diff --git a/ui/src/router/admin/index.ts b/ui/src/router/admin/index.ts
index 3ce08ade5e6..a535a91ff26 100644
--- a/ui/src/router/admin/index.ts
+++ b/ui/src/router/admin/index.ts
@@ -14,6 +14,8 @@ const router = createRouter({
   history: createWebHistory(ADMIN_BASE_PATH),
   routes: [
     { path: '/', redirect: { name: 'workspace-home', params: { workspaceId: 'default' } } },
+    // TODO(demo): 执行详情样式预览，验收后删除
+    { path: '/details-demo', name: 'details-demo', component: () => import('@/views/details-demo/index.vue'), meta: { title: '执行详情 Demo' } },
     ...loginRoutes,
     ...workflowRoutes,
     workspaceRoutes,
@@ -24,7 +26,7 @@ const router = createRouter({
 
 router.beforeEach(async (to, from) => {
   if (to.path !== from.path) NProgress.start()
-  const notAuthRouteNameList = ['login', 'forgot-password', 'not-found']
+  const notAuthRouteNameList = ['login', 'forgot-password', 'not-found', 'details-demo']
   const { auth, user } = useStore()
 
   if (!notAuthRouteNameList.includes(to.name ? to.name.toString() : '')) {
diff --git a/ui/src/views/details-demo/index.vue b/ui/src/views/details-demo/index.vue
new file mode 100644
index 00000000000..f80a80f192e
--- /dev/null
+++ b/ui/src/views/details-demo/index.vue
@@ -0,0 +1,82 @@
+<script setup lang="ts">
+import { ref } from 'vue'
+import ExecutionDetailContent from '@/workflow-canvas/details/index.vue'
+import { WorkflowMode, WorkflowNodeType } from '@/workflow-canvas/types'
+import type { ExecutionNodeDetail } from '@/workflow-canvas/details/types'
+
+defineOptions({ name: 'DetailsDemo' })
+
+// 一个 AI 对话节点的示例执行详情。
+function makeAiChat(overrides: Partial<ExecutionNodeDetail> = {}): ExecutionNodeDetail {
+  return {
+    type: WorkflowNodeType.AiChat,
+    name: 'AI 对话',
+    status: 200,
+    index: 0,
+    run_time: 1.234,
+    message_tokens: 128,
+    answer_tokens: 96,
+    system: '你是一个乐于助人的中文助手，请简洁作答。',
+    history_message: [
+      { role: 'user', content: '你好' },
+      { role: 'ai', content: '你好，有什么可以帮你的吗？' },
+    ],
+    question: '用一句话介绍一下 MaxKB。',
+    reasoning_content: '用户想要一句话简介，需覆盖：开源、企业级、知识库、Agent 三个要点，控制在一句。',
+    answer:
+      '**MaxKB** 是一款开源的企业级 AI 应用平台，支持基于知识库的 RAG 检索、Agent 工作流编排与 MCP 工具调用。\n\n- 开源、可私有化部署\n- RAG + 工作流 + 工具\n- 多模型接入',
+    ...overrides,
+  }
+}
+
+const detail = ref<ExecutionNodeDetail[]>([
+  makeAiChat({ index: 0 }),
+  {
+    type: WorkflowNodeType.LoopNode,
+    name: '循环',
+    status: 200,
+    index: 1,
+    run_time: 4.567,
+    loop_type: 'ARRAY',
+    loop_node_data: {
+      '0': {
+        aiNode: makeAiChat({
+          name: 'AI 对话（第 1 轮）',
+          index: 0,
+          question: '第 1 项：介绍知识库',
+          answer: '知识库把文档切片、向量化后存入 pgvector，用于语义检索。',
+          run_time: 0.98,
+        }),
+      },
+      '1': {
+        aiNode: makeAiChat({
+          name: 'AI 对话（第 2 轮）',
+          index: 0,
+          question: '第 2 项：介绍工作流',
+          answer: '工作流用 LogicFlow 编排节点，支持循环、分支与工具调用。',
+          run_time: 1.12,
+          status: 500,
+          err_message: '模型调用超时（示例错误，用于查看失败态样式）',
