agentworkshop 0.7.66 → 0.7.70

package/server/plugins-builtin/diag-bridge/host/plugin.mjs

@@ -1,420 +1,457 @@
-/**
- * 插件对象(settings / routes / tools / hooks)
- * (由 server/plugins-builtin/diag-bridge/index.mjs 按职责拆出;内容逐行原文搬运)
- */
-import { AUTO_COOLDOWN_MS, DEFAULT_BASE, MIN } from './constants.mjs'
-import { authHeadersOf, authHint, autoEnabledOf, autoRulesOf, baseOf, diagTokenOf, harnessOf, jget, kvRuns, parseAutoRules, runKey } from './helpers.mjs'
-import { snapshotCore } from './snapshot.mjs'
-import { startDiagnosis } from './diagnosis.mjs'
-import { uploadExportDataset } from './exports.mjs'
-import { sweepOnce } from './sweep.mjs'
-import { intentLabel } from './intents.mjs'
-
-export default {
-  name: 'diag-bridge',
-  version: '2.2.0',
-  description: '深度诊断桥接 v2.2:异步任务式深度诊断(提交即返 task_id)+ daq-export 数据集多文件分析(diag_run export_id= 全量时序+manifest)+ 同步等待模式(mode=sync)+ 完成自动回执 Channel(修复发起者身份丢失)+ 报告在线查看。完成后经 rag-bridge kb_agent 入库知识库。',
-  auth: 'user',
-  client: './client.mjs', // 前端面板(插件页注入;i18n 见 i18n.json)
-  // 插件配置分组(声明式):连接 / 执行 / 自动化 三个独立分区
-  configGroups: [
-    { id: 'default', labelKey: 'plugin.diag-bridge.group.conn', label: '诊断服务连接', description: '后端地址与鉴权 Token', order: 420 },
-    { id: 'run', labelKey: 'plugin.diag-bridge.group.run', label: '诊断执行参数', description: 'diag_run 的引擎与预算', order: 430 },
-    { id: 'auto', labelKey: 'plugin.diag-bridge.group.auto', label: '自动诊断', description: '越限自动发起诊断(DAQ 采样事件驱动)', order: 440, collapsed: true },
-  ],
-  // 插件设置声明(key 编址 plugins.diag-bridge.<key>;labelKey 解析 i18n.json)
-  settings: [
-    { key: 'base_url', type: 'string', default: DEFAULT_BASE, group: 'default', labelKey: 'plugin.diag-bridge.settings.base_url', label: '诊断服务地址', description: 'industrial-deep-diagnostic 后端(http/https,host 限 127.0.0.1/localhost);保存即热生效' },
-    { key: 'token', type: 'string', default: '', group: 'default', labelKey: 'plugin.diag-bridge.settings.token', label: '诊断 API Token', description: '诊断服务开启鉴权时(v4 起默认强制)的 API Token(Authorization: Bearer);空=匿名;保存即热生效' },
-    { key: 'harness', type: 'string', default: '', group: 'run', labelKey: 'plugin.diag-bridge.settings.harness', label: '诊断引擎', description: 'diag_run 用的引擎 id(omp/claude/mock 等;空=取 kv/omp);无 Anthropic key 的机器请用 omp' },
-    { key: 'max_turns', type: 'number', default: 0, min: 0, max: 2000, group: 'run', labelKey: 'plugin.diag-bridge.settings.max_turns', label: '诊断最大轮数', description: 'diagnosis/start 的 maxTurns(0=取 kv/内置默认 220)' },
-    { key: 'max_minutes', type: 'number', default: 0, min: 0, max: 720, group: 'run', labelKey: 'plugin.diag-bridge.settings.max_minutes', label: '诊断超时(分钟)', description: 'diagnosis/start 的 timeoutMinutes(0=取 kv/内置默认 40)' },
-    { key: 'auto_enabled', type: 'boolean', default: false, group: 'auto', labelKey: 'plugin.diag-bridge.settings.auto_enabled', label: '自动诊断', description: 'daq:sample 越限命中规则时自动发起深度诊断(每产线冷却 30 分钟)' },
-  ],
-  setup(ctx) {
-    // 运行时追加分区与字段(ctx.config.defineGroup / defineField):
-    // 字段 key 与声明式完全等价(plugins.diag-bridge.auto_rules),已保存的值不丢;
-    // 这里演示「插件在 setup 里按条件调用平台 API 扩展自己的配置面」。
-    ctx.config.defineGroup({
-      id: 'rules',
-      labelKey: 'plugin.diag-bridge.group.rules',
-      label: '自动诊断规则',
-      description: 'JSON 规则;命中即触发自动诊断(仅 auto_enabled=true 时生效)',
-      order: 450,
-    })
-    ctx.config.defineField({
-      key: 'auto_rules',
-      type: 'string',
-      default: '',
-      group: 'rules',
-      labelKey: 'plugin.diag-bridge.settings.auto_rules',
-      label: '自动诊断规则',
-      description: 'JSON: {"节点id":{"op":"gt|lt","value":数值}};命中即触发自动诊断(空=无规则)',
-    })
-    // 轮询器(15s)+ setup 重水化:恢复对 kv 中 running run 的跟踪(热重载安全)
-    const sweep = () => sweepOnce(ctx).catch(err => ctx.logger.warn(`诊断轮询异常(继续): ${err?.message ?? err}`))
-    ctx.timer.setInterval(sweep, 15000)
-    sweep()
-
-    // 自动诊断(默认关):daq:sample 命中规则 → 冷却 30min → 同管线发起(source:'auto')
-    ctx.hooks.on('daq:sample', (s) => {
-      try {
-        if (!autoEnabledOf(ctx)) return
-        if (!s || typeof s.value !== 'number' || !s.nodeId || !s.lineId) return
-        const rule = parseAutoRules(autoRulesOf(ctx))[s.nodeId]
-        if (!rule || typeof rule?.value !== 'number' || !['gt', 'lt'].includes(rule.op)) return
-        const hit = rule.op === 'lt' ? s.value < rule.value : s.value > rule.value
-        if (!hit) return
-        const line = s.lineId
-        const last = Number(ctx.kv.get(`cooldown:${line}`)) || 0
-        if (Date.now() - last < AUTO_COOLDOWN_MS) return
-        const now = Date.now()
-        ctx.kv.set(`cooldown:${line}`, now) // 先占冷却,防同窗连发
-        // 首拍竞态:开线后首个越限样本即刻触发,快照查询早于 tsdb 首次刷盘(TSDB_FLUSH_MS=500ms)
-        // → 时窗查空失败。空窗失败定向重试(≤3 次、间隔 5s,总窗 ≤20s);冷却不回滚,仍只此一次派发。
-        const question = `${line} 产线该时窗数据深度根因诊断(自动触发:节点 ${s.nodeId} 采样值 ${s.value} ${rule.op === 'lt' ? '<' : '>'} ${rule.value})`
-        const attempt = (n) => {
-          startDiagnosis(ctx, {
-            line,
-            fromMs: now - 60 * MIN,
-            toMs: now,
-            question,
-            scene: `${line}_auto`,
-            source: 'auto',
-          }).then((r) => {
-            if (r.ok) ctx.logger.warn(`自动诊断已发起:产线 ${line} runId=${r.runId}(节点 ${s.nodeId}=${s.value} 越限)`)
-            else if (n < 3 && String(r.error ?? '').includes('无数采样本')) setTimeout(() => attempt(n + 1), 5000)
-            else ctx.logger.warn(`自动诊断发起失败(产线 ${line}): ${r.error}`)
-          }).catch(() => {})
-        }
-        attempt(0)
-      }
-      catch { /* 钩子内绝不抛 */ }
-    })
-
-    // 工具 ×2
-    ctx.omp.registerTool({
-      name: 'diag_run',
-      label: '发起分析',
-      description: '发起 IDD 后台分析任务(提交即返 task_id,无需等待;完成后平台自动回执本频道,空闲时也可 diag_status 查询)。analysis_intent 选择分析模式,每种意图由 IDD 侧对应 skill 协议执行(harness 在 IDD 内,分析业务与平台解耦):deep-diagnosis=深度根因诊断(9 阶段管线,最重);stability-report=产线稳定报告(评分+等级+建议,零根因);quick-screen=快速筛查(越限/漂移/冻结异常清单+HTML 看板,最快);doe-analysis=DOE 试验分析(因子效应/显著性/最优工况);experience-mining=调参经验挖掘(结构化经验条目,供知识库沉淀);optimizer-campaign=优化战役(寻优迭代+推荐工况,最重)。数据来源三选一:① line+时间窗(导出产线 DAQ 快照);② data_path:诊断服务内已有数据文件;③ export_id(推荐):daq_export 导出的数据集。',
