feat(note): 添加视频理解功能- 在 GPT 模型中增加 video_img_urls 字段用于存储视频截图

- 在笔记生成请求中添加视频理解相关参数
- 实现视频截图功能，支持按指定间隔生成截图
- 更新笔记生成逻辑，支持视频理解功能- 在前端服务中添加视频理解相关参数

BillNote_frontend/src/pages/HomePage/components/NoteForm.tsx

@@ -1,238 +1,189 @@
+/* NoteForm.tsx ---------------------------------------------------- */
 import {
-  Form,
-  FormControl,
-  FormField,
-  FormItem,
-  FormLabel,
-  FormMessage,
+  Form, FormControl, FormField, FormItem, FormLabel, FormMessage,
 } from '@/components/ui/form.tsx'
 import { useEffect } from 'react'
-import { Input } from '@/components/ui/input.tsx'
-import {
-  Select,
-  SelectContent,
-  SelectItem,
-  SelectTrigger,
-  SelectValue,
-} from '@/components/ui/select.tsx'
-import { Button } from '@/components/ui/button.tsx'
-import { Checkbox } from '@/components/ui/checkbox.tsx'
-import { useForm } from 'react-hook-form'
-import { z } from 'zod'
+import { useForm, useWatch } from 'react-hook-form'
 import { zodResolver } from '@hookform/resolvers/zod'
-import { Info, Clock, Loader2 } from 'lucide-react'
-import { message } from 'antd'
-import {
-  Tooltip,
-  TooltipContent,
-  TooltipProvider,
-  TooltipTrigger,
-} from '@/components/ui/tooltip.tsx'
+import { z } from 'zod'
+
+import { Info, Loader2 } from 'lucide-react'
+import { message, Alert } from 'antd'
+
 import { generateNote } from '@/services/note.ts'
-import { useTaskStore } from '@/store/taskStore'
-import NoteHistory from '@/pages/HomePage/components/NoteHistory.tsx'
-import { useModelStore } from '@/store/modelStore'
-import { Alert } from 'antd'
-import { Textarea } from '@/components/ui/textarea.tsx'
-import { ScrollArea } from '@/components/ui/scroll-area.tsx'
 import { uploadFile } from '@/services/upload.ts'
-// ✅ 定义表单 schema
-const formSchema = z
-  .object({
+import { useTaskStore } from '@/store/taskStore'
+import { useModelStore } from '@/store/modelStore'
+import {Tooltip, TooltipContent, TooltipProvider, TooltipTrigger} from "@/components/ui/tooltip.tsx";
+import {Checkbox} from "@/components/ui/checkbox.tsx";
+import {ScrollArea} from "@/components/ui/scroll-area.tsx";
+import {Button} from "@/components/ui/button.tsx";
+import {Select, SelectContent, SelectItem, SelectTrigger, SelectValue} from "@/components/ui/select.tsx";
+import {Input} from "@/components/ui/input.tsx";
+import {Textarea} from "@/components/ui/textarea.tsx";
+
+/* -------------------- 常量 -------------------- */
+const noteFormats = [
+  { label: '目录', value: 'toc' },
+  { label: '原片跳转', value: 'link' },
+  { label: '原片截图', value: 'screenshot' },
+  { label: 'AI总结', value: 'summary' },
+] as const
+
+const noteStyles = [
+  { label: '精简', value: 'minimal' },
+  { label: '详细', value: 'detailed' },
+  { label: '教程', value: 'tutorial' },
+  { label: '学术', value: 'academic' },
+  { label: '小红书', value: 'xiaohongshu' },
+  { label: '生活向', value: 'life_journal' },
+  { label: '任务导向', value: 'task_oriented' },
+  { label: '商业风格', value: 'business' },
+  { label: '会议纪要', value: 'meeting_minutes' },
+] as const
+
+/* -------------------- 校验 Schema -------------------- */
+const formSchema = z.object({
   video_url: z.string(),
   platform: z.string().nonempty('请选择平台'),
-    quality: z.enum(['fast', 'medium', 'slow'], {
-      required_error: '请选择音频质量',
-    }),
+  quality: z.enum(['fast', 'medium', 'slow']),
   screenshot: z.boolean().optional(),
   link: z.boolean().optional(),
   model_name: z.string().nonempty('请选择模型'),
   format: z.array(z.string()).default([]),
   style: z.string().nonempty('请选择笔记生成风格'),
   extras: z.string().optional(),
-  })
-  .superRefine((data, ctx) => {
-    const { video_url, platform } = data
-
+  video_understanding: z.boolean().optional(),
+  video_interval: z.coerce.number().min(1).max(30).default(4).optional(),
+  grid_size: z.tuple([
+    z.coerce.number().min(1).max(10),
+    z.coerce.number().min(1).max(10),
+  ]).default([3, 3]).optional(),
+}).superRefine(({ video_url, platform }, ctx) => {
   if (platform === 'local' || platform === 'douyin') {
-      if (!video_url || typeof video_url !== 'string') {
-        ctx.addIssue({
-          code: z.ZodIssueCode.custom,
-          message: '本地视频路径不能为空',
-          path: ['video_url'],
-        })
+    if (!video_url) {
+      ctx.addIssue({ code: 'custom', message: '本地视频路径不能为空', path: ['video_url'] })
     }
   } else {
     try {
       const url = new URL(video_url)
-        if (!(url.protocol === 'http:' || url.protocol === 'https:')) {
-          throw new Error()
-        }
+      if (!['http:', 'https:'].includes(url.protocol)) throw new Error()
     } catch {
-        ctx.addIssue({
-          code: z.ZodIssueCode.custom,
-          message: '请输入正确的视频链接',
-          path: ['video_url'],
-        })
+      ctx.addIssue({ code: 'custom', message: '请输入正确的视频链接', path: ['video_url'] })
     }
   }
-  })
+})
 
 type NoteFormValues = z.infer<typeof formSchema>
-const noteFormats = [
-  {
-    label: '目录',
-    value: 'toc',
-  },
-  { label: '原片跳转', value: 'link' },
-  { label: '原片截图', value: 'screenshot' },
-  { label: 'AI总结', value: 'summary' },
-]
-const noteStyles = [
-  {
-    label: '精简',
-    value: 'minimal', // 简洁、快速呈现要点
-  },
-  {
-    label: '详细',
-    value: 'detailed', // 详细记录，包含时间戳、关键点
-  },
-  {
-    label: '教程',
-    value: 'tutorial', // 详细记录，包含时间戳、关键点
-  },
-  {
-    label: '学术',
-    value: 'academic', // 适合学术报告，正式且结构化
-  },
-  {
-    label: '小红书',
-    value: 'xiaohongshu', // 适合社交平台分享，亲切、口语化
-  },
-  {
-    label: '生活向',
-    value: 'life_journal', // 记录个人生活感悟，情感化表达
-  },
-  {
-    label: '任务导向',
-    value: 'task_oriented', // 强调任务、目标，适合工作和待办事项
-  },
-  {
-    label: '商业风格',
-    value: 'business', // 适合商业报告、会议纪要，正式且精准
-  },
-  {
-    label: '会议纪要',
-    value: 'meeting_minutes', // 适合商业报告、会议纪要，正式且精准
-  },
-]
 