+        }),
+      },
+    },
+  } as ExecutionNodeDetail,
+])
+
+const mode = ref<WorkflowMode>(WorkflowMode.Application)
+</script>
+
+<template>
+  <div class="min-h-screen bg-N100 p-6">
+    <div class="mx-auto max-w-[480px]">
+      <div class="mb-4 flex items-center gap-3">
+        <h3>执行详情样式 Demo</h3>
+        <el-radio-group v-model="mode" size="small">
+          <el-radio-button :value="WorkflowMode.Application">application</el-radio-button>
+          <el-radio-button :value="WorkflowMode.Knowledge">knowledge</el-radio-button>
+        </el-radio-group>
+      </div>
+      <ExecutionDetailContent :detail="detail" :workflow-mode="mode" />
+    </div>
+  </div>
+</template>
diff --git a/ui/src/workflow-canvas/WORKFLOW_README.md b/ui/src/workflow-canvas/WORKFLOW_README.md
index cd5dbe36dfa..d5c174f360c 100644
--- a/ui/src/workflow-canvas/WORKFLOW_README.md
+++ b/ui/src/workflow-canvas/WORKFLOW_README.md
@@ -22,6 +22,7 @@
 src/workflow-canvas/
 ├── component/          # 画布内可复用的控制和搜索组件
 ├── config/             # 节点数据、映射、常量及预留的本地化配置
+├── details/            # 执行详情：顶层分发入口与公共卡壳，节点内容由 nodes/*/details/ 提供
 ├── core/               # 稳定的画布内核与所有节点共用的基础能力
 │   ├── edge/           # 普通边、循环边及边删除按钮
 │   └── node-container/ # 节点容器、锚点按钮及私有的条件、操作下拉组件
@@ -225,6 +226,40 @@ MkFormList 的排序、增删均以 `cloneDeep` 回写；MkTable 保留普通行
 带 ID 的新行使用 `default-item` 工厂函数，每次添加生成独立 ID。条件分支等保留锚点 ID、
 固定末尾分支的特殊规则不能直接替换成普通列表移动。
 
+### 执行详情（`details/`）
+
+节点执行详情按三层组织，与画布节点注册同源，通过节点类型分发渲染：
+
+- `details/index.vue`（`ExecutionDetailContent`）是顶层分发层：接收 `detail` 数组和 `workflowMode`，
+  按 `index` 升序，逐项按节点类型渲染对应节点的详情组件。类型到详情组件的映射来自
+  `import.meta.glob('../nodes/*/index.ts')` 收集的默认导出中的 `details` 字段，与画布节点注册同一份
+  清单，不按目录名推断，也不再单独维护映射表。
+- `details/DetailContainer.vue` 是纯布局卡壳（架子），不认识任何具体节点类型：提供 `#header` 具名
+  插槽（透出折叠 `show`）、折叠体默认插槽承载节点内容，以及 `showContentOnError` 决定失败时是否仍
+  展示内容（默认失败只显示错误日志块）。
+- `details/BaseHeader.vue` 是公共头部：折叠箭头、节点图标、名称、耗时和状态图标；是否显示 tokens 由
+  节点通过 `show-tokens` 布尔控制，tokens 数值由 `BaseHeader` 从 `data` 自行计算，容器不参与该判断。
+- 详情载荷类型 `ExecutionNodeDetail` 在 `details/types.ts`，以开放索引签名承载各节点动态字段，仅显式
+  声明外层卡片通用字段；分发和模式统一使用 `WorkflowMode` 枚举，不自造字符串联合类型。
+
+每个节点的详情入口统一为 `nodes/<node-type>/details/index.vue`，并在该节点 `index.ts` 的默认导出中通过
+`details` 字段注册（LogicFlow 注册忽略该额外字段）：
+
+- 内容不随工作流模式变化的节点（如循环），`details/index.vue` 直接就是详情内容。
+- 内容随模式不同的节点（如 AI 对话），`details/index.vue` 按 `workflowMode` 用映射分发到同目录下的
+  `application.vue`、`knowledge.vue` 等模式文件，其它模式兜底到应用视图；各模式文件各自完整、自包含，
+  共用部分不强行抽取。
+
+详情内的灰底标题块直接使用内联 Tailwind（`overflow-hidden rounded-md bg-N100` + `h5.px-3.py-2` +
+`border-t border-dashed px-3 py-2 text-N900`），不引入独立的区块组件。只读 Markdown 回答复用全局
+`MdPreview`，并在节点内通过 `:deep()` 覆盖其固定高度与背景以融入灰底。
+
+循环节点的详情是唯一的递归点：它维护循环设置与轮次选择，选中某轮后把该轮的子节点数组交回顶层
+`ExecutionDetailContent` 渲染（排序与类型分发由顶层负责），并对外层卡壳传 `show-content-on-error`
+以便整体失败时仍能展开已执行的子节点。递归由数据驱动、逐层向下收敛（一个循环的轮次数据不含它
+自身），执行详情是有限树因而必然终止；`details/index.vue` 与节点详情之间的循环 import 仅在渲染期
+使用、不在模块求值期调用，属正常的组件递归引用。
+
 ## View 接入约定
 
 - 所有画布路由页面均放在 `src/views/workflow/`，并以 `XxxWorkflowView.vue` 命名，例如
