From 71feae4bbd9579a100a9c1d63e37b60f08bf7d5a Mon Sep 17 00:00:00 2001
From: lixiaolong <702156524@qq.com>
Date: Wed, 19 Aug 2026 15:00:17 +0800
Subject: [PATCH] =?UTF-8?q?feat(flow):=20=E6=B7=BB=E5=8A=A0=E9=80=9A?=
 =?UTF-8?q?=E7=94=A8=E5=9B=BE=E7=89=87=E5=88=86=E6=9E=90=E5=8A=9F=E8=83=BD?=
 =?UTF-8?q?=E5=B9=B6=E4=BC=98=E5=8C=96=E5=AA=92=E4=BD=93=E5=88=86=E6=9E=90?=
 =?UTF-8?q?=E8=8A=82=E7=82=B9?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- 新增通用图片分析节点类型，支持按提示词分析图片并返回指定内容
- 添加详细的图片分析参数配置界面，包括图片源选择、提示词输入等功能
- 实现图片分析节点的输出参数定义和处理逻辑
- 更新视频分析节点，增加详细输出选项和结果字段
- 重构错误状态恢复功能，更名为运行状态恢复并更新相关UI
- 添加图片分析结果的日志展示支持
- 优化流程资源配置，增加图片分析相关的常量定义
- 更新单节点执行参数构建逻辑，适配新的图片分析功能
---
 src/api/device/system.js                      |  11 +
 src/utils/flowResources.js                    |  25 ++
 src/views/flow/components/ParamsDrawer.vue    |   2 +
 .../flow/components/TaskExecutionLog.vue      |  20 +-
 .../components/params/ImageAnalysisParams.vue | 274 ++++++++++++++++++
 .../components/params/MediaAnalysisParams.vue |  23 ++
 src/views/flow/config.js                      |  23 ++
 src/views/flow/index.vue                      | 133 ++++++---
 .../intelligenceTest/runningTask/task-log.vue |   2 +
 9 files changed, 467 insertions(+), 46 deletions(-)
 create mode 100644 src/views/flow/components/params/ImageAnalysisParams.vue

diff --git a/src/api/device/system.js b/src/api/device/system.js
index 18508dc..55356a9 100644
--- a/src/api/device/system.js
+++ b/src/api/device/system.js
@@ -8,6 +8,17 @@ export function getSafetyState(robotId) {
   })
 }
 
+export function restoreOperationalState(data) {
+  return request({
+    url: '/api/system/restoreOperationalState',
+    method: 'post',
+    data,
+  })
+}
+
+/**
+ * @deprecated Use restoreOperationalState.
+ */
 export function recoverSafetyState(data) {
   return request({
     url: '/api/system/recoverSafetyState',
diff --git a/src/utils/flowResources.js b/src/utils/flowResources.js
index a761b30..b6f8a98 100644
--- a/src/utils/flowResources.js
+++ b/src/utils/flowResources.js
@@ -4,6 +4,10 @@ export const FLOW_SCHEMA_VERSION = 2
 export const DEFAULT_ROLE_KEY = 'default_executor'
 export const FLOW_NODE_RESOURCE_OWNER = 'FLOW_NODE'
 export const USER_RESOURCE_OWNER = 'USER'
+export const IMAGE_ANALYSIS_OUTPUT_PARAMS = Object.freeze([
+  { name: 'result', type: 'object', desc: '按提示词返回的结果', children: [], disabled: true },
+  { name: 'resultText', type: 'string', desc: '模型返回的原始文本', disabled: true },
+])
 
 const DEVICE_KIND_META = Object.freeze({
   1: { key: 'agv', name: '移动底盘' },
@@ -98,6 +102,27 @@ export const normalizeFlowResourceSchema = (sourceGraph) => {
   graph.edges = Array.isArray(graph.edges) ? graph.edges : []
   graph.nodes.forEach((node) => {
     node.type = LEGACY_NODE_TYPE_ALIASES[node.type] || node.type
+    if (node.type === 'imageAnalysis') {
+      const properties = node.properties || (node.properties = {})
+      properties.action = 'IMAGE_ANALYZE'
+      properties.nodeType = 'LLM'
+      properties.outputType = 'json'
+      properties.outputParams = clone(IMAGE_ANALYSIS_OUTPUT_PARAMS)
+    }
+    if (node.properties?.action === 'VIDEO_ANALYZE') {
+      const outputParams = Array.isArray(node.properties.outputParams)
+        ? node.properties.outputParams
+        : []
+      if (!outputParams.some((param) => param?.name === 'result')) {
+        outputParams.unshift({ name: 'result', type: 'string', desc: '分析结果', disabled: true })
+      }
+      if (!outputParams.some((param) => param?.name === 'detailedOutput')) {
+        outputParams.push({
+          name: 'detailedOutput', type: 'boolean', desc: '是否返回详细事件证据', disabled: true,
+        })
+      }
+      node.properties.outputParams = outputParams
+    }
   })
   const start = getStartNode(graph)
   if (!start) return graph
diff --git a/src/views/flow/components/ParamsDrawer.vue b/src/views/flow/components/ParamsDrawer.vue
index 6b307bf..c33a11f 100644
--- a/src/views/flow/components/ParamsDrawer.vue
+++ b/src/views/flow/components/ParamsDrawer.vue
@@ -40,6 +40,7 @@ import DeviceUniversal from './params/DeviceUniversal.vue'
 import SwitchNodeParams from './params/SwitchNodeParams.vue'
 import ActionQueueParams from './params/ActionQueueParams.vue'
 import MediaAnalysisParams from './params/MediaAnalysisParams.vue'
+import ImageAnalysisParams from './params/ImageAnalysisParams.vue'
 