-      parameters: {
-        type: 'object',
-        properties: {
-          line: { type: 'string', description: '产线 id(与 data_path/export_id 三选一;export/data_path 模式可填 any)' },
-          export_id: { type: 'string', description: 'daq_export 返回的数据集 id(daqexp-…):上传 manifest+全部节点全量时序 CSV 做多文件分析' },
-          data_path: { type: 'string', description: '可选:诊断服务内数据文件相对路径(如 data/aw-snapshots/x.csv;跳过快照导出,适合离线数据分析)' },
-          from_ms: { type: 'number', description: '窗口起点(epoch 毫秒;仅 line 模式;缺省 now-60min)' },
-          to_ms: { type: 'number', description: '窗口终点(epoch 毫秒;仅 line 模式;缺省 now)' },
-          question: { type: 'string', description: '分析问题(缺省按意图自动生成情景化提示词;业务要求特殊时才覆写)' },
-          analysis_intent: {
-            type: 'string',
-            enum: ['deep-diagnosis', 'stability-report', 'quick-screen', 'doe-analysis', 'experience-mining', 'optimizer-campaign'],
-            description: '分析模式(缺省 deep-diagnosis)。按业务情景选择:要根因→deep-diagnosis;要运行健康度评分/等级→stability-report;要快速异常清单→quick-screen;要多因子试验数据→doe-analysis;要沉淀调参经验→experience-mining;要寻优推荐工况→optimizer-campaign',
-          },
-          scene: { type: 'string', description: '场景名(缺省 <line>_diag);场景为自由标签,分析管线按数据特征自适应' },
-          mode: { type: 'string', enum: ['async', 'sync'], description: 'async(缺省)=提交即返后台跑;sync=同步等待完成(最长 sync_timeout_s,超时自动转后台继续)' },
-          sync_timeout_s: { type: 'number', description: 'sync 模式等待上限秒(缺省 240,上限 540)' },
-          harness: { type: 'string', description: '分析引擎按次覆盖(omp/claude/mock;缺省用插件设置)' },
-          model: { type: 'string', description: '可选:引擎模型按次覆盖(合法性由 IDD 校验,不合法返回 400 说明)' },
-          permission_mode: { type: 'string', description: '可选:引擎权限模式按次覆盖(合法性由 IDD 校验)' },
-        },
-        required: [],
-      },
-      roles: ['lead', 'worker'],
-      handler: async (args, agent) => {
-        try {
-          const line = String(args.line ?? '').trim()
-          const dataPathDirect = String(args.data_path ?? '').trim()
-          const exportId = String(args.export_id ?? '').trim()
-          if (!line && !dataPathDirect && !exportId) return { text: 'line / data_path / export_id 至少给一个。', isError: true }
-          const toMs = Number(args.to_ms) > 0 ? Number(args.to_ms) : Date.now()
-          const fromMs = Number(args.from_ms) > 0 ? Number(args.from_ms) : toMs - 60 * MIN
-          if (fromMs >= toMs) return { text: 'from_ms 必须小于 to_ms。', isError: true }
-
-          // 数据集模式:上传 manifest + 全部节点 CSV(manifest 首位,IDD multi 模式)
-          let dataPaths
-          let via = ''
-          let manifest = null
-          if (exportId) {
-            const up = await uploadExportDataset(ctx, exportId)
-            if (!up.ok) return { text: up.error, isError: true }
-            dataPaths = up.dataPaths
-            manifest = up.manifest
-            const nodeCount = Array.isArray(manifest?.nodes) ? manifest.nodes.length : dataPaths.length - 1
-            const win = manifest?.window ? `窗口 ${String(manifest.window.from ?? '').slice(0, 16)} ~ ${String(manifest.window.to ?? '').slice(0, 16)}` : ''
-            via = `数据集 ${up.folder}(${nodeCount} 节点全量时序 + manifest,${win})`
-          }
-
-          // 问题缺省:按数据来源自动组题(export 模式带产线与场景备注)
-          let question = args.question ? String(args.question) : ''
-          if (!question) {
-            if (manifest) {
-              const lineNames = (Array.isArray(manifest.lines) ? manifest.lines : []).map(l => l?.lineName || l?.lineId).filter(Boolean).join('、') || line || 'offline'
-              const note = manifest.note ? `;场景备注:${String(manifest.note).slice(0, 200)}` : ''
-              question = `${lineNames} 产线数据集深度根因诊断(${via})${note}`
-            }
-            else {
-              question = `${line || 'offline-data'} 产线该时窗数据深度根因诊断`
-            }
-          }
-          const r = await startDiagnosis(ctx, {
-            line: line || 'offline-data',
-            fromMs, toMs,
-            question,
-            scene: args.scene ? String(args.scene) : '',
-            source: 'agent',
-            dataPath: dataPathDirect || undefined,
-            dataPaths,
-            harness: args.harness ? String(args.harness) : undefined,
-            analysisIntent: args.analysis_intent ? String(args.analysis_intent) : undefined,
-            model: args.model ? String(args.model) : undefined,
-            permissionMode: args.permission_mode ? String(args.permission_mode) : undefined,
-            // 归因(分发层注入的调用方上下文):诊断完成后平台据此把结果回执给发起 Agent
-            agentId: agent?.agentId ?? '',
-            channelId: agent?.channelId ?? '',
-          })
-          if (!r.ok) return { text: r.error, isError: true }
-          if (!via) via = dataPathDirect ? `离线数据 ${r.csvPath}` : `产线 ${line} 窗口 ${new Date(fromMs).toISOString()} ~ ${new Date(toMs).toISOString()} 快照 ${r.rows} 行 × ${r.nodes} 节点`
-
-          // 同步模式:轮询到终态或超时(超时自动转后台,sweep 完成仍会回执)
-          if (String(args.mode ?? '').toLowerCase() === 'sync') {
-            const base = baseOf(ctx)
-            const timeoutS = Math.min(Math.max(Number(args.sync_timeout_s) || 240, 10), 540)
-            if (base) {
-              const deadline = Date.now() + timeoutS * 1000
-              while (Date.now() < deadline) {
-                await new Promise(resolve => setTimeout(resolve, 5000))
-                let st = null
-                try {
-                  st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(r.runId)}`, 8000)).data ?? null
-                }
-                catch { /* 单次轮询失败继续等 */ }
-                const status = String(st?.status ?? 'running')
-                if (status === 'completed') {
-                  // 已同步拿到结果,认领回执防 sweep 补送重复
-                  try {
-                    const notify = await ctx.services.get('notify')
-                    await notify?.claim?.('diag_run', r.runId)
-                  }
-                  catch { /* 未接线时忽略 */ }
-                  const rep = st?.result?.report_md_path ?? '-'