@@ -260,7 +295,9 @@ MkFormList 的排序、增删均以 `cloneDeep` 回写；MkTable 保留普通行
 3. 在 `config/node-mapping.ts` 中维护节点映射；需要作为初始节点时再加入默认节点集合。
 4. 在 `node-menu/menu.ts` 中加入适用画布模式的菜单分组。
 5. 在 `nodes/<node-type>/` 中实现注册文件和节点视图，并按需增加 `icons/` 图标。
-6. 确认节点能被自动注册、从菜单添加、正确连线、校验并导出图数据。
+6. 如需执行详情，在 `nodes/<node-type>/details/index.vue` 中实现，并在 `index.ts` 默认导出中通过
+   `details` 字段注册；内容随模式变化时在该目录下按 `workflowMode` 分发到模式文件。
+7. 确认节点能被自动注册、从菜单添加、正确连线、校验并导出图数据。
 
 节点协议值统一使用 `WorkflowNodeType` 等枚举或画布常量，不在判断和映射中重复书写
 `ai-chat-node`、`tool-custom-node` 等字符串字面量。
diff --git a/ui/src/workflow-canvas/details/BaseHeader.vue b/ui/src/workflow-canvas/details/BaseHeader.vue
new file mode 100644
index 00000000000..276dc4aad22
--- /dev/null
+++ b/ui/src/workflow-canvas/details/BaseHeader.vue
@@ -0,0 +1,45 @@
+<script setup lang="ts">
+import { computed, type Component } from 'vue'
+import { CaretRight, CircleCheck, CircleClose, Loading } from '@element-plus/icons-vue'
+import { iconComponent } from '@/workflow-canvas/icons/utils'
+import type { ExecutionNodeDetail } from './types'
+
+defineOptions({ name: 'BaseHeader' })
+
+const props = defineProps<{
+  data: ExecutionNodeDetail
+  show?: boolean
+  // 是否显示 tokens 由各节点自行决定；tokens 数值统一由 data 计算。
+  showTokens?: boolean
+}>()
+
+const nodeIconComponent = computed<Component | null>(() => iconComponent(`${props.data?.type}-icon`))
+const isSuccess = computed(() => props.data?.status === 200)
+const isRunning = computed(() => props.data?.status === 202)
+const tokenCount = computed(() => (props.data?.message_tokens || 0) + (props.data?.answer_tokens || 0))
+</script>
+
+<template>
+  <div class="flex-between">
+    <div class="flex items-center gap-2">
+      <el-icon class="text-N600 transition-transform" :class="{ 'rotate-90': show }">
+        <CaretRight />
+      </el-icon>
+      <component :is="nodeIconComponent" v-if="nodeIconComponent" :size="24" />
+      <h4>{{ data.name }}</h4>
+    </div>
+    <div class="flex items-center gap-3 text-N600">
+      <span v-if="showTokens">{{ tokenCount }} tokens</span>
+      <span v-if="!isRunning">{{ data?.run_time?.toFixed(2) || '0.00' }} s</span>
+      <el-icon v-if="isSuccess" class="text-success" :size="16">
+        <CircleCheck />
+      </el-icon>
+      <el-icon v-else-if="isRunning" class="is-loading text-primary" :size="16">
+        <Loading />
+      </el-icon>
+      <el-icon v-else class="text-danger" :size="16">
+        <CircleClose />
+      </el-icon>
+    </div>
+  </div>
+</template>
diff --git a/ui/src/workflow-canvas/details/DetailContainer.vue b/ui/src/workflow-canvas/details/DetailContainer.vue
new file mode 100644
index 00000000000..4e1045cc195
--- /dev/null
+++ b/ui/src/workflow-canvas/details/DetailContainer.vue
@@ -0,0 +1,38 @@
+<script setup lang="ts">
+import { computed, ref } from 'vue'
+import type { ExecutionNodeDetail } from './types'
+
+defineOptions({ name: 'DetailContainer' })
+
+const props = withDefaults(
+  defineProps<{
+    data: ExecutionNodeDetail
+    // 循环等节点即使整体失败也需展示已执行的子节点详情，其余节点失败时只展示错误日志。
+    showContentOnError?: boolean