 const props = defineProps({
   drawer: Boolean,
@@ -68,6 +69,7 @@ const currentComponent = computed(() => {
   if (['deviceUniversal', 'device_universal'].includes(type)) return DeviceUniversal
   if (type === 'actionQueue') return ActionQueueParams
   if (type === 'mediaAnalysis') return MediaAnalysisParams
+  if (type === 'imageAnalysis') return ImageAnalysisParams
   if (type === 'branch') return SwitchNodeParams
   return null
 })
diff --git a/src/views/flow/components/TaskExecutionLog.vue b/src/views/flow/components/TaskExecutionLog.vue
index 00e1846..c088627 100644
--- a/src/views/flow/components/TaskExecutionLog.vue
+++ b/src/views/flow/components/TaskExecutionLog.vue
@@ -178,6 +178,7 @@ const analysisOf = (log) => {
   const profileCode = String(output.profileCode || '').toLowerCase()
   const isAudio = analysisType.includes('AUDIO') || profileCode.includes('audio') || Number(log.logType) === 4
   const isVideo = analysisType.includes('VIDEO') || profileCode.includes('video') || Number(log.logType) === 3
+  const isImage = analysisType.includes('IMAGE') || profileCode.includes('image')
 
   // 仅有媒体类型还不足以展示分析面板，至少要存在一个分析状态或业务结果字段。
   const hasResult = Boolean(
@@ -188,11 +189,13 @@ const analysisOf = (log) => {
     || output.labelName
     || output.summary
     || output.conclusion
+    || output.resultText
+    || output.result !== undefined
     || output.passed !== undefined
   )
 
   // 非分析日志也缓存 null，避免之后每次渲染都重复执行同一套识别逻辑。
-  if (!hasResult || (!isAudio && !isVideo)) {
+  if (!hasResult || (!isAudio && !isVideo && !isImage)) {
     analysisCache.set(log, null)
     return null
   }
@@ -223,9 +226,9 @@ const analysisOf = (log) => {
   // 它与已经结构化展示的分析字段一起从“详情”中移除，避免用户看到重复且难以理解的底层上下文。
   ;[
     'analysisStatus', 'analysisType', 'profileCode', 'analysisModel', 'analysisTimingMs',
-    'mediaUrl', 'audioUrl', 'videoUrl', 'imageUrl', 'label', 'labelName', 'matched',
+    'mediaUrl', 'mediaUrls', 'audioUrl', 'videoUrl', 'imageUrl', 'label', 'labelName', 'matched',
     'confidence', 'passed', 'summary', 'conclusion', 'events', '_executionContext',
-    'expectedLabel', 'expectedLabelName', 'decisionReason',
+    'expectedLabel', 'expectedLabelName', 'decisionReason', 'resultText', 'result',
   ].forEach((key) => delete detailSource[key])
 
   // 所有已知字段删除后还有剩余内容时，才生成详情 JSON，避免显示空对象。
@@ -233,7 +236,7 @@ const analysisOf = (log) => {
 
   // 统一生成展示模型，模板不需要了解各版本日志字段和状态枚举的差异。
   const result = {
-    title: isAudio ? '声音类型检测' : '视频分析',
+    title: isAudio ? '声音类型检测' : isImage ? '通用图片分析' : '视频分析',
     statusLabel,
 
     // 分析失败或业务判断未通过使用红色；仅匹配失败使用警告色，其余使用成功色。
@@ -249,7 +252,10 @@ const analysisOf = (log) => {
     matchLabel: isAudio ? '' : matched === true ? '匹配' : matched === false ? '未匹配' : '',
     confidence,
     passedLabel: passed === true ? '通过' : passed === false ? '未通过' : '',
-    summary: output.decisionReason || output.summary || output.conclusion || '',
+    summary: output.summary || output.conclusion || output.resultText
+      || (typeof output.result === 'string' ? output.result : '')
+      || (/^[A-Z_]+$/.test(String(output.decisionReason || '')) ? '' : output.decisionReason)
+      || '',
     isAudio,
     events: Array.isArray(output.events)
       ? output.events.slice(0, 20).map((event) => typeof event === 'string' ? event : JSON.stringify(event))
@@ -289,6 +295,10 @@ const mediaOf = (log) => {
   // mediaUrl 本身没有类型，必须结合分析类型或 profileCode 后才能作为视频/音频加入。
   if (analysisType.includes('VIDEO') || profileCode.includes('video')) addMedia(result, seen, 'video', output.mediaUrl)
   if (analysisType.includes('AUDIO') || profileCode.includes('audio')) addMedia(result, seen, 'audio', output.mediaUrl)
+  if (analysisType.includes('IMAGE') || profileCode.includes('image')) {
+    addMedia(result, seen, 'image', output.mediaUrls)
+    addMedia(result, seen, 'image', output.mediaUrl)
+  }
   addMedia(result, seen, 'image', log.screenshotUrl)
   addMedia(result, seen, 'video', log.videoUrl)
 