-                  return { text: `✅ 同步诊断完成 task_id=${r.runId}(${via})\n评分=${st?.result?.score ?? '-'} 结论=${st?.result?.verdict ?? '-'}\nreport_md_path=${rep}\n总结:${String(st?.result?.summary ?? '').slice(0, 800)}\n\n入库知识库:调用 rag-bridge 的 kb_agent(mode=async,prompt=「请把这份深度诊断报告入库到 aw-industrial 知识库:读取文件 ${rep} 的全文内容并入库,完成后报告 doc_path 与分块数。标签:诊断、深度诊断。」),完成后 kb_agent_status(task_id) 查询。` }
-                }
-                if (status === 'failed' || status === 'stopped') {
-                  return { text: `分析任务 ${r.runId} 以 ${status} 结束:${st?.result?.error ?? '详见 diag_status'}`, isError: true }
-                }
-              }
-              return { text: `已发起${intentLabel(r.intent)}任务 task_id=${r.runId}(${via})。同步等待 ${timeoutS}s 超时,分析仍在后台运行——完成后平台会自动把结果回执到你的信箱并唤醒你;也可 diag_status(run_id=${r.runId}) 主动查询。注入的分析问题:${r.question}` }
-            }
-          }
-          return {
-            text: `已发起「${intentLabel(r.intent)}」异步分析任务 task_id=${r.runId}(${via})。IDD 在后台以对应 skill 协议执行,Channel 无需等待——可继续其他任务;完成后平台会自动把结果回执到你的信箱并唤醒你(空闲时也可用 diag_status(run_id=${r.runId}) 主动查询;完成后按指引调 kb_agent 把报告入库知识库)。注入的分析问题:${r.question}`,
-          }
-        }
-        catch (err) {
-          return { text: `diag_run 异常: ${err?.message ?? err}`, isError: true }
-        }
-      },
-    })
-
-    ctx.omp.registerTool({
-      name: 'diag_status',
-      label: '分析任务状态',
-      description: '查询后台分析任务进度:传 run_id 看单条实时状态(完成时含评分/结论/总结/报告路径),不传看最近 5 条概要。适用 diag_run 发起的全部意图(深度诊断/产线报告/快速筛查/DOE/经验挖掘/优化战役)。',
-      parameters: {
-        type: 'object',
-        properties: {
-          run_id: { type: 'string', description: '诊断 runId(缺省列最近 5 条概要)' },
-        },
-      },
-      roles: ['lead', 'worker'],
-      handler: async (args) => {
-        try {
-          const id = String(args.run_id ?? '').trim()
-          if (!id) {
-            const runs = kvRuns(ctx)
-              .sort((a, b) => (b.meta?.createdAt ?? 0) - (a.meta?.createdAt ?? 0))
-              .slice(0, 5)
-            if (!runs.length) return { text: '暂无诊断记录。用 diag_run 发起一次深度诊断。' }
-            const lines = runs.map(({ id: rid, meta }) =>
-              `- ${rid} | ${meta.line} | [${intentLabel(meta.intent ?? 'deep-diagnosis')}] ${meta.status}${meta.stored ? '(已入库)' : ''} | ${meta.source ?? 'agent'} | ${meta.createdAt ? new Date(meta.createdAt).toISOString() : '?'} | ${meta.question ?? ''}`)
-            return { text: `最近 ${runs.length} 条诊断:\n${lines.join('\n')}` }
-          }
-          const meta = ctx.kv.get(runKey(id))
-          const base = baseOf(ctx)
-          let st = null
-          if (base) {
-            try {
-              // v2.1:走任务管理面 /tasks/:id——result.report_md_path 是绝对路径,
-              // 可直接交给 kb_agent 读取入库(旧 /status 的相对路径会让 kb_agent 找不到文件)
-              st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(id)}`, 8000)).data ?? null
-              if (st && st.result) {
-                st = {
-                  ...st,
-                  engineStatus: st.status,
-                  score: st.result.score ?? null,
-                  judge_verdict: st.result.verdict ?? null,
-                  summary: st.result.summary ?? '',
-                  report_path: st.result.report_md_path ?? st.report_path ?? null,
-                  error_message: st.result.error ?? st.error_message ?? null,
-                }
-              }
-            }
-            catch { /* 3210 不可达时回退 kv 元数据 */ }
-          }
-          if (!st && !meta) return { text: `诊断 ${id} 不存在或诊断服务不可达。`, isError: true }
-          const status = st?.engineStatus || st?.status || meta?.status || 'unknown'
-          // 终态经轮询面亲自看到 → 认领结果所有权,取消平台的事后补送(先到先得)
-          if (['completed', 'failed', 'stopped'].includes(status)) {
-            const notify = await ctx.services.get('notify').catch(() => null)
-            notify?.claim?.('diag_run', id)
-          }
-          const parts = [
-            `run_id=${id}`,
-            `name=${st?.name ?? meta?.name ?? '?'}`,
-            `产线=${meta?.line ?? st?.scene_name ?? '?'}`,
-            `意图=${meta?.intent ?? 'deep-diagnosis'}(${intentLabel(meta?.intent ?? 'deep-diagnosis')})`,
-            `状态=${status}`,
-            `来源=${meta?.source ?? '?'}`,
-          ]
-          if (meta?.fromMs && meta?.toMs) parts.push(`窗口=${new Date(meta.fromMs).toISOString()} ~ ${new Date(meta.toMs).toISOString()}`)
-          if (meta?.question) parts.push(`问题=${meta.question}`)
-          if (status === 'completed') {
-            const reportPath = st?.report_path ?? meta?.reportPath ?? '-'
-            parts.push(`评分=${st?.score ?? meta?.score ?? '-'}`)
-            parts.push(`结论=${st?.judge_verdict ?? meta?.verdict ?? '-'}`)
-            if (st?.summary) parts.push(`总结=${String(st.summary).slice(0, 600)}`)
-            parts.push(`report_md_path=${reportPath}`)
-            parts.push('')
-            parts.push('✅ 分析完成,报告已就绪。入库知识库请调用 rag-bridge 的 kb_agent 工具:')
-            parts.push(`  mode=async, prompt=「请把这份分析报告入库到 aw-industrial 知识库:读取文件 ${reportPath} 的全文内容并入库(写盘+索引),完成后报告 doc_path 与分块数。标签:分析、${intentLabel(meta?.intent ?? 'deep-diagnosis')}。」`)
-            parts.push('提交后频道无需等待,空闲时用 kb_agent_status(task_id) 查询入库结果。')
-          }
-          if (status === 'failed' || status === 'stopped') parts.push(`错误=${st?.error_message ?? meta?.error ?? '-'}`)
-          if (!st) parts.push('(诊断服务不可达,以上为本地记录)')
-          return { text: parts.join('\n') }
-        }
-        catch (err) {
-          return { text: `diag_status 异常: ${err?.message ?? err}`, isError: true }
-        }
-      },
-    })
-
-    // 路由 ×3(挂 /api/plugins/diag-bridge/*)
-    ctx.route('GET', '/health', async () => {
-      const runs = kvRuns(ctx)
-      const byStatus = {}
-      for (const { meta } of runs) byStatus[meta?.status ?? 'unknown'] = (byStatus[meta?.status ?? 'unknown'] ?? 0) + 1
-      const base = baseOf(ctx)
-      let remote
-      let tokenState = { configured: Boolean(diagTokenOf(ctx)), accepted: null }
-      if (base) {