+  }>(),
+  {
+    showContentOnError: false,
+  },
+)
+
+const show = ref(false)
+const isSuccess = computed(() => props.data?.status === 200)
+const showContent = computed(() => isSuccess.value || props.showContentOnError)
+</script>
+
+<template>
+  <el-card class="execution-detail-card mb-2" shadow="never" style="--el-card-padding: 8px 12px">
+    <div class="cursor-pointer" @click="show = !show">
+      <slot name="header" :show="show" />
+    </div>
+    <el-collapse-transition>
+      <div v-if="show" class="mt-3 flex flex-col gap-2">
+        <slot v-if="showContent" />
+        <div v-else class="overflow-hidden rounded-md bg-N100">
+          <h5 class="px-3 py-2">错误日志</h5>
+          <div class="border-t border-dashed px-3 py-2 text-N900">{{ data.err_message || '-' }}</div>
+        </div>
+      </div>
+    </el-collapse-transition>
+  </el-card>
+</template>
diff --git a/ui/src/workflow-canvas/details/index.vue b/ui/src/workflow-canvas/details/index.vue
new file mode 100644
index 00000000000..47e12c53261
--- /dev/null
+++ b/ui/src/workflow-canvas/details/index.vue
@@ -0,0 +1,55 @@
+<script setup lang="ts">
+import { computed, type Component } from 'vue'
+import { WorkflowMode, WorkflowNodeType } from '@/workflow-canvas/types'
+import type { ExecutionNodeDetail } from './types'
+
+defineOptions({ name: 'ExecutionDetailContent' })
+
+const props = withDefaults(
+  defineProps<{
+    detail?: ExecutionNodeDetail[]
+    workflowMode?: WorkflowMode
+  }>(),
+  {
+    detail: () => [],
+    workflowMode: WorkflowMode.Application,
+  },
+)
+
+// 复用节点注册表：每个节点的 index.ts 默认导出携带注册的 WorkflowNodeType 及其执行详情视图
+// details（各节点的整卡），与画布注册节点同源。据此建立 type -> 详情组件 映射，按节点类型渲染。
+const nodeRegistryModules = import.meta.glob<{ default: { type: WorkflowNodeType; details?: Component } }>(
+  '../nodes/*/index.ts',
+  { eager: true },
+)
+
+const nodeDetailComponentMap: Map<WorkflowNodeType, Component> = (() => {
+  const map = new Map<WorkflowNodeType, Component>()
+  for (const { default: registration } of Object.values(nodeRegistryModules)) {
+    if (registration?.details) map.set(registration.type, registration.details)
+  }
+  return map
+})()
+
+function nodeDetailComponent(type?: WorkflowNodeType): Component | null {
+  return type ? (nodeDetailComponentMap.get(type) ?? null) : null
+}
+
+// 按 index 升序展示各节点执行详情。
+const sortedDetail = computed(() =>
+  [...props.detail].sort((a, b) => ((a?.index as number) || 0) - ((b?.index as number) || 0)),
+)
+</script>
+
+<template>
+  <div class="flex flex-col gap-2">
+    <template v-for="(item, index) in sortedDetail" :key="index">
+      <component
+        :is="nodeDetailComponent(item.type)"
+        v-if="nodeDetailComponent(item.type)"
+        :data="item"
+        :workflow-mode="workflowMode"
+      />
+    </template>
+  </div>
+</template>
diff --git a/ui/src/workflow-canvas/details/types.ts b/ui/src/workflow-canvas/details/types.ts
new file mode 100644
index 00000000000..36564124e11
--- /dev/null
+++ b/ui/src/workflow-canvas/details/types.ts