diff --git a/src/views/flow/components/params/ImageAnalysisParams.vue b/src/views/flow/components/params/ImageAnalysisParams.vue
new file mode 100644
index 0000000..9c1d4df
--- /dev/null
+++ b/src/views/flow/components/params/ImageAnalysisParams.vue
@@ -0,0 +1,274 @@
+<template>
+  <div class="image-analysis-params">
+    <div class="intro">
+      <div class="intro__title">通用图片分析</div>
+      <div class="intro__desc">根据提示词分析一张或多张图片，模型结果保持提示词要求的内容和格式。</div>
+    </div>
+
+    <el-form ref="formRef" :model="formData" :rules="rules" label-position="top">
+      <el-form-item label="图片来源" prop="imageMode">
+        <el-segmented v-model="formData.imageMode" :options="sourceOptions" />
+      </el-form-item>
+
+      <el-form-item v-if="formData.imageMode === 'quote'" label="引用上游图片" prop="imageQuote">
+        <el-cascader
+          ref="imageCascaderRef"
+          v-model="formData.imageQuote"
+          :options="quoteOptions"
+          :props="{ checkStrictly: true }"
+          placeholder="选择图片节点的 imageUrl 输出"
+          style="width: 100%"
+          @visible-change="loadQuoteOptions"
+          @change="handleQuoteChange"
+        />
+        <div class="field-hint">上游输出为图片数组时，将按原顺序全部分析，最多 12 张。</div>
+      </el-form-item>
+
+      <el-form-item v-else label="图片地址" prop="imageUrlsText">
+        <el-input
+          v-model="formData.imageUrlsText"
+          type="textarea"
+          :rows="4"
+          placeholder="每行一个 HTTP/HTTPS 图片地址，最多 12 张"
+        />
+      </el-form-item>
+
+      <el-form-item label="分析提示词" prop="prompt">
+        <el-input
+          v-model="formData.prompt"
+          type="textarea"
+          :rows="5"
+          maxlength="4000"
+          show-word-limit
+          placeholder="例如：读取图片中的仪表盘数值，只输出数字。或：识别所有仪表，按 JSON 输出名称、数值和单位。"
+        />
+      </el-form-item>
+
+      <el-form-item label="分析模式">
+        <el-segmented v-model="formData.analysisMode" :options="modeOptions" block />
+      </el-form-item>
+
+      <el-collapse v-model="advancedSections" class="advanced-settings">
+        <el-collapse-item title="参考图片（可选）" name="reference">
+          <el-form-item label="参考图片来源">
+            <el-segmented v-model="formData.referenceMode" :options="sourceOptions" />
+          </el-form-item>
+          <el-form-item v-if="formData.referenceMode === 'quote'" label="引用参考图片" prop="referenceQuote">
+            <el-cascader
+              ref="referenceCascaderRef"
+              v-model="formData.referenceQuote"
+              :options="quoteOptions"
+              :props="{ checkStrictly: true }"
+              placeholder="选择参考图片输出"
+              style="width: 100%"
+              clearable
+              @visible-change="loadQuoteOptions"
+              @change="handleReferenceQuoteChange"
+            />
+          </el-form-item>
+          <el-form-item v-else label="参考图片地址" prop="referenceImageUrl">
+            <el-input v-model="formData.referenceImageUrl" clearable placeholder="HTTP/HTTPS 图片地址" />
+          </el-form-item>
+          <div class="field-hint">参考图片会作为最后一张图片传入；请在提示词中说明它的用途。</div>
+        </el-collapse-item>
+
+        <el-collapse-item title="高级设置" name="advanced">
+          <el-form-item label="图片最大宽度">
+            <el-slider v-model="formData.analysisTuning.maxWidth" :min="640" :max="1600" :step="64" show-input />
+          </el-form-item>
+          <el-form-item label="最多分析图片">
+            <el-input-number v-model="formData.analysisTuning.maxImages" :min="1" :max="12" />
+          </el-form-item>
+          <el-form-item label="最大输出长度">
+            <el-input-number v-model="formData.analysisTuning.maxOutputTokens" :min="32" :max="4096" :step="32" />
+          </el-form-item>
+        </el-collapse-item>
+      </el-collapse>
+    </el-form>
+
+    <el-divider content-position="left">节点输出</el-divider>
+    <div class="output-list">
+      <div v-for="item in outputSummary" :key="item.name" class="output-item">
+        <code>{{ item.name }}</code><span>{{ item.desc }}</span>
+      </div>
+    </div>
+  </div>
+</template>
+
+<script setup>
+import { reactive, ref, watch } from 'vue'
+import { getInput } from '@/utils/flow'
+import { IMAGE_ANALYSIS_OUTPUT_PARAMS } from '@/utils/flowResources'
+
+const props = defineProps({ data: Object })
+const emit = defineEmits(['save-success', 'save-error'])
+
+const formRef = ref()
+const imageCascaderRef = ref()
+const referenceCascaderRef = ref()
+const quoteOptions = ref([])
+const advancedSections = ref([])
+
+const sourceOptions = [
+  { label: '引用节点输出', value: 'quote' },
+  { label: '输入图片地址', value: 'input' },
+]
+const modeOptions = [
+  { label: '快速', value: 'FAST' },
+  { label: '精准', value: 'ACCURATE' },
+]
+
+const formData = reactive({
+  imageMode: 'quote',
+  imageQuote: [],
+  imageQuoteType: 'output',
+  imageUrlsText: '',
+  referenceMode: 'input',
+  referenceQuote: [],
+  referenceQuoteType: 'output',
+  referenceImageUrl: '',
+  prompt: '',
+  analysisMode: 'FAST',
+  analysisTuning: { maxWidth: 896, maxImages: 12, maxOutputTokens: 384 },
+})
+
+const outputSummary = [
+  { name: 'result', desc: '按提示词返回的结果；JSON、数组和数字会自动解析' },
+  { name: 'resultText', desc: '模型返回的原始文本' },
+]
+
+const parseUrls = value => String(value || '').split(/\r?\n|,/).map(item => item.trim()).filter(Boolean)
+const validUrl = value => /^https?:\/\//i.test(value)
+const hasReference = () => formData.referenceMode === 'quote'
+  ? formData.referenceQuote.length > 0
+  : Boolean(formData.referenceImageUrl.trim())
+
+const rules = {
+  imageQuote: [{
+    validator: (_, value) => formData.imageMode !== 'quote' || value?.length
+      ? Promise.resolve()
+      : Promise.reject(new Error('请选择图片输出')),
+    trigger: 'change',
+  }],
+  imageUrlsText: [{
+    validator: (_, value) => {
+      if (formData.imageMode !== 'input') return Promise.resolve()
+      const urls = parseUrls(value)
+      if (!urls.length) return Promise.reject(new Error('请至少输入一个图片地址'))