-const NoteForm = () => {
-  useTaskStore(state => state.tasks)
-  const setCurrentTask = useTaskStore(state => state.setCurrentTask)
-  const currentTaskId = useTaskStore(state => state.currentTaskId)
-  const getCurrentTask = useTaskStore(state => state.getCurrentTask)
-  const loadEnabledModels = useModelStore(state => state.loadEnabledModels)
-  const modelList = useModelStore(state => state.modelList)
-  const showFeatureHint = useModelStore(state => state.showFeatureHint)
-  const setShowFeatureHint = useModelStore(state => state.setShowFeatureHint)
-  const form = useForm<NoteFormValues>({
-    resolver: zodResolver(formSchema),
-    defaultValues: {
-      video_url: '',
-      platform: 'bilibili',
-      quality: 'medium', // 默认中等质量
-      screenshot: false,
-      model_name: modelList[0]?.model_name || '', // 确保有值
-      format: [], // 初始化为空数组
-      style: 'minimal', // 默认选择精简风格
-      extras: '', // 初始化为空字符串
-    },
-  })
-  const platform = form.watch('platform')
-
-  const onClose = () => {
-    setShowFeatureHint(false)
-  }
-  const isGenerating = () => {
-    console.log('🚀 isGenerating', getCurrentTask()?.status)
-    return (
-      getCurrentTask()?.status != 'SUCCESS' &&
-      getCurrentTask()?.status != 'FAILED' &&
-      getCurrentTask()?.status != undefined
-    )
-  }
-  const handleFileUpload = async (file: File, onSuccess: (url: string) => void) => {
-    if (!file) return
-    const formData = new FormData()
-    formData.append('file', file)
-
-    try {
-      const res = await uploadFile(formData)
-      if (res.data.code === 0) {
-        const uploadedUrl = res.data.data.url
-        console.log('✅ 上传成功', uploadedUrl)
-
-        onSuccess(uploadedUrl)
-      }
-    } catch (error) {
-      console.error('上传失败', error)
-      // 可以弹个 toast 或者提示上传失败
-    }
-  }
-  // TODO 修复选择其他视频平台以后再选择本地视频还可以选择 Link 的问题
-  const onSubmit = async (data: NoteFormValues) => {
-    console.log('🎯 提交内容：', data)
-    message.success('提交任务')
-
-    const payload = {
-      video_url: data.video_url,
-      platform: data.platform,
-      quality: data.quality,
-      model_name: data.model_name,
-      provider_id: modelList.find(model => model.model_name === data.model_name).provider_id,
-      format: data.format,
-      style: data.style,
-      extras: data.extras,
-    }
-    const res = await generateNote(payload)
-    const taskId = res.data.task_id
-    useTaskStore.getState().addPendingTask(taskId, data.platform, payload)
-  }
-  useEffect(() => {
-    loadEnabledModels()
-  }, [])
-
-  return (
-    <>
-      <ScrollArea className="sm:h-[400px] md:h-[800px]">
-        <Form {...form}>
-          <form onSubmit={form.handleSubmit(onSubmit)} className="space-y-2">
-            <div className="flex w-full items-center gap-2 py-1.5">
-              <Button
-                type="submit"
-                className="bg-primary w-full sm:w-full"
-                disabled={isGenerating()}
-              >
-                {isGenerating() && <Loader2 className="mr-2 h-4 w-4 animate-spin" />}
-                {isGenerating() ? '正在生成…' : '生成笔记'}
-              </Button>
-            </div>
-            <div className="space-y-2">
+/* -------------------- 可复用子组件 -------------------- */
+const SectionHeader = ({ title, tip }: { title: string; tip?: string }) => (
     <div className="my-3 flex items-center justify-between">
-                <h2 className="block">视频链接</h2>
+      <h2 className="block">{title}</h2>
+      {tip && (
           <TooltipProvider>
             <Tooltip>
               <TooltipTrigger asChild>
                 <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />
               </TooltipTrigger>
-                    <TooltipContent>
-                      <p className="text-xs">输入视频链接，支持哔哩哔哩、YouTube等平台</p>
-                    </TooltipContent>
+              <TooltipContent className="text-xs">{tip}</TooltipContent>
             </Tooltip>
           </TooltipProvider>
+      )}
     </div>
+)
 