@@ -0,0 +1,18 @@
+/* eslint-disable @typescript-eslint/no-explicit-any */
+import type { WorkflowNodeType } from '@/workflow-canvas/types'
+
+/**
+ * 节点执行详情的运行时载荷。字段随节点类型（含循环、工具工作流的嵌套子节点）差异极大，
+ * 由后端按节点动态返回，因此以开放索引签名承载动态字段，仅显式声明外层卡片通用的字段。
+ * 各节点的 details.vue 直接读取本类型所需字段，待接口契约稳定后再逐节点收敛为精确类型。
+ */
+export interface ExecutionNodeDetail {
+  [key: string]: any
+  answer_tokens?: number
+  err_message?: string
+  message_tokens?: number
+  name?: string
+  run_time?: number
+  status?: number
+  type?: WorkflowNodeType
+}
diff --git a/ui/src/workflow-canvas/nodes/ai-chat-node/details/application.vue b/ui/src/workflow-canvas/nodes/ai-chat-node/details/application.vue
new file mode 100644
index 00000000000..4d273a6999f
--- /dev/null
+++ b/ui/src/workflow-canvas/nodes/ai-chat-node/details/application.vue
@@ -0,0 +1,102 @@
+<script setup lang="ts">
+import DetailContainer from '@/workflow-canvas/details/DetailContainer.vue'
+import BaseHeader from '@/workflow-canvas/details/BaseHeader.vue'
+import type { ExecutionNodeDetail } from '@/workflow-canvas/details/types'
+
+defineOptions({ name: 'AiChatApplicationDetails' })
+
+defineProps<{
+  data: ExecutionNodeDetail
+}>()
+</script>
+
+<template>
+  <DetailContainer :data="data">
+    <template #header="{ show }">
+      <BaseHeader :data="data" :show="show" show-tokens />
+    </template>
+
+    <!-- 系统提示词 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">系统提示词</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">{{ data.system || '-' }}</div>
+    </div>
+
+    <!-- 历史记录 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">历史记录</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">
+        <template v-if="data.history_message?.length > 0">
+          <p v-for="(history, historyIndex) in data.history_message" :key="historyIndex" class="my-1">
+            <span class="mr-1 text-N600">{{ history.role }}:</span>
+            <span>{{ history.content }}</span>
+          </p>
+        </template>
+        <template v-else>-</template>
+      </div>
+    </div>
+
+    <!-- 本次对话 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">本次对话</h5>
+      <div class="whitespace-pre-wrap border-t border-dashed px-3 py-2 text-N900">
+        <template v-if="Array.isArray(data.question)">
+          <div v-for="(item, qIndex) in data.question" :key="qIndex" class="mb-2">
+            <el-image
+              v-if="item.type === 'image_url'"
+              :src="item.image_url?.url || item.image_url"
+              fit="cover"
+              class="block h-10 w-10 rounded-md"
+              :zoom-rate="1.2"
+              :max-scale="7"
+              :min-scale="0.2"
+            />
+            <video
+              v-else-if="item.type === 'video_url'"
+              :src="item.video_url?.url || item.video_url"
+              class="block w-[170px] rounded-md"
+              autoplay
+              controls
+            />
+            <div v-else-if="item.type === 'text'">{{ item.text }}</div>
+            <div v-else>{{ item }}</div>
+          </div>