+      if (urls.length > 12) return Promise.reject(new Error('最多输入 12 张图片'))
+      return urls.every(validUrl)
+        ? Promise.resolve()
+        : Promise.reject(new Error('图片地址必须是 HTTP/HTTPS 地址'))
+    },
+    trigger: 'blur',
+  }],
+  referenceImageUrl: [{
+    validator: (_, value) => !value || formData.referenceMode !== 'input' || validUrl(String(value).trim())
+      ? Promise.resolve()
+      : Promise.reject(new Error('请输入有效的参考图片地址')),
+    trigger: 'blur',
+  }],
+  prompt: [{ required: true, message: '请输入分析提示词', trigger: 'blur' }],
+}
+
+const loadQuoteOptions = visible => { if (visible) quoteOptions.value = getInput(props.data.id) || [] }
+const handleQuoteChange = () => {
+  formData.imageQuoteType = imageCascaderRef.value?.getCheckedNodes?.(true)?.[0]?.data?.type || 'output'
+}
+const handleReferenceQuoteChange = () => {
+  formData.referenceQuoteType = referenceCascaderRef.value?.getCheckedNodes?.(true)?.[0]?.data?.type || 'output'
+}
+const getParam = (params, name) => params.find(item => item.name === name)
+
+const legacyPrompt = (params) => {
+  const target = String(getParam(params, 'targetDescription')?.input || '').trim()
+  const instruction = String(getParam(params, 'instruction')?.input || '').trim()
+  const taskType = String(getParam(params, 'taskType')?.input || '').toUpperCase()
+  const parts = [target || instruction]
+  if (taskType === 'ICON_TEMPLATE_LOCATE') {
+    parts.push('最后一张图片是参考图片，请在其他图片中查找对应目标并返回坐标。')
+  }
+  parts.push('只返回分析结果，不输出通过、未通过或置信度。')
+  return parts.filter(Boolean).join('\n')
+}
+
+const initData = () => {
+  const params = props.data?.properties?.nodeParams || []
+  const images = getParam(params, 'imageUrl')
+  const reference = getParam(params, 'referenceImageUrl')
+  const tuning = getParam(params, 'analysisTuning')?.input || {}
+  formData.imageMode = images?.type === 'input' ? 'input' : 'quote'
+  formData.imageQuote = Array.isArray(images?.quote) ? [...images.quote] : []
+  formData.imageQuoteType = images?.quoteType || 'output'
+  formData.imageUrlsText = Array.isArray(images?.input) ? images.input.join('\n') : (images?.input || '')
+  formData.referenceMode = reference?.type === 'quote' ? 'quote' : 'input'
+  formData.referenceQuote = Array.isArray(reference?.quote) ? [...reference.quote] : []
+  formData.referenceQuoteType = reference?.quoteType || 'output'
+  formData.referenceImageUrl = reference?.input || ''
+  formData.prompt = getParam(params, 'prompt')?.input || legacyPrompt(params)
+  const savedMode = String(getParam(params, 'analysisMode')?.input || '').toUpperCase()
+  formData.analysisMode = savedMode === 'ACCURATE' ? 'ACCURATE' : 'FAST'
+  formData.analysisTuning.maxWidth = Number(tuning.maxWidth) || (formData.analysisMode === 'FAST' ? 896 : 1280)
+  formData.analysisTuning.maxImages = Number(tuning.maxImages) || 12
+  formData.analysisTuning.maxOutputTokens = Number(tuning.maxOutputTokens)
+    || (formData.analysisMode === 'FAST' ? 384 : 512)
+  advancedSections.value = hasReference() ? ['reference'] : []
+}
+watch(() => props.data, initData, { immediate: true, deep: true })
+
+const buildMediaParam = (name, mode, input, quote, quoteType, required = true) => ({
+  name,
+  type: mode,
+  input: mode === 'input' ? input : '',
+  quote: mode === 'quote' ? [...quote] : '',
+  quoteType,
+  required,
+  disabled: true,
+})
+
+const validateAndSave = async () => {
+  try {
+    await formRef.value.validate()
+    const nodeParams = [
+      { name: 'profileCode', type: 'input', input: 'common.image_analysis.v1', required: true, disabled: true },
+      buildMediaParam('imageUrl', formData.imageMode, parseUrls(formData.imageUrlsText), formData.imageQuote, formData.imageQuoteType),
+      { name: 'prompt', type: 'input', input: formData.prompt.trim(), required: true, disabled: true, componentType: 'textarea' },
+      { name: 'analysisMode', type: 'input', input: formData.analysisMode, required: false, disabled: true, componentType: 'select', selectOptions: modeOptions },
+      { name: 'analysisTuning', type: 'input', input: { ...formData.analysisTuning }, required: false, disabled: true },
+    ]
+    if (hasReference()) {
+      nodeParams.splice(2, 0, buildMediaParam(
+        'referenceImageUrl', formData.referenceMode, formData.referenceImageUrl.trim(),
+        formData.referenceQuote, formData.referenceQuoteType, false,
+      ))
+    }
+    const properties = {
+      ...props.data.properties,
+      name: '通用图片分析',
+      action: 'IMAGE_ANALYZE',
+      nodeType: 'LLM',
+      nodeParams,
+      outputType: 'json',
+      outputParams: JSON.parse(JSON.stringify(IMAGE_ANALYSIS_OUTPUT_PARAMS)),
+    }
+    delete properties.executionTarget
+    lf.setProperties(props.data.id, properties)
+    emit('save-success')
+  } catch {
+    emit('save-error')
+  }
+}
+
+defineExpose({ validateAndSave })
+</script>
+
+<style scoped lang="scss">
+.image-analysis-params { padding-top: 16px; }
+.intro { margin-bottom: 22px; padding: 14px 16px; border-left: 3px solid var(--el-color-primary); background: var(--el-fill-color-light); }
+.intro__title { color: var(--el-text-color-primary); font-weight: 650; }
+.intro__desc, .field-hint { margin-top: 5px; color: var(--el-text-color-secondary); font-size: 12px; line-height: 18px; }
+.advanced-settings { margin-bottom: 18px; }
+.output-list { display: grid; gap: 8px; }
+.output-item { display: flex; align-items: baseline; gap: 12px; color: var(--el-text-color-regular); font-size: 13px; }
+.output-item code { min-width: 92px; color: var(--el-color-primary); }
+</style>
diff --git a/src/views/flow/components/params/MediaAnalysisParams.vue b/src/views/flow/components/params/MediaAnalysisParams.vue
index e8674c8..d3fd061 100644
--- a/src/views/flow/components/params/MediaAnalysisParams.vue
+++ b/src/views/flow/components/params/MediaAnalysisParams.vue
@@ -100,6 +100,16 @@
         <div class="field-hint">{{ modeDescription }}</div>
       </el-form-item>
 