+const CheckboxGroup = ({
+                         value = [], onChange, disabledMap,
+                       }: {
+  value?: string[]
+  onChange: (v: string[]) => void
+  disabledMap: Record<string, boolean>
+}) => (
+    <div className="flex flex-wrap space-x-1.5">
+      {noteFormats.map(({ label, value: v }) => (
+          <label key={v} className="flex items-center space-x-2">
+            <Checkbox
+                checked={value.includes(v)}
+                disabled={disabledMap[v]}
+                onCheckedChange={checked =>
+                    onChange(checked ? [...value, v] : value.filter(x => x !== v))}
+            />
+            <span>{label}</span>
+          </label>
+      ))}
+    </div>
+)
+
+/* -------------------- 主组件 -------------------- */
+const NoteForm = () => {
+  /* ---- 全局状态 ---- */
+  const { addPendingTask, currentTaskId, getCurrentTask } = useTaskStore()
+  const { loadEnabledModels, modelList, showFeatureHint, setShowFeatureHint } = useModelStore()
+
+  /* ---- 表单 ---- */
+  const form = useForm<NoteFormValues>({
+    resolver: zodResolver(formSchema),
+    defaultValues: {
+      platform: 'bilibili',
+      quality: 'medium',
+      model_name: modelList[0]?.model_name || '',
+      style: 'minimal',
+      video_interval: 4,
+      grid_size: [3, 3],
+      format: [],
+    },
+  })
+
+  /* ---- 派生状态（只 watch 一次，提高性能） ---- */
+  const platform = useWatch({ control: form.control, name: 'platform' }) as string
+  const videoUnderstandingEnabled = useWatch({ control: form.control, name: 'video_understanding' })
+
+  /* ---- 副作用 ---- */
+  useEffect(() => {
+    loadEnabledModels()
+
+  return}, [])
+
+  /* ---- 帮助函数 ---- */
+  const isGenerating = () =>
+      !['SUCCESS', 'FAILED', undefined].includes(getCurrentTask()?.status)
+
+  const handleFileUpload = async (file: File, cb: (url: string) => void) => {
+    const formData = new FormData()
+    formData.append('file', file)
+    try {
+      const { data } = await uploadFile(formData)
+      if (data.code === 0) cb(data.data.url)
+    } catch (err) {
+      console.error('上传失败:', err)
+      message.error('上传失败，请重试')
+    }
+  }
+
+  const onSubmit = async (values: NoteFormValues) => {
+    const payload:NoteFormValues = {
+      ...values,
+      provider_id: modelList.find(m => m.model_name === values.model_name)!.provider_id,
+    }
+    message.success('已提交任务')
+    const { data } = await generateNote(payload)
+    addPendingTask(data.task_id, values.platform, payload)
+  }
+
+  /* -------------------- 渲染 -------------------- */
+  return (
+      <ScrollArea className="sm:h-[400px] md:h-[800px]">
+        <Form {...form}>
+          <form onSubmit={form.handleSubmit(onSubmit)} className="space-y-4">
+            {/* 顶部按钮 */}
+            <Button type="submit" className="w-full bg-primary" disabled={isGenerating()}>
+              {isGenerating() && <Loader2 className="mr-2 h-4 w-4 animate-spin" />}
+              {isGenerating() ? '正在生成…' : '生成笔记'}
+            </Button>
+
+            {/* 视频链接 & 平台 */}
+            <SectionHeader title="视频链接" tip="支持 B 站、YouTube 等平台" />
             <div className="flex gap-2">
               {/* 平台选择 */}
               <FormField
@@ -242,13 +193,11 @@ const NoteForm = () => {
                       <FormItem>
                         <Select onValueChange={field.onChange} defaultValue={field.value}>
                           <FormControl>
-                          <SelectTrigger className="w-32">
-                            <SelectValue placeholder="选择平台" />
-                          </SelectTrigger>
+                            <SelectTrigger className="w-32"><SelectValue /></SelectTrigger>
                           </FormControl>
                           <SelectContent>
                             <SelectItem value="bilibili">哔哩哔哩</SelectItem>
-                          <SelectItem value="youtube">Youtube</SelectItem>
+                            <SelectItem value="youtube">YouTube</SelectItem>
                             <SelectItem value="douyin">抖音</SelectItem>
                             <SelectItem value="local">本地视频</SelectItem>
                           </SelectContent>
@@ -257,274 +206,198 @@ const NoteForm = () => {
                       </FormItem>
                   )}
               />
-
-                {/* 视频地址 */}
+              {/* 链接输入 / 上传框 */}
               <FormField
                   control={form.control}
                   name="video_url"
                   render={({ field }) => (
                       <FormItem className="flex-1">
-                      <FormControl>
-                        {form.watch('platform') === 'local' ? (
-                          <div className="flex flex-col gap-2">
-                            {/* 第一行：本地路径输入框 */}
-                            <Input placeholder="请输入本地视频路径" {...field} className="w-full" />
-                          </div>
-                        ) : (
-                          <Input placeholder="请输入视频网站链接" {...field} />
-                        )}
-                      </FormControl>
-                      <FormMessage />
-                    </FormItem>
-                  )}
-                />
-              </div>
-
-              <FormField
-                control={form.control}
-                name="video_url"
-                render={({ field }) => (
-                  <FormItem className="w-full">
-                    <FormControl>
-                      {form.watch('platform') === 'local' ? (
+                        {platform === 'local' ? (
+                            <>
+                              <Input placeholder="请输入本地视频路径" {...field} />
                               <div
-                          className="hover:border-primary flex h-40 w-full cursor-pointer items-center justify-center rounded-md border-2 border-dashed border-gray-300 transition-colors"
-                          onDragOver={e => {
-                            e.preventDefault()
-                            e.stopPropagation()
-                          }}
+                                  className="hover:border-primary mt-2 flex h-40 cursor-pointer items-center justify-center rounded-md border-2 border-dashed border-gray-300 transition-colors"
+                                  onDragOver={e => { e.preventDefault(); e.stopPropagation() }}
                                   onDrop={e => {
                                     e.preventDefault()
-                            e.stopPropagation()
                                     const file = e.dataTransfer.files?.[0]
-                            if (file) {
-                              handleFileUpload(file, uploadedUrl => {
-                                field.onChange(uploadedUrl)
-                              })
-                            }
+                                    if (file) handleFileUpload(file, field.onChange)
                                   }}
                                   onClick={() => {
                                     const input = document.createElement('input')
                                     input.type = 'file'
                                     input.accept = 'video/*'
-                            input.onchange = (e: any) => {
-                              const file = e.target.files?.[0]
-                              if (file) {
-                                handleFileUpload(file, uploadedUrl => {
-                                  field.onChange(uploadedUrl)
-                                })
-                              }
+                                    input.onchange = e => {
+                                      const file = (e.target as HTMLInputElement).files?.[0]
+                                      if (file) handleFileUpload(file, field.onChange)
                                     }
                                     input.click()
                                   }}
                               >
-                          <div className="text-center text-sm text-gray-500">
-                            <p className="mb-2">拖拽文件到这里上传</p>
-                            <p className="text-xs text-gray-400">或点击选择文件</p>
-                          </div>
+                                <p className="text-center text-sm text-gray-500">
+                                  拖拽文件到这里上传 <br />
+                                  <span className="text-xs text-gray-400">或点击选择文件</span>
+                                </p>
                               </div>
+                            </>
                         ) : (
-                        <></>
+                            <Input placeholder="请输入视频网站链接" {...field} />
                         )}
-                    </FormControl>
-                    {/* ❗可以不要FormMessage，不然重复两次报错提示 */}
+                        <FormMessage />
                       </FormItem>
                   )}
               />
-              {/*<p className="text-xs text-neutral-500">*/}
-              {/*    支持哔哩哔哩视频链接，例如：*/}
-              {/*    https://www.bilibili.com/video/BV1vc25YQE9X/*/}
-              {/*</p>*/}
-              {/*<FormField*/}
-              {/*  control={form.control}*/}
-              {/*  name="quality"*/}
-              {/*  render={({ field }) => (*/}
-              {/*    <FormItem>*/}
-              {/*      <div className="my-3 flex items-center justify-between">*/}
-              {/*        <h2 className="block">音频质量</h2>*/}
-              {/*        <TooltipProvider>*/}
-              {/*          <Tooltip>*/}
-              {/*            <TooltipTrigger asChild>*/}
-              {/*              <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />*/}
-              {/*            </TooltipTrigger>*/}
-              {/*            <TooltipContent>*/}
-              {/*              <p className="max-w-[200px] text-xs">*/}
-              {/*                质量越高，下载体积越大，速度越慢*/}
-              {/*              </p>*/}
-              {/*            </TooltipContent>*/}
-              {/*          </Tooltip>*/}
-              {/*        </TooltipProvider>*/}
-              {/*      </div>*/}
-              {/*      <Select onValueChange={field.onChange} defaultValue={field.value}>*/}
-              {/*        <FormControl>*/}
-              {/*          <SelectTrigger className="w-full">*/}
-              {/*            <SelectValue placeholder="选择质量" />*/}
-              {/*          </SelectTrigger>*/}
-              {/*        </FormControl>*/}
-              {/*        <SelectContent>*/}
-              {/*          <SelectItem value="fast">快速（压缩）</SelectItem>*/}
-              {/*          <SelectItem value="medium">中等（推荐）</SelectItem>*/}
-              {/*          <SelectItem value="slow">高质量（清晰）</SelectItem>*/}
-              {/*        </SelectContent>*/}
-              {/*      </Select>*/}
-              {/*      /!*<FormDescription className="text-xs text-neutral-500">*!/*/}
-              {/*      /!*    质量越高，下载体积越大，速度越慢*!/*/}
-              {/*      /!*</FormDescription>*!/*/}
-              {/*      <FormMessage />*/}
-              {/*    </FormItem>*/}
-              {/*  )}*/}
-              {/*/>*/}
+            </div>
 