+        </template>
+        <template v-else>
+          {{ data.question || '-' }}
+        </template>
+      </div>
+    </div>
+
+    <!-- 思考过程 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">思考过程</h5>
+      <div class="whitespace-pre-wrap border-t border-dashed px-3 py-2 text-N900">
+        {{ data.reasoning_content || '-' }}
+      </div>
+    </div>
+
+    <!-- AI 回答 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">AI 回答</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">
+        <MdPreview
+          v-if="data.answer"
+          editor-id="preview-only"
+          :model-value="data.answer"
+          class="ai-chat-answer"
+          noImgZoomIn
+        />
+        <template v-else>-</template>
+      </div>
+    </div>
+  </DetailContainer>
+</template>
+
+<style scoped lang="scss">
+/* 只读回答需要自适应高度并融入区块灰底，覆盖全局 Markdown 编辑器的固定高度与背景。 */
+:deep(.ai-chat-answer.md-editor) {
+  background: transparent;
+  height: auto;
+}
+</style>
diff --git a/ui/src/workflow-canvas/nodes/ai-chat-node/details/index.vue b/ui/src/workflow-canvas/nodes/ai-chat-node/details/index.vue
new file mode 100644
index 00000000000..04350d6baf5
--- /dev/null
+++ b/ui/src/workflow-canvas/nodes/ai-chat-node/details/index.vue
@@ -0,0 +1,28 @@
+<script setup lang="ts">
+import ApplicationDetails from './application.vue'
+import KnowledgeDetails from './knowledge.vue'
+import { WorkflowMode } from '@/workflow-canvas/types'
+import type { ExecutionNodeDetail } from '@/workflow-canvas/details/types'
+
+defineOptions({ name: 'AiChatNodeDetail' })
+
+withDefaults(
+  defineProps<{
+    data: ExecutionNodeDetail
+    workflowMode?: WorkflowMode
+  }>(),
+  {
+    workflowMode: WorkflowMode.Application,
+  },
+)
+
+// 按工作流模式选择详情视图；循环体等其它模式沿用应用视图。
+const kv: Partial<Record<WorkflowMode, typeof ApplicationDetails>> = {
+  [WorkflowMode.Application]: ApplicationDetails,
+  [WorkflowMode.Knowledge]: KnowledgeDetails,
+}
+</script>
+
+<template>
+  <component :is="kv[workflowMode] ?? ApplicationDetails" :data="data" />
+</template>
diff --git a/ui/src/workflow-canvas/nodes/ai-chat-node/details/knowledge.vue b/ui/src/workflow-canvas/nodes/ai-chat-node/details/knowledge.vue
new file mode 100644
index 00000000000..599dc3026c3
--- /dev/null
+++ b/ui/src/workflow-canvas/nodes/ai-chat-node/details/knowledge.vue
@@ -0,0 +1,88 @@
+<script setup lang="ts">
+import DetailContainer from '@/workflow-canvas/details/DetailContainer.vue'
+import BaseHeader from '@/workflow-canvas/details/BaseHeader.vue'
+import type { ExecutionNodeDetail } from '@/workflow-canvas/details/types'
+
+defineOptions({ name: 'AiChatKnowledgeDetails' })
+
+defineProps<{
+  data: ExecutionNodeDetail
+}>()
+</script>
+
+<template>
+  <DetailContainer :data="data">
+    <template #header="{ show }">
+      <BaseHeader :data="data" :show="show" show-tokens />
+    </template>
+
+    <!-- 系统提示词 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">系统提示词</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">{{ data.system || '-' }}</div>
+    </div>
+