+      <el-form-item v-if="isVideo" label="输出内容">
+        <el-switch
+          v-model="formData.detailedOutput"
+          inline-prompt
+          active-text="详细"
+          inactive-text="精简"
+        />
+        <div class="field-hint">精简结果响应更快；需要逐步事件和时间范围时再启用详细输出。</div>
+      </el-form-item>
+
       <!--
         高级参数只对视频抽帧分析生效。默认折叠以降低普通用户的配置负担，isTuned 用于提示
         当前值已经偏离模式预设；“恢复默认”只重置当前模式的参数，不切换分析模式。
@@ -189,6 +199,7 @@ const formData = reactive({
   mediaUrl: '',
   instruction: '',
   analysisMode: 'AUTO',
+  detailedOutput: false,
   analysisTuning: { ...analysisPresets.AUTO },
 })
 
@@ -245,6 +256,7 @@ const profileOptions = computed(() => isVideo.value
  */
 const outputSummary = computed(() => isVideo.value
   ? [
+      { name: 'result', desc: '分析结果' },
       { name: 'passed', desc: '是否通过' },
       { name: 'confidence', desc: '判定置信度' },
       { name: 'evidenceSufficient', desc: '证据是否充分' },
@@ -253,6 +265,7 @@ const outputSummary = computed(() => isVideo.value
       { name: 'events', desc: '关键事件列表' },
       { name: 'analysisMode', desc: '实际使用的分析模式' },
       { name: 'fallback', desc: '是否切换到精确分析' },
+      { name: 'detailedOutput', desc: '是否返回详细事件证据' },
     ]
   : [
       { name: 'passed', desc: '是否检测到目标声音' },
@@ -379,6 +392,7 @@ const initData = () => {
   const media = params.find(item => item.name === mediaName.value)
   const instruction = params.find(item => item.name === 'instruction')
   const analysisMode = params.find(item => item.name === 'analysisMode')
+  const detailedOutput = params.find(item => item.name === 'detailedOutput')
   const analysisTuning = params.find(item => item.name === 'analysisTuning')
   formData.profileCode = profile?.input || profileOptions.value[0]?.value || ''
   formData.expectedLabel = expectedLabel?.input || ''
@@ -392,6 +406,7 @@ const initData = () => {
   formData.mediaUrl = media?.input || ''
   formData.instruction = instruction?.input || ''
   formData.analysisMode = analysisPresets[analysisMode?.input] ? analysisMode.input : 'AUTO'
+  formData.detailedOutput = detailedOutput?.input === true || detailedOutput?.input === 'true'
   const preset = analysisPresets[formData.analysisMode] || analysisPresets.AUTO
   Object.assign(formData.analysisTuning, normalizeStoredTuning(analysisTuning?.input, preset))
   if (formData.analysisMode !== 'AUTO') formData.analysisTuning.fallbackToAccurate = false
@@ -474,6 +489,14 @@ const validateAndSave = async () => {
         componentType: 'select',
         selectOptions: analysisModes,
       })
+      nodeParams.push({
+        name: 'detailedOutput',
+        type: 'input',
+        input: formData.detailedOutput,
+        required: false,
+        disabled: true,
+        componentType: 'switch',
+      })
       nodeParams.push({
         name: 'analysisTuning',
         type: 'input',
diff --git a/src/views/flow/config.js b/src/views/flow/config.js
index cfdda8d..8613001 100644
--- a/src/views/flow/config.js
+++ b/src/views/flow/config.js
@@ -556,10 +556,12 @@ function handler(params) {
           { name: "videoUrl", type: "quote", input: "", quote: "", required: true, disabled: true },
           { name: "instruction", type: "input", input: "", required: true, disabled: true },
           { name: "analysisMode", type: "input", input: "AUTO", required: false, disabled: true },
+          { name: "detailedOutput", type: "input", input: false, required: false, disabled: true, componentType: "switch" },
           { name: "analysisTuning", type: "input", input: { sampleFps: 2, maxWidth: 960, confidenceThreshold: 0.75, fallbackToAccurate: true }, required: false, disabled: true },
         ],
         outputType: "json",
         outputParams: [
+          { name: "result", type: "string", desc: "分析结果", disabled: true },
           { name: "passed", type: "boolean", desc: "是否通过", disabled: true },
           { name: "confidence", type: "number", desc: "判定置信度", disabled: true },
           { name: "evidenceSufficient", type: "boolean", desc: "证据是否充分", disabled: true },
@@ -569,6 +571,27 @@ function handler(params) {
           { name: "events", type: "array<object>", desc: "关键事件", children: [], disabled: true },
           { name: "analysisMode", type: "string", desc: "实际分析模式", disabled: true },
           { name: "fallback", type: "boolean", desc: "是否切换精确分析", disabled: true },
+          { name: "detailedOutput", type: "boolean", desc: "是否返回详细事件证据", disabled: true },
+        ],
+      },
+      {
+        icon: cameraSvg,
+        name: "通用图片分析",
+        type: "imageAnalysis",
+        desc: "按照提示词分析一张或多张图片并返回指定内容",
+        action: 'IMAGE_ANALYZE',
+        nodeType: "LLM",
+        nodeParams: [
+          { name: "profileCode", type: "input", input: "common.image_analysis.v1", required: true, disabled: true },
+          { name: "imageUrl", type: "quote", input: "", quote: "", required: true, disabled: true },
+          { name: "prompt", type: "input", input: "", required: true, disabled: true, componentType: "textarea" },
+          { name: "analysisMode", type: "input", input: "FAST", required: false, disabled: true },
+          { name: "analysisTuning", type: "input", input: { maxWidth: 896, maxImages: 12, maxOutputTokens: 384 }, required: false, disabled: true },
+        ],
+        outputType: "json",
+        outputParams: [
+          { name: "result", type: "object", desc: "按提示词返回的结果", children: [], disabled: true },
+          { name: "resultText", type: "string", desc: "模型返回的原始文本", disabled: true },
         ],
       },
       {
diff --git a/src/views/flow/index.vue b/src/views/flow/index.vue
index 7b532df..5aa5dd3 100644
--- a/src/views/flow/index.vue
+++ b/src/views/flow/index.vue
@@ -24,9 +24,9 @@
             <el-button
               type="warning"
               plain
-              :loading="recoverSafetyLoading"
+              :loading="restoreOperationalLoading"
               :disabled="flowState === 'testRunning'"
-              @click="openRecoverSafetyDialog"
+              @click="openRestoreOperationalDialog"
             >
               <el-icon><Warning /></el-icon>清除错误状态
             </el-button>
@@ -172,21 +172,21 @@
     />
   </div>
 