-        try {
-          // 连通性:/api/health 是公开端点(无鉴权),只回答「服务在不在」。
-          const res = await ctx.http.get(`${base}/api/health`, { timeoutMs: 5000 })
-          const body = await res.json().catch(() => null)
-          remote = res.ok && body
-            ? { status: body.status ?? 'ok', activeRuns: body.checks?.activeRuns ?? null }
-            : { status: `HTTP ${res.status}`, activeRuns: null }
-          // Token 有效性:/api/health 不校验鉴权,必须另打一个受保护端点,
-          // 否则 token 配错时健康面依旧 ok —— 配置错误完全不可见(实测踩过)。
-          if (tokenState.configured) {
-            try {
-              const auth = await ctx.http.get(`${base}/api/files/workspace`, {
-                timeoutMs: 5000,
-                headers: authHeadersOf(ctx),
-              })
-              if (auth.status === 401 || auth.status === 403) {
-                tokenState = { configured: true, accepted: false, hint: authHint(auth.status, 'Token 被上游拒绝') }
-                remote = { ...remote, auth: `HTTP ${auth.status}(鉴权被拒)`, hint: tokenState.hint }
-              }
-              else if (auth.ok) {
-                tokenState = { configured: true, accepted: true }
-                remote = { ...remote, auth: 'ok' }
-              }
-              else {
-                remote = { ...remote, auth: `HTTP ${auth.status}` }
-              }
-            }
-            catch (err) {
-              remote = { ...remote, auth: 'unreachable', authError: String(err?.message ?? err) }
-            }
-          }
-          else {
-            tokenState = { configured: false, accepted: null }
-          }
-        }
-        catch (err) {
-          remote = { status: 'unreachable', error: String(err?.message ?? err) }
-        }
-      }
-      else {
-        remote = { status: 'bad_base_url' }
-      }
-      return {
-        plugin: ctx.name,
-        base: base ?? String(ctx.kv.get('diag.base_url') || DEFAULT_BASE),
-        harness: harnessOf(ctx),
-        auth: diagTokenOf(ctx) ? 'bearer' : 'anonymous',
-        /** Token 配置态 + 实测是否被上游接受(401/403 → false;网络不通 → null) */
-        token: tokenState,
-        remote,
-        runs: { total: runs.length, byStatus },
-      }
-    })
-
-    ctx.route('GET', '/runs', () => ({
-      success: true,
-      runs: kvRuns(ctx)
-        .sort((a, b) => (b.meta?.createdAt ?? 0) - (a.meta?.createdAt ?? 0))
-        .map(({ id, meta }) => ({ runId: id, ...meta })),
-    }))
-
-    ctx.route('POST', '/snapshot', async (event) => {
-      try {
-        const b = event?.awBody ?? {}
-        const line = String(b.line ?? '').trim()
-        if (!line) return { success: false, error: 'line 必填(产线 id)' }
-        const toMs = Number(b.to_ms) > 0 ? Number(b.to_ms) : Date.now()
-        const fromMs = Number(b.from_ms) > 0 ? Number(b.from_ms) : toMs - 60 * MIN
-        if (fromMs >= toMs) return { success: false, error: 'from_ms 必须小于 to_ms' }
-        const snap = await snapshotCore(ctx, line, fromMs, toMs)
-        if (!snap.ok) return { success: false, error: snap.error }
-        return { success: true, csvPath: snap.csvPath, rows: snap.rows, nodes: snap.nodes, fromMs: snap.fromMs, toMs: snap.toMs }
-      }
-      catch (err) {
-        return { success: false, error: String(err?.message ?? err) }
-      }
-    })
-
-    // 报告查看(面板/前端):run_id → 诊断任务 → workspace run 目录 → report md 内容
-    ctx.route('GET', '/report', async (event) => {
-      try {
-        const u = new URL(String(event?.path ?? event?.node?.req?.url ?? ''), 'http://local')
-        const runId = String(u.searchParams.get('run_id') ?? '').trim()
-        if (!runId) return { success: false, error: 'run_id 必填' }
-        const base = baseOf(ctx)
-        if (!base) return { success: false, error: 'diag.base_url 非法' }
-        const st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(runId)}`, 8000)).data ?? {}
-        if (String(st.status ?? '') !== 'completed') {
-          return { success: false, error: `诊断 ${runId} 状态 ${st.status ?? 'unknown'},报告未就绪` }
-        }
-        const ws = (await jget(ctx, `${base}/api/files/workspace`, 8000)).data ?? []
-        const names = (Array.isArray(ws) ? ws : [])
-        const meta = ctx.kv.get(runKey(runId)) ?? {}
-        // 目录命名不统一(<ts>_<scene> 或 <ts>_<scene>_<runId>):runId → scene → 最新 逐级兜底
-        const hit = names.find(e => String(e?.name ?? '').includes(runId))
-          ?? names.find(e => meta?.scene && String(e?.name ?? '').toLowerCase().includes(String(meta.scene).toLowerCase()))
-          ?? (names.length ? names.slice().sort((a, b) => Number(b?.created ?? 0) - Number(a?.created ?? 0))[0] : null)
-        if (!hit) return { success: false, error: `诊断服务 workspace 未找到 ${runId} 的运行目录` }
-        const rep = (await jget(ctx, `${base}/api/files/workspace/report/${encodeURIComponent(hit.name)}`, 15000)).data
-        return { success: true, run_id: runId, run_name: hit.name, report: rep ?? null }
-      }
-      catch (err) {
-        return { success: false, error: String(err?.message ?? err) }
-      }
-    })
-
-    ctx.logger.info('diag-bridge 就绪:工具 diag_run/diag_status + 路由 health/runs/snapshot/report,轮询 15s 已启动')
-  },
-}
+/**
+ * 插件对象(settings / routes / tools / hooks)
+ * (由 server/plugins-builtin/diag-bridge/index.mjs 按职责拆出;内容逐行原文搬运)
+ */
+import { AUTO_COOLDOWN_MS, DEFAULT_BASE, MIN } from './constants.mjs'
+import { authHeadersOf, authHint, autoEnabledOf, autoRulesOf, baseOf, diagTokenOf, harnessOf, jget, kvRuns, parseAutoRules, runKey } from './helpers.mjs'
+import { snapshotCore } from './snapshot.mjs'
+import { startDiagnosis } from './diagnosis.mjs'
+import { uploadExportDataset } from './exports.mjs'
+import { sweepOnce } from './sweep.mjs'
+import { intentLabel } from './intents.mjs'
+
+export default {
+  name: 'diag-bridge',
+  version: '2.2.0',
+  description: '深度诊断桥接 v2.2:异步任务式深度诊断(提交即返 task_id)+ daq-export 数据集多文件分析(diag_run export_id= 全量时序+manifest)+ 同步等待模式(mode=sync)+ 完成自动回执 Channel(修复发起者身份丢失)+ 报告在线查看。完成后经 rag-bridge kb_agent 入库知识库。',
+  auth: 'user',