+    <!-- 用户提示词 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">用户提示词</h5>
+      <div class="whitespace-pre-wrap border-t border-dashed px-3 py-2 text-N900">
+        <template v-if="Array.isArray(data.question)">
+          <div v-for="(item, qIndex) in data.question" :key="qIndex" class="mb-2">
+            <el-image
+              v-if="item.type === 'image_url'"
+              :src="item.image_url?.url || item.image_url"
+              fit="cover"
+              class="block h-10 w-10 rounded-md"
+              :zoom-rate="1.2"
+              :max-scale="7"
+              :min-scale="0.2"
+            />
+            <video
+              v-else-if="item.type === 'video_url'"
+              :src="item.video_url?.url || item.video_url"
+              class="block w-[170px] rounded-md"
+              autoplay
+              controls
+            />
+            <div v-else-if="item.type === 'text'">{{ item.text }}</div>
+            <div v-else>{{ item }}</div>
+          </div>
+        </template>
+        <template v-else>
+          {{ data.question || '-' }}
+        </template>
+      </div>
+    </div>
+
+    <!-- 思考过程 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">思考过程</h5>
+      <div class="whitespace-pre-wrap border-t border-dashed px-3 py-2 text-N900">
+        {{ data.reasoning_content || '-' }}
+      </div>
+    </div>
+
+    <!-- AI 回答 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">AI 回答</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">
+        <MdPreview
+          v-if="data.answer"
+          editor-id="preview-only"
+          :model-value="data.answer"
+          class="ai-chat-answer"
+          noImgZoomIn
+        />
+        <template v-else>-</template>
+      </div>
+    </div>
+  </DetailContainer>
+</template>
+
+<style scoped lang="scss">
+/* 只读回答需要自适应高度并融入区块灰底，覆盖全局 Markdown 编辑器的固定高度与背景。 */
+:deep(.ai-chat-answer.md-editor) {
+  background: transparent;
+  height: auto;
+}
+</style>
diff --git a/ui/src/workflow-canvas/nodes/ai-chat-node/index.ts b/ui/src/workflow-canvas/nodes/ai-chat-node/index.ts
index 3970bf68978..63ad7e2d14e 100644
--- a/ui/src/workflow-canvas/nodes/ai-chat-node/index.ts
+++ b/ui/src/workflow-canvas/nodes/ai-chat-node/index.ts
@@ -1,4 +1,5 @@
 import AiChatNode from './index.vue'
+import AiChatNodeDetail from './details/index.vue'
 import { WorkflowNodeModel, WorkflowNodeView } from '@/workflow-canvas/core/workflow-node'
 import { WorkflowNodeType } from '@/workflow-canvas/types.ts'
 
@@ -8,4 +9,4 @@ class AiChatNodeView extends WorkflowNodeView {
   }
 }
 
-export default { type: WorkflowNodeType.AiChat, model: WorkflowNodeModel, view: AiChatNodeView }
+export default { type: WorkflowNodeType.AiChat, model: WorkflowNodeModel, view: AiChatNodeView, details: AiChatNodeDetail }
diff --git a/ui/src/workflow-canvas/nodes/loop-node/details/index.vue b/ui/src/workflow-canvas/nodes/loop-node/details/index.vue
new file mode 100644
index 00000000000..e982650b6ac
--- /dev/null
+++ b/ui/src/workflow-canvas/nodes/loop-node/details/index.vue