-  <el-dialog v-model="recoverSafetyDialogVisible" title="清除错误状态" width="480" append-to-body>
+  <el-dialog v-model="restoreOperationalDialogVisible" title="恢复运行状态" width="480" append-to-body>
     <el-form label-position="top">
       <el-form-item label="机器人">
         <el-select
-          v-model="recoverSafetyRobotIds"
+          v-model="restoreOperationalRobotIds"
           multiple
           filterable
           collapse-tags
           collapse-tags-tooltip
-          :loading="recoverRobotLoading"
+          :loading="restoreRobotLoading"
           placeholder="请选择需要恢复的在线机器人"
           style="width: 100%"
         >
           <el-option
-            v-for="item in recoverRobotOptions"
+            v-for="item in restoreRobotOptions"
             :key="item.value"
             :label="item.label"
             :value="item.value"
@@ -195,13 +195,13 @@
       </el-form-item>
     </el-form>
     <template #footer>
-      <el-button @click="recoverSafetyDialogVisible = false">取消</el-button>
+      <el-button @click="restoreOperationalDialogVisible = false">取消</el-button>
       <el-button
         type="warning"
-        :loading="recoverSafetyLoading"
-        :disabled="recoverSafetyRobotIds.length === 0"
-        @click="confirmRecoverSafetyState"
-      >确认清除</el-button>
+        :loading="restoreOperationalLoading"
+        :disabled="restoreOperationalRobotIds.length === 0"
+        @click="confirmRestoreOperationalState"
+      >恢复运行状态</el-button>
     </template>
   </el-dialog>
 