+            {/* 模型选择 */}
             <FormField
                 control={form.control}
                 name="model_name"
                 render={({ field }) => (
                     <FormItem>
-                    <div className="my-3 flex items-center justify-between">
-                      <h2 className="block">模型选择</h2>
-                      <TooltipProvider>
-                        <Tooltip>
-                          <TooltipTrigger asChild>
-                            <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />
-                          </TooltipTrigger>
-                          <TooltipContent>
-                            <p className="max-w-[200px] text-xs">
-                              不同模型返回质量不同，可自行测试
-                            </p>
-                          </TooltipContent>
-                        </Tooltip>
-                      </TooltipProvider>
-                    </div>
+                      <SectionHeader title="模型选择" tip="不同模型效果不同，建议自行测试" />
                       <Select onValueChange={field.onChange} defaultValue={field.value}>
                         <FormControl>
-                        <SelectTrigger className="w-full">
-                          <SelectValue placeholder="选择配置好的模型" />
-                        </SelectTrigger>
+                          <SelectTrigger className="w-full"><SelectValue /></SelectTrigger>
                         </FormControl>
                         <SelectContent>
-                        {modelList.map(item => {
-                          return <SelectItem value={item.model_name}>{item.model_name}</SelectItem>
-                        })}
+                          {modelList.map(m => (
+                              <SelectItem key={m.model_name} value={m.model_name}>
+                                {m.model_name}
+                              </SelectItem>
+                          ))}
                         </SelectContent>
                       </Select>
-                    {/*<FormDescription className="text-xs text-neutral-500">*/}
-                    {/*    质量越高，下载体积越大，速度越慢*/}
-                    {/*</FormDescription>*/}
+                      <FormMessage />
+                    </FormItem>
+                )}
+            />
+
+            {/* 视频理解 */}
+            <SectionHeader title="视频理解" tip="将视频截图发给多模态模型辅助分析" />
+            <div className="flex flex-col gap-2">
+
+              <FormField
+                  control={form.control}
+                  name="video_understanding"
+                  render={({ field }) => (
+                      <FormItem>
+                        <div  className="flex items-center gap-2">
+                        <FormLabel>启用</FormLabel>
+                        <Checkbox
+                            checked={videoUnderstandingEnabled}
+                            onCheckedChange={v => form.setValue('video_understanding', v)}
+                        />
+                        </div>
+                        <FormMessage />
+                      </FormItem>
+                  )}
+              />
+              <Alert
+                  type="info"
+                  message="推荐多模态模型：qwen2.5-vl-72b-instruct / gpt-4o"
+                  className="text-sm"
+              />
+              <div className="grid grid-cols-2 gap-4">
+                {/* 采样间隔 */}
+                <FormField
+                    control={form.control}
+                    name="video_interval"
+                    render={({ field }) => (
+                        <FormItem>
+                          <FormLabel>采样间隔（秒）</FormLabel>
+                          <Input
+                              disabled={!videoUnderstandingEnabled}
+                              type="number"
+                              {...field}
+                          />
+                          <FormMessage />
+                        </FormItem>
+                    )}
+                />
+                {/* 拼图大小 */}
+                <FormField
+                    control={form.control}
+                    name="grid_size"
+                    render={({ field }) => (
+                        <FormItem>
+                          <FormLabel>拼图尺寸（列 × 行）</FormLabel>
+                          <div className="flex items-center space-x-2">
+                            <Input
+                                disabled={!videoUnderstandingEnabled}
+                                type="number"
+                                value={field.value?.[0] || 3}
+                                onChange={e =>
+                                    field.onChange([+e.target.value, field.value?.[1] || 3])}
+                                className="w-16"
+                            />
+                            <span>x</span>
+                            <Input
+                                disabled={!videoUnderstandingEnabled}
+                                type="number"
+                                value={field.value?.[1] || 3}
+                                onChange={e =>
+                                    field.onChange([field.value?.[0] || 3, +e.target.value])}
+                                className="w-16"
+                            />
+                          </div>
                           <FormMessage />
                         </FormItem>
                     )}
                 />
               </div>
+            </div>
 