+  client: './client.mjs', // 前端面板(插件页注入;i18n 见 i18n.json)
+  // 插件配置分组(声明式):连接 / 执行 / 自动化 三个独立分区
+  configGroups: [
+    { id: 'default', labelKey: 'plugin.diag-bridge.group.conn', label: '诊断服务连接', description: '后端地址与鉴权 Token', order: 420 },
+    { id: 'run', labelKey: 'plugin.diag-bridge.group.run', label: '诊断执行参数', description: 'diag_run 的引擎与预算', order: 430 },
+    { id: 'auto', labelKey: 'plugin.diag-bridge.group.auto', label: '自动诊断', description: '越限自动发起诊断(DAQ 采样事件驱动)', order: 440, collapsed: true },
+  ],
+  // 插件设置声明(key 编址 plugins.diag-bridge.<key>;labelKey 解析 i18n.json)
+  settings: [
+    { key: 'base_url', type: 'string', default: DEFAULT_BASE, group: 'default', labelKey: 'plugin.diag-bridge.settings.base_url', label: '诊断服务地址', description: 'industrial-deep-diagnostic 后端(http/https,host 限 127.0.0.1/localhost);保存即热生效' },
+    { key: 'token', type: 'string', default: '', group: 'default', labelKey: 'plugin.diag-bridge.settings.token', label: '诊断 API Token', description: '诊断服务开启鉴权时(v4 起默认强制)的 API Token(Authorization: Bearer);空=匿名;保存即热生效' },
+    { key: 'harness', type: 'string', default: '', group: 'run', labelKey: 'plugin.diag-bridge.settings.harness', label: '诊断引擎', description: 'diag_run 用的引擎 id(omp/claude/mock 等;空=取 kv/omp);无 Anthropic key 的机器请用 omp' },
+    { key: 'max_turns', type: 'number', default: 0, min: 0, max: 2000, group: 'run', labelKey: 'plugin.diag-bridge.settings.max_turns', label: '诊断最大轮数', description: 'diagnosis/start 的 maxTurns(0=取 kv/内置默认 220)' },
+    { key: 'max_minutes', type: 'number', default: 0, min: 0, max: 720, group: 'run', labelKey: 'plugin.diag-bridge.settings.max_minutes', label: '诊断超时(分钟)', description: 'diagnosis/start 的 timeoutMinutes(0=取 kv/内置默认 40)' },
+    { key: 'auto_enabled', type: 'boolean', default: false, group: 'auto', labelKey: 'plugin.diag-bridge.settings.auto_enabled', label: '自动诊断', description: 'daq:sample 越限命中规则时自动发起深度诊断(每产线冷却 30 分钟)' },
+  ],
+  setup(ctx) {
+    // 运行时追加分区与字段(ctx.config.defineGroup / defineField):
+    // 字段 key 与声明式完全等价(plugins.diag-bridge.auto_rules),已保存的值不丢;
+    // 这里演示「插件在 setup 里按条件调用平台 API 扩展自己的配置面」。
+    ctx.config.defineGroup({
+      id: 'rules',
+      labelKey: 'plugin.diag-bridge.group.rules',
+      label: '自动诊断规则',
+      description: 'JSON 规则;命中即触发自动诊断(仅 auto_enabled=true 时生效)',
+      order: 450,
+    })
+    ctx.config.defineField({
+      key: 'auto_rules',
+      type: 'string',
+      default: '',
+      group: 'rules',
+      labelKey: 'plugin.diag-bridge.settings.auto_rules',
+      label: '自动诊断规则',
+      description: 'JSON: {"节点id":{"op":"gt|lt","value":数值}};命中即触发自动诊断(空=无规则)',
+    })
+    // 轮询器(15s)+ setup 重水化:恢复对 kv 中 running run 的跟踪(热重载安全)
+    const sweep = () => sweepOnce(ctx).catch(err => ctx.logger.warn(`诊断轮询异常(继续): ${err?.message ?? err}`))
+    ctx.timer.setInterval(sweep, 15000)
+    sweep()
+
+    // 自动诊断(默认关):daq:sample 命中规则 → 冷却 30min → 同管线发起(source:'auto')
+    ctx.hooks.on('daq:sample', (s) => {
+      try {
+        if (!autoEnabledOf(ctx)) return
+        if (!s || typeof s.value !== 'number' || !s.nodeId || !s.lineId) return
+        const rule = parseAutoRules(autoRulesOf(ctx))[s.nodeId]
+        if (!rule || typeof rule?.value !== 'number' || !['gt', 'lt'].includes(rule.op)) return
+        const hit = rule.op === 'lt' ? s.value < rule.value : s.value > rule.value
+        if (!hit) return
+        const line = s.lineId
+        const last = Number(ctx.kv.get(`cooldown:${line}`)) || 0
+        if (Date.now() - last < AUTO_COOLDOWN_MS) return
+        const now = Date.now()
+        ctx.kv.set(`cooldown:${line}`, now) // 先占冷却,防同窗连发
+        // 首拍竞态:开线后首个越限样本即刻触发,快照查询早于 tsdb 首次刷盘(TSDB_FLUSH_MS=500ms)
+        // → 时窗查空失败。空窗失败定向重试(≤3 次、间隔 5s,总窗 ≤20s);冷却不回滚,仍只此一次派发。
+        const question = `${line} 产线该时窗数据深度根因诊断(自动触发:节点 ${s.nodeId} 采样值 ${s.value} ${rule.op === 'lt' ? '<' : '>'} ${rule.value})`
+        const attempt = (n) => {
+          startDiagnosis(ctx, {
+            line,
+            fromMs: now - 60 * MIN,
+            toMs: now,
+            question,
+            scene: `${line}_auto`,
+            source: 'auto',
+          }).then((r) => {
+            if (r.ok) ctx.logger.warn(`自动诊断已发起:产线 ${line} runId=${r.runId}(节点 ${s.nodeId}=${s.value} 越限)`)
+            else if (n < 3 && String(r.error ?? '').includes('无数采样本')) setTimeout(() => attempt(n + 1), 5000)
+            else ctx.logger.warn(`自动诊断发起失败(产线 ${line}): ${r.error}`)
+          }).catch(() => {})
+        }
+        attempt(0)
+      }
+      catch { /* 钩子内绝不抛 */ }
+    })
+
+    // 工具 ×2
+    ctx.omp.registerTool({
+      name: 'diag_run',
+      label: '发起分析',
+      description: '发起 IDD 后台分析任务(提交即返 task_id,无需等待;完成后平台自动回执本频道,空闲时也可 diag_status 查询)。analysis_intent 选择分析模式,每种意图由 IDD 侧对应 skill 协议执行(harness 在 IDD 内,分析业务与平台解耦):deep-diagnosis=深度根因诊断(9 阶段管线,最重);stability-report=产线稳定报告(评分+等级+建议,零根因);quick-screen=快速筛查(越限/漂移/冻结异常清单+HTML 看板,最快);doe-analysis=DOE 试验分析(因子效应/显著性/最优工况);experience-mining=调参经验挖掘(结构化经验条目,供知识库沉淀);optimizer-campaign=优化战役(寻优迭代+推荐工况,最重)。数据来源三选一:① line+时间窗(导出产线 DAQ 快照);② data_path:诊断服务内已有数据文件;③ export_id(推荐):daq_export 导出的数据集。',
+      parameters: {
+        type: 'object',
+        properties: {
+          line: { type: 'string', description: '产线 id(与 data_path/export_id 三选一;export/data_path 模式可填 any)' },
+          export_id: { type: 'string', description: 'daq_export 返回的数据集 id(daqexp-…):上传 manifest+全部节点全量时序 CSV 做多文件分析' },
+          data_path: { type: 'string', description: '可选:诊断服务内数据文件相对路径(如 data/aw-snapshots/x.csv;跳过快照导出,适合离线数据分析)' },
+          from_ms: { type: 'number', description: '窗口起点(epoch 毫秒;仅 line 模式;缺省 now-60min)' },
+          to_ms: { type: 'number', description: '窗口终点(epoch 毫秒;仅 line 模式;缺省 now)' },
+          question: { type: 'string', description: '分析问题(缺省按意图自动生成情景化提示词;业务要求特殊时才覆写)' },
+          analysis_intent: {
+            type: 'string',
+            enum: ['deep-diagnosis', 'stability-report', 'quick-screen', 'doe-analysis', 'experience-mining', 'optimizer-campaign'],