@@ -249,7 +249,7 @@
             <el-input
               :disabled="field.param?.disabled || false"
               v-model="field.param.name"
-              placeholder="请输入"
+              :placeholder="field.param.placeholder || '请输入'"
               clearable
             />
           </el-form-item>
@@ -280,6 +280,13 @@
                 :value="item.value"
               />
             </el-select>
+            <el-switch
+              v-model="field.param.input"
+              v-else-if="field.param.componentType === 'switch'"
+              inline-prompt
+              active-text="是"
+              inactive-text="否"
+            />
             <el-input
               v-else
               v-model="field.param.input"
@@ -350,7 +357,7 @@ import {
 } from "@/api/flow/flow";
 import { getDetect } from "@/api/test/detect";
 import { getRobotList, getRobotDevicesByRobotId } from '@/api/inspection/robot'
-import { recoverSafetyState } from '@/api/device/system'
+import { restoreOperationalState } from '@/api/device/system'
 import { getDeviceKindForAction, toDeviceOptions } from '@/utils/robotDevice'
 import {
   CircleClose,
@@ -377,53 +384,53 @@ const queueDropTargetId = ref("");
 const { onFlowEvent } = useFlowEvents();
 window.lf = lf;
 
-const recoverSafetyDialogVisible = ref(false);
-const recoverRobotLoading = ref(false);
-const recoverSafetyLoading = ref(false);
-const recoverRobotOptions = ref([]);
-const recoverSafetyRobotIds = ref([]);
+const restoreOperationalDialogVisible = ref(false);
+const restoreRobotLoading = ref(false);
+const restoreOperationalLoading = ref(false);
+const restoreRobotOptions = ref([]);
+const restoreOperationalRobotIds = ref([]);
 
-const openRecoverSafetyDialog = async () => {
-  recoverSafetyDialogVisible.value = true;
-  recoverRobotLoading.value = true;
+const openRestoreOperationalDialog = async () => {
+  restoreOperationalDialogVisible.value = true;
+  restoreRobotLoading.value = true;
   try {
     const response = await getRobotList({ pageNum: 1, pageSize: 1000, connectStatus: '1' });
-    recoverRobotOptions.value = (response.rows || []).filter((robot) => robot.robotId).map((robot) => ({
+    restoreRobotOptions.value = (response.rows || []).filter((robot) => robot.robotId).map((robot) => ({
       value: robot.robotId,
       label: `${robot.robotName || robot.robotId} (${robot.robotId})`,
     }));
-    recoverSafetyRobotIds.value = recoverSafetyRobotIds.value.filter((robotId) => (
-      recoverRobotOptions.value.some((option) => option.value === robotId)
+    restoreOperationalRobotIds.value = restoreOperationalRobotIds.value.filter((robotId) => (
+      restoreRobotOptions.value.some((option) => option.value === robotId)
     ));
-    if (!recoverSafetyRobotIds.value.length && recoverRobotOptions.value.length === 1) {
-      recoverSafetyRobotIds.value = [recoverRobotOptions.value[0].value];
+    if (!restoreOperationalRobotIds.value.length && restoreRobotOptions.value.length === 1) {
+      restoreOperationalRobotIds.value = [restoreRobotOptions.value[0].value];
     }
   } catch (error) {
     console.error("加载在线机器人失败:", error);
-    recoverRobotOptions.value = [];
+    restoreRobotOptions.value = [];
   } finally {
-    recoverRobotLoading.value = false;
+    restoreRobotLoading.value = false;
   }
 };
 
-const confirmRecoverSafetyState = async () => {
-  if (!recoverSafetyRobotIds.value.length) {
+const confirmRestoreOperationalState = async () => {
+  if (!restoreOperationalRobotIds.value.length) {
     ElMessage.warning("请选择需要恢复的机器人");
     return;
   }
   try {
     await ElMessageBox.confirm(
-      `将清除 ${recoverSafetyRobotIds.value.length} 台机器人的可恢复错误状态，确认继续？`,
+      `将恢复 ${restoreOperationalRobotIds.value.length} 台机器人的运行状态，确认继续？`,
       "安全状态确认",
       { type: "warning", confirmButtonText: "确认清除", cancelButtonText: "取消" },
     );
   } catch {
     return;
   }
-  recoverSafetyLoading.value = true;
-  const robotIds = [...recoverSafetyRobotIds.value];
+  restoreOperationalLoading.value = true;
+  const robotIds = [...restoreOperationalRobotIds.value];
   try {
-    const results = await Promise.allSettled(robotIds.map((robotId) => recoverSafetyState({
+    const results = await Promise.allSettled(robotIds.map((robotId) => restoreOperationalState({
       robotId,
       reason: "工作流设计器手动清除错误状态",
       timeoutMs: 15000,
@@ -432,18 +439,18 @@ const confirmRecoverSafetyState = async () => {
       result.status === "rejected" ? [robotIds[index]] : []
     ));
     if (!failedRobotIds.length) {
-      ElMessage.success(`已清除 ${robotIds.length} 台机器人的可恢复错误状态`);
-      recoverSafetyDialogVisible.value = false;
-      recoverSafetyRobotIds.value = [];
+      ElMessage.success(`已恢复 ${robotIds.length} 台机器人的运行状态`);
+      restoreOperationalDialogVisible.value = false;
+      restoreOperationalRobotIds.value = [];
     } else {
       const successCount = robotIds.length - failedRobotIds.length;
-      recoverSafetyRobotIds.value = failedRobotIds;
+      restoreOperationalRobotIds.value = failedRobotIds;
       ElMessage.error(successCount
         ? `${successCount} 台成功，${failedRobotIds.length} 台失败，请重试失败项`
         : "清除错误状态失败，请检查机器人状态");
     }
   } finally {
-    recoverSafetyLoading.value = false;
+    restoreOperationalLoading.value = false;
   }
 };
 