@@ -0,0 +1,66 @@
+<script setup lang="ts">
+import { computed, ref } from 'vue'
+import DetailContainer from '@/workflow-canvas/details/DetailContainer.vue'
+import BaseHeader from '@/workflow-canvas/details/BaseHeader.vue'
+import ExecutionDetailContent from '@/workflow-canvas/details/index.vue'
+import { WorkflowMode } from '@/workflow-canvas/types'
+import type { ExecutionNodeDetail } from '@/workflow-canvas/details/types'
+
+defineOptions({ name: 'LoopNodeDetail' })
+
+const props = withDefaults(
+  defineProps<{
+    data: ExecutionNodeDetail
+    workflowMode?: WorkflowMode
+  }>(),
+  {
+    workflowMode: WorkflowMode.Application,
+  },
+)
+
+// 每一轮循环的子节点执行详情，键为轮次下标。
+const loopKeys = computed(() => Object.keys(props.data?.loop_node_data || {}))
+const currentLoop = ref(loopKeys.value[0] ?? '')
+
+// 当前选中轮次的子节点数组，交回顶层 dispatcher 递归渲染（其内部已按 index 排序、按类型分发）。
+const currentLoopDetail = computed<ExecutionNodeDetail[]>(() =>
+  Object.values(props.data?.loop_node_data?.[currentLoop.value] || {}),
+)
+</script>
+
+<template>
+  <!-- 循环节点整体失败也需展示已执行轮次的子节点详情。 -->
+  <DetailContainer :data="data" show-content-on-error>
+    <template #header="{ show }">
+      <BaseHeader :data="data" :show="show" />
+    </template>
+
+    <!-- 循环设置 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">循环设置</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">
+        <p class="mb-2"><span class="mr-1 text-N600">循环类型:</span>{{ data.loop_type || '-' }}</p>
+        <p>
+          <span class="mr-1 text-N600">循环数组:</span>
+          {{ data.loop_type === 'NUMBER' ? data.number : loopKeys.join(', ') || '-' }}
+        </p>
+      </div>
+    </div>
+
+    <!-- 循环详情 -->
+    <div class="overflow-hidden rounded-md bg-N100">
+      <h5 class="px-3 py-2">循环详情</h5>
+      <div class="border-t border-dashed px-3 py-2 text-N900">
+        <template v-if="loopKeys.length > 0">
+          <el-radio-group v-model="currentLoop" class="mb-2">
+            <el-radio-button v-for="key in loopKeys" :key="key" :value="key">
+              第 {{ Number(key) + 1 }} 轮
+            </el-radio-button>
+          </el-radio-group>
+          <ExecutionDetailContent :detail="currentLoopDetail" :workflow-mode="workflowMode" />
+        </template>
+        <template v-else>-</template>
+      </div>
+    </div>
+  </DetailContainer>
+</template>
diff --git a/ui/src/workflow-canvas/nodes/loop-node/index.ts b/ui/src/workflow-canvas/nodes/loop-node/index.ts
index 5dcb664435b..8b5ee6f07fc 100644
--- a/ui/src/workflow-canvas/nodes/loop-node/index.ts
+++ b/ui/src/workflow-canvas/nodes/loop-node/index.ts
@@ -1,4 +1,5 @@
 import LoopNodeVue from './index.vue'
+import LoopNodeDetail from './details/index.vue'
 import { WorkflowNodeModel, WorkflowNodeView } from '@/workflow-canvas/core/workflow-node'
 import { WorkflowNodeType } from '@/workflow-canvas/types'
 import type { Model } from '@logicflow/core'
@@ -25,4 +26,5 @@ export default {
   type: WorkflowNodeType.LoopNode,
   model: LoopNodeModel,
   view: LoopNodeView,
+  details: LoopNodeDetail,
 }