+            description: '分析模式(缺省 deep-diagnosis)。按业务情景选择:要根因→deep-diagnosis;要运行健康度评分/等级→stability-report;要快速异常清单→quick-screen;要多因子试验数据→doe-analysis;要沉淀调参经验→experience-mining;要寻优推荐工况→optimizer-campaign',
+          },
+          scene: { type: 'string', description: '场景名(缺省 <line>_diag);场景为自由标签,分析管线按数据特征自适应' },
+          mode: { type: 'string', enum: ['async', 'sync'], description: 'async(缺省)=提交即返后台跑;sync=同步等待完成(最长 sync_timeout_s,超时自动转后台继续)' },
+          sync_timeout_s: { type: 'number', description: 'sync 模式等待上限秒(缺省 240,上限 540)' },
+          harness: { type: 'string', description: '分析引擎按次覆盖(omp/claude/mock;缺省用插件设置)' },
+          model: { type: 'string', description: '可选:引擎模型按次覆盖(合法性由 IDD 校验,不合法返回 400 说明)' },
+          permission_mode: { type: 'string', description: '可选:引擎权限模式按次覆盖(合法性由 IDD 校验)' },
+        },
+        required: [],
+      },
+      roles: ['lead', 'worker'],
+      handler: async (args, agent) => {
+        try {
+          const line = String(args.line ?? '').trim()
+          const dataPathDirect = String(args.data_path ?? '').trim()
+          const exportId = String(args.export_id ?? '').trim()
+          if (!line && !dataPathDirect && !exportId) return { text: 'line / data_path / export_id 至少给一个。', isError: true }
+          // 产线模式按调用者可读面授权(插件工具不走工业工具 grant/verify 两道闸;
+          // 此前任意 Agent 可对任意产线导出快照做诊断,构成跨线数据读取)。
+          if (line && agent?.agentId) {
+            try {
+              const scope = await ctx.services.get('agentScope')
+              const readable = scope.readableLines(agent.agentId)
+              if (!readable.includes(line)) {
+                return {
+                  text: `产线「${line}」不在你的可读范围。你的可读产线:${readable.length ? readable.join(', ') : '(无)'}。如需诊断该产线,请让用户绑定该产线节点/把频道绑定到该产线,或改用上述产线重试。`,
+                  isError: true,
+                }
+              }
+            }
+            catch (err) {
+              // fail-closed:无法确认授权时不读取产线数据
+              return { text: `产线授权校验不可用,已拒绝本次分析:${err?.message ?? err}。请联系管理员检查插件运行时服务。`, isError: true }
+            }
+          }
+          const toMs = Number(args.to_ms) > 0 ? Number(args.to_ms) : Date.now()
+          const fromMs = Number(args.from_ms) > 0 ? Number(args.from_ms) : toMs - 60 * MIN
+          if (fromMs >= toMs) return { text: 'from_ms 必须小于 to_ms。', isError: true }
+
+          // 数据集模式:上传 manifest + 全部节点 CSV(manifest 首位,IDD multi 模式)
+          let dataPaths
+          let via = ''
+          let manifest = null
+          if (exportId) {
+            const up = await uploadExportDataset(ctx, exportId)
+            if (!up.ok) return { text: up.error, isError: true }
+            dataPaths = up.dataPaths
+            manifest = up.manifest
+            const nodeCount = Array.isArray(manifest?.nodes) ? manifest.nodes.length : dataPaths.length - 1
+            const win = manifest?.window ? `窗口 ${String(manifest.window.from ?? '').slice(0, 16)} ~ ${String(manifest.window.to ?? '').slice(0, 16)}` : ''
+            via = `数据集 ${up.folder}(${nodeCount} 节点全量时序 + manifest,${win})`
+          }
+
+          // 问题缺省:按数据来源自动组题(export 模式带产线与场景备注)
+          let question = args.question ? String(args.question) : ''
+          if (!question) {
+            if (manifest) {
+              const lineNames = (Array.isArray(manifest.lines) ? manifest.lines : []).map(l => l?.lineName || l?.lineId).filter(Boolean).join('、') || line || 'offline'
+              const note = manifest.note ? `;场景备注:${String(manifest.note).slice(0, 200)}` : ''
+              question = `${lineNames} 产线数据集深度根因诊断(${via})${note}`
+            }
+            else {
+              question = `${line || 'offline-data'} 产线该时窗数据深度根因诊断`
+            }
+          }
+          const r = await startDiagnosis(ctx, {
+            line: line || 'offline-data',
+            fromMs, toMs,
+            question,
+            scene: args.scene ? String(args.scene) : '',
+            source: 'agent',
+            dataPath: dataPathDirect || undefined,
+            dataPaths,
+            harness: args.harness ? String(args.harness) : undefined,
+            analysisIntent: args.analysis_intent ? String(args.analysis_intent) : undefined,
+            model: args.model ? String(args.model) : undefined,
+            permissionMode: args.permission_mode ? String(args.permission_mode) : undefined,
+            // 归因(分发层注入的调用方上下文):诊断完成后平台据此把结果回执给发起 Agent
+            agentId: agent?.agentId ?? '',
+            channelId: agent?.channelId ?? '',
+          })
+          if (!r.ok) return { text: r.error, isError: true }
+          if (!via) via = dataPathDirect ? `离线数据 ${r.csvPath}` : `产线 ${line} 窗口 ${new Date(fromMs).toISOString()} ~ ${new Date(toMs).toISOString()} 快照 ${r.rows} 行 × ${r.nodes} 节点`
+
+          // 同步模式:轮询到终态或超时(超时自动转后台,sweep 完成仍会回执)
+          if (String(args.mode ?? '').toLowerCase() === 'sync') {
+            const base = baseOf(ctx)
+            const timeoutS = Math.min(Math.max(Number(args.sync_timeout_s) || 240, 10), 540)
+            if (base) {
+              const deadline = Date.now() + timeoutS * 1000
+              let lastPollError = ''
+              while (Date.now() < deadline) {
+                await new Promise(resolve => setTimeout(resolve, 5000))
+                let st = null
+                try {
+                  st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(r.runId)}`, 8000)).data ?? null
+                  lastPollError = ''
+                }
+                catch (err) {
+                  lastPollError = String(err?.message ?? err)
+                  // 鉴权失败/任务不存在:继续等只会空耗到超时,直接如实返回
+                  if (/HTTP (401|403|404)/.test(lastPollError)) {
+                    return { text: `同步等待中止 task_id=${r.runId}:${lastPollError}。任务可能仍在后台运行;修复后可用 diag_status(run_id=${r.runId}) 查询,完成后平台仍会回执。`, isError: true }
+                  }
+                }
+                const status = String(st?.status ?? r.status ?? 'running')
+                if (status === 'completed') {