+            {/* 笔记风格 */}
             <FormField
                 control={form.control}
                 name="style"
                 render={({ field }) => (
                     <FormItem>
-                  <div className="my-3 flex items-center justify-between">
-                    <h2 className="block">笔记风格</h2>
-                    <TooltipProvider>
-                      <Tooltip>
-                        <TooltipTrigger asChild>
-                          <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />
-                        </TooltipTrigger>
-                        <TooltipContent>
-                          <p className="max-w-[200px] text-xs">选择你希望生成的笔记风格</p>
-                        </TooltipContent>
-                      </Tooltip>
-                    </TooltipProvider>
-                  </div>
-
+                      <SectionHeader title="笔记风格" tip="选择生成笔记的呈现风格" />
                       <Select onValueChange={field.onChange} defaultValue={field.value}>
                         <FormControl>
-                      <SelectTrigger className="w-full">
-                        <SelectValue placeholder="选择笔记风格" />
-                      </SelectTrigger>
+                          <SelectTrigger className="w-full"><SelectValue /></SelectTrigger>
                         </FormControl>
                         <SelectContent>
-                      {noteStyles.map(item => (
-                        <SelectItem key={item.value} value={item.value}>
-                          {item.label}
-                        </SelectItem>
+                          {noteStyles.map(({ label, value }) => (
+                              <SelectItem key={value} value={value}>{label}</SelectItem>
                           ))}
                         </SelectContent>
                       </Select>
-
                       <FormMessage />
                     </FormItem>
                 )}
             />
 
+            {/* 笔记格式 */}
             <FormField
                 control={form.control}
                 name="format"
                 render={({ field }) => (
                     <FormItem>
-                  <div className="my-3 flex items-center justify-between">
-                    <h2 className="block">笔记格式</h2>
-                    <TooltipProvider>
-                      <Tooltip>
-                        <TooltipTrigger asChild>
-                          <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />
-                        </TooltipTrigger>
-                        <TooltipContent>
-                          <p className="text-xs">
-                            选择要包含的笔记元素，比如时间戳、截图提示或总结
-                          </p>
-                        </TooltipContent>
-                      </Tooltip>
-                    </TooltipProvider>
-                  </div>
-
-                  <FormControl>
-                    <div className="flex flex-wrap space-x-1.5">
-                      {noteFormats.map(item => (
-                        <label key={item.value} className="flex items-center space-x-2">
-                          <Checkbox
-                            checked={field.value?.includes(item.value)}
-                            disabled={item.value === 'link' && platform === 'local'}
-                            onCheckedChange={checked => {
-                              const currentValue = field.value ?? [] // ✨ 保底是数组
-                              if (checked) {
-                                field.onChange([...currentValue, item.value])
-                              } else {
-                                field.onChange(currentValue.filter(v => v !== item.value))
-                              }
+                      <SectionHeader title="笔记格式" tip="选择要包含的笔记元素" />
+                      <CheckboxGroup
+                          value={field.value}
+                          onChange={field.onChange}
+                          disabledMap={{
+                            link: platform === 'local',
+                            screenshot: !videoUnderstandingEnabled,
                           }}
                       />
-                          <span>{item.label}</span>
-                        </label>
-                      ))}
-                    </div>
-                  </FormControl>
-                  <FormField
-                    control={form.control}
-                    name="extras"
-                    render={({ field }) => (
-                      <FormItem>
-                        <div className="my-3 flex items-center justify-between">
-                          <h2 className="block">备注</h2>
-                          <TooltipProvider>
-                            <Tooltip>
-                              <TooltipTrigger asChild>
-                                <Info className="hover:text-primary h-4 w-4 cursor-pointer text-neutral-400" />
-                              </TooltipTrigger>
-                              <TooltipContent>
-                                <p className="text-xs">会把这段加入到Prompt最后 可自行测试</p>
-                              </TooltipContent>
-                            </Tooltip>
-                          </TooltipProvider>
-                        </div>
-                        <Textarea placeholder={'笔记需要罗列出 xxx 关键点'} {...field} />
-
-                        {/*<FormDescription className="text-xs text-neutral-500">*/}
-                        {/*    质量越高，下载体积越大，速度越慢*/}
-                        {/*</FormDescription>*/}
                       <FormMessage />
                     </FormItem>
                 )}
             />
 
+            {/* 备注 */}
+            <FormField
+                control={form.control}
+                name="extras"
+                render={({ field }) => (
+                    <FormItem>
+                      <SectionHeader title="备注" tip="可在 Prompt 结尾附加自定义说明" />
+                      <Textarea placeholder="笔记需要罗列出 xxx 关键点…" {...field} />
                       <FormMessage />
                     </FormItem>
                 )}
@@ -532,11 +405,6 @@ const NoteForm = () => {
           </form>
         </Form>
       </ScrollArea>
-
-      {/* 添加一些额外的说明或功能介绍 */}
-
-      {/*<div className="bg-primary-light mt-6 rounded-lg p-4"></div>*/}
-    </>
   )
 }
 

BillNote_frontend/src/services/note.ts

@@ -12,6 +12,9 @@ export const generateNote = async (data: {
   format: Array<string>
   style: string
   extras?: string
+  video_understand?: boolean
+  video_interval?: number
+  grid_size:Array<number>
 }) => {
   try {
     const response = await request.post('/generate_note', data)

backend/app/gpt/prompt_builder.py

@@ -40,7 +40,6 @@ def generate_base_prompt(title, segment_text, tags, _format=None, style=None, ex
     # 添加额外内容
     if extras:
         prompt += f"\n{extras}"
-    print(prompt)
     return prompt
 
 
@@ -104,9 +103,9 @@ def get_link_format():
 
 def get_screenshot_format():
     return '''
-    11. **原片截图**: 如果某个部分涉及**视觉演示**或任何能帮助理解的内容，插入截图提示：
-      - 格式：`*Screenshot-[mm:ss]`
-      至少插入 1-3张截图
+11. **原片截图**:你收到的截图一般是一个网格，网格的每张图片就是一个时间点，左上角会包含时间mm:ss的格式，请你结合我发你的图片插入截图提示，请你帮助用户更好的理解视频内容，请你认真的分析每个图片和对应的转写文案，插入最合适的内容来备注用户理解，请一定按照这个格式 返回否则系统无法解析：
+- 格式：`*Screenshot-[mm:ss]`
+
     '''
 
 

backend/app/gpt/universal_gpt.py

@@ -7,6 +7,7 @@ from app.models.transcriber_model import TranscriptSegment
 from datetime import timedelta
 from typing import List
 
