alarm 进入沿:落库(同节点同量未确认幂等)→ WS 广播 → webhook 外送;失败绝不影响采集。
+   *  metricKey:帧派生指标告警的键(模板 metrics 规则;缺省 = 模板 key,标量越限) */
+  protected handleAlarm(node: DaqNode, value: number, rule: 'lt-min' | 'gt-max', threshold: number, metricKey?: string): void {
+    this.alarmsRaised++
+    const repo = getOps()?.alarmEvents
+    let id = newAlarmId()
+    const createdAt = new Date().toISOString()
+    const metric = metricKey ?? node.templateKey
+    if (repo) {
+      try {
+        const raised = repo.raise({
+          id, nodeId: node.id, nodeName: node.name, metric,
+          value, rule, threshold, createdAt,
+        })
+        if (!raised) id = '' // 已有同源未确认报警:不重复广播/外送
+      }
+      catch (err) {
+        log.error('[daq-alarm] 报警落库失败(不影响采集):', err instanceof Error ? err.message : err)
+      }
+    }
+    if (!id) return
+    const payload = {
+      id, nodeId: node.id, nodeName: node.name, metric,
+      value, rule, threshold, escalation: 0, createdAt,
+      // S14:告警帧必须带产线归属,否则 scene-events 会当作"线无关"广播给所有 peer
+      lineId: node.lineId || null,
+    }
+    this.broadcast?.('daq.alarm', payload)
+    notifyAlarm(payload)
+    // 运维日志:越限告警入实时事件轨(恢复/详情在告警面板与日志管理);
+    // 告警发生在活动批次内时带上该批次的 product/recipe,产线日志才能按配方检索
+    const zh = rule === 'gt-max' ? '高于上限' : '低于下限'
+    const run = getActiveLineRun(node.lineId)
+    recordOps({
+      actor: 'system',
+      actorName: 'system',
+      actorKind: 'system',
+      action: 'daq.alarm.raise',
+      kind: 'alarm',
+      targetKind: 'daq-node',
+      targetId: node.id,
+      summary: `告警:「${node.name}」${metric} ${zh}阈值 ${threshold}(当前 ${value})`,
+      lineId: node.lineId ?? '',
+      productId: run?.productId ?? '',
+      recipeId: run?.recipeId ?? '',
+      detail: { alarmId: id, metric, value, rule, threshold },
+    })
+  }
+
+  /** alarm 恢复沿:广播恢复(报警保持 open 待人工 ack) */
+  protected handleAlarmRecover(node: DaqNode, value: number): void {
+    this.broadcast?.('daq.alarm.changed', { nodeId: node.id, nodeName: node.name, recovered: true, value, at: new Date().toISOString(), lineId: node.lineId || null })
+  }
+
+  /** 报警确认(HITL 闭环的人为一步;幂等:已确认返回 false) */
+  ackAlarm(id: string, byUserId: string, byName: string): boolean {
+    const repo = getOps()?.alarmEvents
+    if (!repo) throw new AppError(503, 'UNAVAILABLE', '报警持久化未就绪')
+    const ok = repo.ack(id, byUserId, byName, new Date().toISOString())
+    if (ok) {
+      this.broadcast?.('daq.alarm.changed', { id, ackedBy: byName, ackedAt: new Date().toISOString() })
+      // 确认是人对产线告警的处置动作:回填报警所属节点/产线/活动批次,避免落成无归属条目
+      const nodeId = repo.nodeIdById(id) ?? ''
+      const node = nodeId ? this.repo.byId(nodeId) : undefined
+      const run = node ? getActiveLineRun(node.lineId) : null
+      recordOps({
+        actor: byUserId,
+        actorName: byName,
+        actorKind: 'user',
+        action: 'daq.alarm.ack',
+        kind: 'alarm',
+        targetKind: 'alarm',
+        targetId: id,
+        lineId: node?.lineId ?? '',
+        productId: run?.productId ?? '',
+        recipeId: run?.recipeId ?? '',
+        summary: `确认报警 ${id}${node ? `(节点「${node.name}」)` : ''}(处理完毕,移出实时告警)`,
+      })
+    }
+    return ok
+  }
+
+  listAlarms(scope: 'open' | 'all', limit = 100) {
+    const repo = getOps()?.alarmEvents
+    if (!repo) return []
+    return scope === 'open' ? repo.listOpen(limit) : repo.list(limit)
+  }
+
+  // ---------- 消费者:队列帧 → 存储/广播/回写 ----------
+}

package/server/services/workshop/daq/daq-controller/types.ts

@@ -27,6 +27,11 @@ export interface DaqCreateInput {
   /** 节点级采集语义备注(覆盖模板) */
   semantics?: string
   deviceBindingId?: string | null
+  /** MQTT 单 topic 多字段:声明后展开为多个子信号节点(每字段一个,独立量程/告警)。
+   *  也可放在 driverConfig.fields(前端动态表单的文本域形态) */
+  fields?: unknown
+  /** 多字段路径公共前缀(可选;与 fields 的 path 拼接) */
+  fieldsBase?: string
 }
 
 export interface DaqPatchInput {

package/server/services/workshop/daq/daq-node.ts

@@ -281,6 +281,13 @@ export class DaqNode {
       semantics: this.semantics,
       chamber: this.chamber,
       unitPath: this.unitPath,
+      // 多字段组:driverConfig 里存 groupId/label(随节点持久化),视图面提升为一级字段
+      fieldGroup: typeof this.driverConfig.fieldGroup === 'string' && this.driverConfig.fieldGroup
+        ? this.driverConfig.fieldGroup
+        : undefined,
+      fieldLabel: typeof this.driverConfig.fieldLabel === 'string' && this.driverConfig.fieldLabel
+        ? this.driverConfig.fieldLabel
+        : undefined,
       value: this.value,
       state: this.state,
       lastAt: this.lastAt,