+                  // 已同步拿到结果,认领回执防 sweep 补送重复
+                  try {
+                    const notify = await ctx.services.get('notify')
+                    await notify?.claim?.('diag_run', r.runId)
+                  }
+                  catch { /* 未接线时忽略 */ }
+                  const rep = st?.result?.report_md_path ?? '-'
+                  return { text: `✅ 同步诊断完成 task_id=${r.runId}(${via})\n评分=${st?.result?.score ?? '-'} 结论=${st?.result?.verdict ?? '-'}\nreport_md_path=${rep}\n总结:${String(st?.result?.summary ?? '').slice(0, 800)}\n\n入库知识库:调用 rag-bridge 的 kb_agent(mode=async,prompt=「请把这份深度诊断报告入库到 aw-industrial 知识库:读取文件 ${rep} 的全文内容并入库,完成后报告 doc_path 与分块数。标签:诊断、深度诊断。」),完成后 kb_agent_status(task_id) 查询。` }
+                }
+                if (status === 'failed' || status === 'stopped') {
+                  return { text: `分析任务 ${r.runId} 以 ${status} 结束:${st?.result?.error ?? '详见 diag_status'}`, isError: true }
+                }
+              }
+              return { text: `已发起${intentLabel(r.intent)}任务 task_id=${r.runId}(${via})。同步等待 ${timeoutS}s 超时,分析仍在后台运行${lastPollError ? `(最近一次轮询失败:${lastPollError})` : ''}——完成后平台会自动把结果回执到你的信箱并唤醒你;也可 diag_status(run_id=${r.runId}) 主动查询。注入的分析问题:${r.question}` }
+            }
+          }
+          return {
+            text: `已发起「${intentLabel(r.intent)}」异步分析任务 task_id=${r.runId}(${via})。IDD 在后台以对应 skill 协议执行,Channel 无需等待——可继续其他任务;完成后平台会自动把结果回执到你的信箱并唤醒你(空闲时也可用 diag_status(run_id=${r.runId}) 主动查询;完成后按指引调 kb_agent 把报告入库知识库)。注入的分析问题:${r.question}`,
+          }
+        }
+        catch (err) {
+          return { text: `diag_run 异常: ${err?.message ?? err}`, isError: true }
+        }
+      },
+    })
+
+    ctx.omp.registerTool({
+      name: 'diag_status',
+      label: '分析任务状态',
+      description: '查询后台分析任务进度:传 run_id 看单条实时状态(完成时含评分/结论/总结/报告路径),不传看最近 5 条概要。适用 diag_run 发起的全部意图(深度诊断/产线报告/快速筛查/DOE/经验挖掘/优化战役)。',
+      parameters: {
+        type: 'object',
+        properties: {
+          run_id: { type: 'string', description: '诊断 runId(缺省列最近 5 条概要)' },
+        },
+      },
+      roles: ['lead', 'worker'],
+      handler: async (args) => {
+        try {
+          const id = String(args.run_id ?? '').trim()
+          if (!id) {
+            const runs = kvRuns(ctx)
+              .sort((a, b) => (b.meta?.createdAt ?? 0) - (a.meta?.createdAt ?? 0))
+              .slice(0, 5)
+            if (!runs.length) return { text: '暂无诊断记录。用 diag_run 发起一次深度诊断。' }
+            const lines = runs.map(({ id: rid, meta }) =>
+              `- ${rid} | ${meta.line} | [${intentLabel(meta.intent ?? 'deep-diagnosis')}] ${meta.status}${meta.stored ? '(已入库)' : ''} | ${meta.source ?? 'agent'} | ${meta.createdAt ? new Date(meta.createdAt).toISOString() : '?'} | ${meta.question ?? ''}`)
+            return { text: `最近 ${runs.length} 条诊断:\n${lines.join('\n')}` }
+          }
+          const meta = ctx.kv.get(runKey(id))
+          const base = baseOf(ctx)
+          let st = null
+          let remoteError = ''
+          if (base) {
+            try {
+              // v2.1:走任务管理面 /tasks/:id——result.report_md_path 是绝对路径,
+              // 可直接交给 kb_agent 读取入库(旧 /status 的相对路径会让 kb_agent 找不到文件)
+              st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(id)}`, 8000)).data ?? null
+              if (st && st.result) {
+                st = {
+                  ...st,
+                  engineStatus: st.status,
+                  score: st.result.score ?? null,
+                  judge_verdict: st.result.verdict ?? null,
+                  summary: st.result.summary ?? '',
+                  report_path: st.result.report_md_path ?? st.report_path ?? null,
+                  error_message: st.result.error ?? st.error_message ?? null,
+                }
+              }
+            }
+            catch (err) {
+              // 保留原始错误(404=任务不存在 / 401=token / 网络不可达),不再一律吞成"不可达"
+              remoteError = String(err?.message ?? err)
+            }
+          }
+          if (!st && !meta) {
+            return { text: `诊断 ${id} 不存在或诊断服务不可达${remoteError ? `(${remoteError})` : ''}。`, isError: true }
+          }
+          const status = st?.engineStatus || st?.status || meta?.status || 'unknown'
+          // 终态经轮询面亲自看到 → 认领结果所有权,取消平台的事后补送(先到先得)
+          if (['completed', 'failed', 'stopped'].includes(status)) {
+            const notify = await ctx.services.get('notify').catch(() => null)
+            notify?.claim?.('diag_run', id)
+          }
+          const parts = [
+            `run_id=${id}`,
+            `name=${st?.name ?? meta?.name ?? '?'}`,
+            `产线=${meta?.line ?? st?.scene_name ?? '?'}`,
+            `意图=${meta?.intent ?? 'deep-diagnosis'}(${intentLabel(meta?.intent ?? 'deep-diagnosis')})`,
+            `状态=${status}`,
+            `来源=${meta?.source ?? '?'}`,
+          ]
+          if (meta?.fromMs && meta?.toMs) parts.push(`窗口=${new Date(meta.fromMs).toISOString()} ~ ${new Date(meta.toMs).toISOString()}`)
+          if (meta?.question) parts.push(`问题=${meta.question}`)
+          if (status === 'completed') {
+            const reportPath = st?.report_path ?? meta?.reportPath ?? '-'
+            parts.push(`评分=${st?.score ?? meta?.score ?? '-'}`)
+            parts.push(`结论=${st?.judge_verdict ?? meta?.verdict ?? '-'}`)
+            if (st?.summary) parts.push(`总结=${String(st.summary).slice(0, 600)}`)
+            parts.push(`report_md_path=${reportPath}`)
+            parts.push('')
+            parts.push('✅ 分析完成,报告已就绪。入库知识库请调用 rag-bridge 的 kb_agent 工具:')
+            parts.push(`  mode=async, prompt=「请把这份分析报告入库到 aw-industrial 知识库:读取文件 ${reportPath} 的全文内容并入库(写盘+索引),完成后报告 doc_path 与分块数。标签:分析、${intentLabel(meta?.intent ?? 'deep-diagnosis')}。」`)
+            parts.push('提交后频道无需等待,空闲时用 kb_agent_status(task_id) 查询入库结果。')
+          }
+          if (status === 'failed' || status === 'stopped') parts.push(`错误=${st?.error_message ?? meta?.error ?? '-'}`)
+          if (!st) {
+            parts.push(`(远端查询失败:${remoteError || '诊断服务不可达'};以上为本地记录)`)