+
 class UniversalGPT(GPT):
     def __init__(self, client, model: str, temperature: float = 0.7):
         self.client = client
@@ -28,21 +29,41 @@ class UniversalGPT(GPT):
     def ensure_segments_type(self, segments) -> List[TranscriptSegment]:
         return [TranscriptSegment(**seg) if isinstance(seg, dict) else seg for seg in segments]
 
-    def create_messages(self, segments: List[TranscriptSegment],**kwargs):
-        print("UniversalGPT",kwargs)
-        content =generate_base_prompt(
+    def create_messages(self, segments: List[TranscriptSegment], **kwargs):
+
+        content_text = generate_base_prompt(
             title=kwargs.get('title'),
             segment_text=self._build_segment_text(segments),
             tags=kwargs.get('tags'),
             _format=kwargs.get('_format'),
             style=kwargs.get('style'),
-            extras=kwargs.get('extras')
+            extras=kwargs.get('extras'),
         )
 
-        return [{"role": "user", "content": content }]
+        # ⛳ 组装 content 数组，支持 text + image_url 混合
+        content = [{"type": "text", "text": content_text}]
+        video_img_urls = kwargs.get('video_img_urls', [])
+
+        for url in video_img_urls:
+            content.append({
+                "type": "image_url",
+                "image_url": {
+                    "url": url,
+                    "detail": "auto"
+                }
+            })
+
+        # ✅ 正确格式：整体包在一个 message 里，role + content array
+        messages = [{
+            "role": "user",
+            "content": content
+        }]
+
+        return messages
 
     def list_models(self):
         return self.client.models.list()
+
     def summarize(self, source: GPTSource) -> str:
         self.screenshot = source.screenshot
         self.link = source.link
@@ -51,8 +72,8 @@ class UniversalGPT(GPT):
         messages = self.create_messages(
             source.segment,
             title=source.title,
-            tags=source.tags
-            ,
+            tags=source.tags,
+            video_img_urls=source.video_img_urls,
             _format=source._format,
             style=source.style,
             extras=source.extras
@@ -63,7 +84,3 @@ class UniversalGPT(GPT):
             temperature=0.7
         )
         return response.choices[0].message.content.strip()
-
-if __name__ == '__main__':
-    print('s')
-

backend/app/models/gpt_model.py

@@ -14,4 +14,5 @@ class GPTSource:
     style: Optional[str] = None
     extras: Optional[str] = None
     _format: Optional[list] = None
+    video_img_urls:  Optional[list] = None
 

backend/app/routers/note.py

@@ -37,12 +37,15 @@ class VideoRequest(BaseModel):
     quality: DownloadQuality
     screenshot: Optional[bool] = False
     link: Optional[bool] = False
-    model_name:str
-    provider_id:str
+    model_name: str
+    provider_id: str
     task_id: Optional[str] = None
-    format:Optional[list]=[]
-    style:str=None
-    extras:Optional[str]
+    format: Optional[list] = []
+    style: str = None
+    extras: Optional[str]
+    video_understanding: Optional[bool] = False
+    video_interval: Optional[int] = 0
+    grid_size: Optional[list] = []
 
     @field_validator("video_url")
     def validate_supported_url(cls, v):
@@ -59,6 +62,7 @@ class VideoRequest(BaseModel):
 NOTE_OUTPUT_DIR = "note_results"
 UPLOAD_DIR = "uploads"
 
+
 def save_note_to_file(task_id: str, note):
     os.makedirs(NOTE_OUTPUT_DIR, exist_ok=True)
     with open(os.path.join(NOTE_OUTPUT_DIR, f"{task_id}.json"), "w", encoding="utf-8") as f:
@@ -66,8 +70,10 @@ def save_note_to_file(task_id: str, note):
 
 
 def run_note_task(task_id: str, video_url: str, platform: str, quality: DownloadQuality,
-                  link: bool = False,screenshot: bool = False,model_name:str=None,provider_id:str=None,
-                  _format:list=None,style:str=None,extras:str=None):
+                  link: bool = False, screenshot: bool = False, model_name: str = None, provider_id: str = None,
+                  _format: list = None, style: str = None, extras: str = None, video_understanding: bool = False,
+                  video_interval=0, grid_size=[]
+                  ):
     try:
         if not model_name or not provider_id:
             raise HTTPException(status_code=400, detail="请选择模型和提供者")
@@ -84,25 +90,26 @@ def run_note_task(task_id: str, video_url: str, platform: str, quality: Download
             style=style,
             extras=extras,
             screenshot=screenshot
+            , video_understanding=video_understanding,
+            video_interval=video_interval,
+            grid_size=grid_size
         )
-        print('Note 结果',note)
+        logger.info(f"Note generated: {task_id}")
         save_note_to_file(task_id, note)
     except Exception as e:
         save_note_to_file(task_id, {"error": str(e)})
 
 
 @router.post('/delete_task')
-def delete_task(data:RecordRequest):
+def delete_task(data: RecordRequest):
     try:
 
-        NoteGenerator().delete_note(video_id=data.video_id,platform=data.platform)
+        NoteGenerator().delete_note(video_id=data.video_id, platform=data.platform)
         return R.success(msg='删除成功')
     except Exception as e:
         return R.error(msg=e)
 
 
-
-
 @router.post("/upload")
 async def upload(file: UploadFile = File(...)):
     os.makedirs(UPLOAD_DIR, exist_ok=True)
@@ -114,6 +121,7 @@ async def upload(file: UploadFile = File(...)):
     # 假设你静态目录挂载了 /uploads
     return R.success({"url": f"/uploads/{file.filename}"})
 
+
 @router.post("/generate_note")
 def generate_note(data: VideoRequest, background_tasks: BackgroundTasks):
     try:
@@ -128,7 +136,6 @@ def generate_note(data: VideoRequest, background_tasks: BackgroundTasks):
         #
         #     )
 
-
         if data.task_id:
             # 如果传了task_id，说明是重试！
             task_id = data.task_id
@@ -139,14 +146,14 @@ def generate_note(data: VideoRequest, background_tasks: BackgroundTasks):
             # 正常新建任务
             task_id = str(uuid.uuid4())
 