@@ -1257,6 +1264,46 @@ const buildSingleNodePayload = (params) => {
   });
   return payload;
 };
+const normalizeImageAnalysisSingleNodeParams = (params) => {
+  const normalized = params || [];
+  const findValue = (name) => normalized.find((param) => param?.name === name)?.input;
+  if (!findValue('prompt')) {
+    const target = String(findValue('targetDescription') || '').trim();
+    const instruction = String(findValue('instruction') || '').trim();
+    const taskType = String(findValue('taskType') || '').toUpperCase();
+    const prompt = [target || instruction];
+    if (taskType === 'ICON_TEMPLATE_LOCATE') {
+      prompt.push('最后一张图片是参考图片，请在其他图片中查找对应目标并返回坐标。');
+    }
+    prompt.push('只返回分析结果，不输出通过、未通过或置信度。');
+    normalized.push({
+      name: 'prompt', type: 'input', input: prompt.filter(Boolean).join('\n'),
+      componentType: 'textarea', required: true, disabled: true,
+    });
+  }
+  for (let index = normalized.length - 1; index >= 0; index -= 1) {
+    if (['taskType', 'targetDescription', 'instruction'].includes(normalized[index]?.name)) {
+      normalized.splice(index, 1);
+    }
+  }
+  normalized.forEach((param) => {
+    if (!['imageUrl', 'referenceImageUrl'].includes(param?.name)) return;
+    param.type = 'input';
+    param.componentType = 'textarea';
+    param.input = Array.isArray(param.input) ? param.input.join('\n') : (param.input || '');
+    param.quote = '';
+    if (param.name === 'imageUrl') param.placeholder = '每行一个图片地址';
+  });
+  return normalized;
+};
+const buildImageAnalysisSingleNodePayload = (params) => {
+  const payload = buildSingleNodePayload(params);
+  payload.imageUrl = String(payload.imageUrl || '')
+    .split(/\r?\n|,/)
+    .map((item) => item.trim())
+    .filter(Boolean);
+  return payload;
+};
 const validateSingleNodeParams = () => {
   for (const { param } of singleNodeFields.value) {
     if (!String(param?.name || "").trim()) {
@@ -1430,7 +1477,9 @@ const singleNodeExecution = async (e) => {
       await handleSingleNodeRobotChange(cachedRobotId);
     }
   } else {
-    formData.nodeParams = executionParams;
+    formData.nodeParams = normalizedAction === 'IMAGE_ANALYZE'
+      ? normalizeImageAnalysisSingleNodeParams(executionParams)
+      : executionParams;
   }
   if (nodeAction.value === "AGV_MOVE_TO_STATION") {
     const stationParam = formData.nodeParams.find((item) => item.name === "stationId");
@@ -1476,7 +1525,9 @@ const execute = async () => {
   try {
     const res = await flowAction({
       action: nodeAction.value,
-      payload: buildSingleNodePayload(formData.nodeParams),
+      payload: nodeAction.value === 'IMAGE_ANALYZE'
+        ? buildImageAnalysisSingleNodePayload(formData.nodeParams)
+        : buildSingleNodePayload(formData.nodeParams),
     });
     if (res.code === 200) {
       ElMessage.success("执行成功");
diff --git a/src/views/intelligenceTest/runningTask/task-log.vue b/src/views/intelligenceTest/runningTask/task-log.vue
index 8090d07..7f1fe58 100644
--- a/src/views/intelligenceTest/runningTask/task-log.vue
+++ b/src/views/intelligenceTest/runningTask/task-log.vue
@@ -290,6 +290,8 @@ const analysisVerdictOf = (log) => {
   const testStatus = String(log?.testStatus ?? '')
   const analysisType = String(output.analysisType || '').toUpperCase()
   const profileCode = String(output.profileCode || '').toLowerCase()
+  // 通用图片分析只返回提示词要求的内容，不承担检测项通过/不通过判定。
+  if (analysisType.includes('IMAGE') || profileCode.includes('image')) return null
   const hasAnalysis = Boolean(
     analysisType.includes('AUDIO') || analysisType.includes('VIDEO')
     || profileCode.includes('audio') || profileCode.includes('video')