+            if (remoteError.includes('HTTP 404') && ['queued', 'running'].includes(String(meta?.status))) {
+              parts.push('远端任务已不存在(可能被 IDD 清理或服务重启);sweep 将在连续 3 轮确认后把本地状态收敛为 failed 并回执,不会永久占用产线单飞锁。')
+            }
+          }
+          return { text: parts.join('\n') }
+        }
+        catch (err) {
+          return { text: `diag_status 异常: ${err?.message ?? err}`, isError: true }
+        }
+      },
+    })
+
+    // 路由 ×3(挂 /api/plugins/diag-bridge/*)
+    ctx.route('GET', '/health', async () => {
+      const runs = kvRuns(ctx)
+      const byStatus = {}
+      for (const { meta } of runs) byStatus[meta?.status ?? 'unknown'] = (byStatus[meta?.status ?? 'unknown'] ?? 0) + 1
+      const base = baseOf(ctx)
+      let remote
+      let tokenState = { configured: Boolean(diagTokenOf(ctx)), accepted: null }
+      if (base) {
+        try {
+          // 连通性:/api/health 是公开端点(无鉴权),只回答「服务在不在」。
+          const res = await ctx.http.get(`${base}/api/health`, { timeoutMs: 5000 })
+          const body = await res.json().catch(() => null)
+          remote = res.ok && body
+            ? { status: body.status ?? 'ok', activeRuns: body.checks?.activeRuns ?? null }
+            : { status: `HTTP ${res.status}`, activeRuns: null }
+          // Token 有效性:/api/health 不校验鉴权,必须另打一个受保护端点,
+          // 否则 token 配错时健康面依旧 ok —— 配置错误完全不可见(实测踩过)。
+          if (tokenState.configured) {
+            try {
+              const auth = await ctx.http.get(`${base}/api/files/workspace`, {
+                timeoutMs: 5000,
+                headers: authHeadersOf(ctx),
+              })
+              if (auth.status === 401 || auth.status === 403) {
+                tokenState = { configured: true, accepted: false, hint: authHint(auth.status, 'Token 被上游拒绝') }
+                remote = { ...remote, auth: `HTTP ${auth.status}(鉴权被拒)`, hint: tokenState.hint }
+              }
+              else if (auth.ok) {
+                tokenState = { configured: true, accepted: true }
+                remote = { ...remote, auth: 'ok' }
+              }
+              else {
+                remote = { ...remote, auth: `HTTP ${auth.status}` }
+              }
+            }
+            catch (err) {
+              remote = { ...remote, auth: 'unreachable', authError: String(err?.message ?? err) }
+            }
+          }
+          else {
+            tokenState = { configured: false, accepted: null }
+          }
+        }
+        catch (err) {
+          remote = { status: 'unreachable', error: String(err?.message ?? err) }
+        }
+      }
+      else {
+        remote = { status: 'bad_base_url' }
+      }
+      return {
+        plugin: ctx.name,
+        base: base ?? String(ctx.kv.get('diag.base_url') || DEFAULT_BASE),
+        harness: harnessOf(ctx),
+        auth: diagTokenOf(ctx) ? 'bearer' : 'anonymous',
+        /** Token 配置态 + 实测是否被上游接受(401/403 → false;网络不通 → null) */
+        token: tokenState,
+        remote,
+        runs: { total: runs.length, byStatus },
+      }
+    })
+
+    ctx.route('GET', '/runs', () => ({
+      success: true,
+      runs: kvRuns(ctx)
+        .sort((a, b) => (b.meta?.createdAt ?? 0) - (a.meta?.createdAt ?? 0))
+        .map(({ id, meta }) => ({ runId: id, ...meta })),
+    }))
+
+    ctx.route('POST', '/snapshot', async (event) => {
+      try {
+        const b = event?.awBody ?? {}
+        const line = String(b.line ?? '').trim()
+        if (!line) return { success: false, error: 'line 必填(产线 id)' }
+        const toMs = Number(b.to_ms) > 0 ? Number(b.to_ms) : Date.now()
+        const fromMs = Number(b.from_ms) > 0 ? Number(b.from_ms) : toMs - 60 * MIN
+        if (fromMs >= toMs) return { success: false, error: 'from_ms 必须小于 to_ms' }
+        const snap = await snapshotCore(ctx, line, fromMs, toMs)
+        if (!snap.ok) return { success: false, error: snap.error }
+        return { success: true, csvPath: snap.csvPath, rows: snap.rows, nodes: snap.nodes, fromMs: snap.fromMs, toMs: snap.toMs }
+      }
+      catch (err) {
+        return { success: false, error: String(err?.message ?? err) }
+      }
+    })
+
+    // 报告查看(面板/前端):run_id → 诊断任务 → workspace run 目录 → report md 内容
+    ctx.route('GET', '/report', async (event) => {
+      try {
+        const u = new URL(String(event?.path ?? event?.node?.req?.url ?? ''), 'http://local')
+        const runId = String(u.searchParams.get('run_id') ?? '').trim()
+        if (!runId) return { success: false, error: 'run_id 必填' }
+        const base = baseOf(ctx)
+        if (!base) return { success: false, error: 'diag.base_url 非法' }
+        const st = (await jget(ctx, `${base}/api/diagnosis/tasks/${encodeURIComponent(runId)}`, 8000)).data ?? {}
+        if (String(st.status ?? '') !== 'completed') {
+          return { success: false, error: `诊断 ${runId} 状态 ${st.status ?? 'unknown'},报告未就绪` }
+        }
+        const ws = (await jget(ctx, `${base}/api/files/workspace`, 8000)).data ?? []
+        const names = (Array.isArray(ws) ? ws : [])
+        const meta = ctx.kv.get(runKey(runId)) ?? {}
+        // 目录命名不统一(<ts>_<scene> 或 <ts>_<scene>_<runId>):runId → scene → 最新 逐级兜底
+        const hit = names.find(e => String(e?.name ?? '').includes(runId))
+          ?? names.find(e => meta?.scene && String(e?.name ?? '').toLowerCase().includes(String(meta.scene).toLowerCase()))
+          ?? (names.length ? names.slice().sort((a, b) => Number(b?.created ?? 0) - Number(a?.created ?? 0))[0] : null)
+        if (!hit) return { success: false, error: `诊断服务 workspace 未找到 ${runId} 的运行目录` }
+        const rep = (await jget(ctx, `${base}/api/files/workspace/report/${encodeURIComponent(hit.name)}`, 15000)).data
+        return { success: true, run_id: runId, run_name: hit.name, report: rep ?? null }
+      }
+      catch (err) {
+        return { success: false, error: String(err?.message ?? err) }
+      }
+    })
+
+    ctx.logger.info('diag-bridge 就绪:工具 diag_run/diag_status + 路由 health/runs/snapshot/report,轮询 15s 已启动')
+  },
+}