-        background_tasks.add_task(run_note_task, task_id, data.video_url, data.platform, data.quality,data.link ,data.screenshot,data.model_name,data.provider_id,data.format,data.style,data.extras)
+        background_tasks.add_task(run_note_task, task_id, data.video_url, data.platform, data.quality, data.link,
+                                  data.screenshot, data.model_name, data.provider_id, data.format, data.style,
+                                  data.extras, data.video_understanding, data.video_interval, data.grid_size)
         return R.success({"task_id": task_id})
     except Exception as e:
         raise HTTPException(status_code=500, detail=str(e))
 
 
-
-
 @router.get("/task_status/{task_id}")
 def get_task_status(task_id: str):
     status_path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}.status.json")

backend/app/services/note.py

@@ -28,7 +28,7 @@ from app.services.constant import SUPPORT_PLATFORM_MAP
 
 from app.services.provider import ProviderService
 from app.transcriber.base import Transcriber
-from app.transcriber.transcriber_provider import get_transcriber,_transcribers
+from app.transcriber.transcriber_provider import get_transcriber, _transcribers
 from app.transcriber.whisper import WhisperTranscriber
 import re
 
@@ -39,12 +39,13 @@ from app.utils.video_helper import generate_screenshot
 # from app.services.gpt import summarize_text
 from dotenv import load_dotenv
 from app.utils.logger import get_logger
+from app.utils.video_reader import VideoReader
 from events import transcription_finished
 
 logger = get_logger(__name__)
 load_dotenv()
 api_path = os.getenv("API_BASE_URL", "http://localhost")
-BACKEND_PORT= os.getenv("BACKEND_PORT", 8000)
+BACKEND_PORT = os.getenv("BACKEND_PORT", 8000)
 
 BACKEND_BASE_URL = f"{api_path}:{BACKEND_PORT}"
 output_dir = os.getenv('OUT_DIR')
@@ -53,11 +54,12 @@ logger.info("starting up")
 
 NOTE_OUTPUT_DIR = "note_results"
 
+
 class NoteGenerator:
     def __init__(self):
         self.model_size: str = 'base'
         self.device: Union[str, None] = None
-        self.transcriber_type = os.getenv('TRANSCRIBER_TYPE','fast-whisper')
+        self.transcriber_type = os.getenv('TRANSCRIBER_TYPE', 'fast-whisper')
         self.transcriber = self.get_transcriber()
         self.video_path = None
         logger.info("初始化NoteGenerator")
@@ -94,7 +96,7 @@ class NoteGenerator:
         return gpt
 
     def get_downloader(self, platform: str) -> Downloader:
-        downloader =SUPPORT_PLATFORM_MAP[platform]
+        downloader = SUPPORT_PLATFORM_MAP[platform]
         if downloader:
             logger.info(f"使用{downloader}下载器")
             return downloader
@@ -120,7 +122,7 @@ class NoteGenerator:
         insert_video_task(video_id=video_id, platform=platform, task_id=task_id)
 
     def insert_screenshots_into_markdown(self, markdown: str, video_path: str, image_base_url: str,
-                                         output_dir: str,_format:list) -> str:
+                                         output_dir: str, _format: list) -> str:
         """
         扫描 markdown 中的 *Screenshot-xx:xx，生成截图并插入 markdown 图片
         :param markdown:
@@ -128,7 +130,7 @@ class NoteGenerator:
         """
         matches = self.extract_screenshot_timestamps(markdown)
         new_markdown = markdown
-        print(f"匹配到的截图：{matches}")
+
         logger.info(f"开始为笔记生成截图")
         try:
             for idx, (marker, ts) in enumerate(matches):
@@ -137,7 +139,7 @@ class NoteGenerator:
                 image_url = f"{BACKEND_BASE_URL.rstrip('/')}/{image_relative_path.lstrip('/')}"
                 replacement = f"![]({image_url})"
                 new_markdown = new_markdown.replace(marker, replacement, 1)
-            print(f"替换后的 markdown：{new_markdown}")
+
 
             return new_markdown
         except Exception as e:
@@ -180,14 +182,18 @@ class NoteGenerator:
             _format: list = None,
             style: str = None,
             extras: str = None,
-            path: Union[str, None] = None
+            path: Union[str, None] = None,
+            video_understanding: bool = False,
+            video_interval=0,
+            grid_size=[]
     ) -> NoteResult:
+
         try:
             logger.info(f"🎯 开始解析并生成笔记，task_id={task_id}")
             self.update_task_status(task_id, TaskStatus.PARSING)
             downloader = self.get_downloader(platform)
             gpt = self.get_gpt(model_name=model_name, provider_id=provider_id)
-
+            video_img_urls = []
             audio_cache_path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}_audio.json")
             transcript_cache_path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}_transcript.json")
             markdown_cache_path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}_markdown.md")
@@ -201,11 +207,20 @@ class NoteGenerator:
                         audio_data = json.load(f)
                     audio = AudioDownloadResult(**audio_data)
                 else:
-                    if 'screenshot' in _format:
+                    if 'screenshot' in _format or video_understanding:
                         video_path = downloader.download_video(video_url)
                         self.video_path = video_path
                         logger.info(f"成功下载视频文件: {video_path}")
-                    screenshot= 'screenshot' in _format
+                        video_img_urls = VideoReader(
+                            video_path=video_path,
+                            grid_size=tuple(grid_size),
+                            frame_interval=video_interval,
+                            unit_width=1280,
+                            unit_height=720,
+                            save_quality=90,
+                        ).run()
+
+                    screenshot = 'screenshot' in _format
                     audio: AudioDownloadResult = downloader.download(
                         video_url=video_url,
                         quality=quality,
@@ -261,6 +276,7 @@ class NoteGenerator:
                         segment=transcript.segments,
                         tags=audio.raw_info.get('tags'),
                         screenshot=screenshot,
+                        video_img_urls=video_img_urls,
                         link=link,
                         _format=_format,
                         style=style,
@@ -279,12 +295,13 @@ class NoteGenerator:
             # -------- 4. 插入截图 --------
             if _format and 'screenshot' in _format:
                 try:
-                    markdown = self.insert_screenshots_into_markdown(markdown, self.video_path, image_base_url, output_dir,_format)
+                    markdown = self.insert_screenshots_into_markdown(markdown, self.video_path, image_base_url,
+                                                                     output_dir, _format)
                 except Exception as e:
                     logger.warning(f"⚠️ 插入截图失败，跳过处理，task_id={task_id}，错误信息：{e}")
             if _format and 'link' in _format:
                 try:
-                    markdown = replace_content_markers(markdown, video_id=audio.video_id,platform=platform)
+                    markdown = replace_content_markers(markdown, video_id=audio.video_id, platform=platform)
                 except Exception as e:
                     logger.warning(f"⚠️ 插入链接失败，跳过处理，task_id={task_id}，错误信息：{e}")
                 # 注意：截图失败不终止整体流程
@@ -310,7 +327,3 @@ class NoteGenerator:
             logger.error(f"❌ 笔记生成流程异常终止，task_id={task_id}，错误信息：{e}")
             self.update_task_status(task_id, TaskStatus.FAILED, message=str(e))
             raise f'❌ 笔记生成流程异常终止，task_id={task_id}，错误信息：{e}'
-
-
-
-

backend/app/utils/video_reader.py

@@ -0,0 +1,134 @@
+import base64
+import os
+import re
+import subprocess
+import ffmpeg
+from PIL import Image, ImageDraw, ImageFont
+
+from app.utils.logger import get_logger
+
+logger = get_logger(__name__)
+class VideoReader:
+    def __init__(self,
+                 video_path: str,
+                 grid_size=(3, 3),
+                 frame_interval=2,
+                 unit_width=960,
+                 unit_height=540,
+                 save_quality=90,
+                 font_path="fonts/arial.ttf",
+                 frame_dir="data/output_frames",
+                 grid_dir="data/grid_output"):
+        self.video_path = video_path
+        self.grid_size = grid_size
+        self.frame_interval = frame_interval
+        self.unit_width = unit_width
+        self.unit_height = unit_height
+        self.save_quality = save_quality
+        self.font_path = font_path
+        self.frame_dir = frame_dir
+        self.grid_dir = grid_dir
+
+    def format_time(self, seconds: float) -> str:
+        mm = int(seconds // 60)
+        ss = int(seconds % 60)
+        return f"{mm:02d}_{ss:02d}"
+
+    def extract_time_from_filename(self, filename: str) -> float:
+        match = re.search(r"frame_(\d{2})_(\d{2})\.jpg", filename)
+        if match:
+            mm, ss = map(int, match.groups())
+            return mm * 60 + ss
+        return float('inf')
+
+    def extract_frames(self, max_frames=1000) -> list[str]:
+
+        try:
+            os.makedirs(self.frame_dir, exist_ok=True)
+            duration = float(ffmpeg.probe(self.video_path)["format"]["duration"])
+            timestamps = [i for i in range(0, int(duration), self.frame_interval)][:max_frames]
+
+            image_paths = []
+            for ts in timestamps:
+                time_label = self.format_time(ts)
+                output_path = os.path.join(self.frame_dir, f"frame_{time_label}.jpg")
+                cmd = ["ffmpeg", "-ss", str(ts), "-i", self.video_path, "-frames:v", "1", "-q:v", "2", "-y", output_path,
+                       "-hide_banner", "-loglevel", "error"]
+                subprocess.run(cmd, check=True)
+                image_paths.append(output_path)
+            return image_paths
+        except Exception as e:
+            logger.error(f"分割帧发生错误：{str(e)}")
+            raise ValueError("视频处理失败")
+
+    def group_images(self) -> list[list[str]]:
+        image_files = [os.path.join(self.frame_dir, f) for f in os.listdir(self.frame_dir) if
+                       f.startswith("frame_") and f.endswith(".jpg")]
+        image_files.sort(key=lambda f: self.extract_time_from_filename(os.path.basename(f)))
+        group_size = self.grid_size[0] * self.grid_size[1]
+        return [image_files[i:i + group_size] for i in range(0, len(image_files), group_size)]
+
+    def concat_images(self, image_paths: list[str], name: str) -> str:
+        os.makedirs(self.grid_dir, exist_ok=True)
+        font = ImageFont.truetype(self.font_path, 48) if os.path.exists(self.font_path) else ImageFont.load_default()
+        images = []
+
+        for path in image_paths:
+            img = Image.open(path).convert("RGB").resize((self.unit_width, self.unit_height), Image.Resampling.LANCZOS)
+            timestamp = re.search(r"frame_(\d{2})_(\d{2})\.jpg", os.path.basename(path))
+            time_text = f"{timestamp.group(1)}:{timestamp.group(2)}" if timestamp else ""
+            draw = ImageDraw.Draw(img)
+            draw.text((10, 10), time_text, fill="yellow", font=font, stroke_width=1, stroke_fill="black")
+            images.append(img)
+
+        cols, rows = self.grid_size
+        grid_img = Image.new("RGB", (self.unit_width * cols, self.unit_height * rows), (255, 255, 255))
+
+        for i, img in enumerate(images):
+            x = (i % cols) * self.unit_width
+            y = (i // cols) * self.unit_height
+            grid_img.paste(img, (x, y))
+
+        save_path = os.path.join(self.grid_dir, f"{name}.jpg")
+        grid_img.save(save_path, quality=self.save_quality)
+        return save_path
+
+    def encode_images_to_base64(self, image_paths: list[str]) -> list[str]:
+        base64_images = []
+        for path in image_paths:
+            with open(path, "rb") as img_file:
+                encoded_string = base64.b64encode(img_file.read()).decode("utf-8")
+                base64_images.append(f"data:image/jpeg;base64,{encoded_string}")
+        return base64_images
+
+    def run(self)->list[str]:
+        logger.info("🚀 开始提取视频帧...")
+        try:
+            #清空帧文件夹
+            for file in os.listdir(self.frame_dir):
+                if file.startswith("frame_"):
+                    os.remove(os.path.join(self.frame_dir, file))
+            #清空网格文件夹
+            for file in os.listdir(self.grid_dir):
+                if file.startswith("grid_"):
+                    os.remove(os.path.join(self.grid_dir, file))
+            self.extract_frames()
+
+            logger.info("🧩 开始拼接网格图...")
+            image_paths = []
+            groups = self.group_images()
+            for idx, group in enumerate(groups, start=1):
+                if len(group) < self.grid_size[0] * self.grid_size[1]:
+                    logger.warning(f"⚠️ 跳过第 {idx} 组，图片不足 {self.grid_size[0] * self.grid_size[1]} 张")
+                    continue
+                out_path = self.concat_images(group, f"grid_{idx}")
+                image_paths.append(out_path)
+
+            logger.info("📤 开始编码图像...")
+            urls = self.encode_images_to_base64(image_paths)
+            return urls
+        except Exception as e:
+            logger.error(f"发生错误：{str(e)}")
+            raise ValueError("视频处理失败")
+
+