fix: 性能优化、前端转写器配置、任务进度丢失及 MLX Whisper 回退问题修复

### 性能优化
- 后端任务执行从串行锁改为 ThreadPoolExecutor 并发执行（默认3线程）
- 添加 GZipMiddleware 响应压缩 + Nginx gzip 配置
- 数据库连接池参数优化（pool_size=10, max_overflow=20）
- 视频帧提取并行化（ThreadPoolExecutor）
- LLM 重试配置缓存到实例，避免每次请求读 env var
- 前端路由级代码拆分（React.lazy + Suspense）
- Vite manualChunks 拆分 markdown/markmap/vendor
- MarkdownViewer 用 React.memo + useMemo 减少不必要渲染
- NoteHistory Fuse.js 实例 useMemo 缓存
- useTaskPolling 无待处理任务时跳过轮询
- 移除 antd 依赖（NoteForm Alert、modelForm Tag），改用 shadcn/ui

### 前端转写器配置（新功能）
- 新增 TranscriberConfigManager（JSON 文件存储，替代环境变量）
- 新增 GET/POST /transcriber_config API 端点
- 新增 GET /transcriber_models_status 模型下载状态查询
- 新增 POST /transcriber_download 后台模型下载触发
- 前端转写器设置页面：引擎选择、模型大小选择、模型下载管理
- deploy_status 端点同步从配置文件读取

### Bug 修复
- 修复任务进行中切换页面后进度丢失：Home.tsx status 派生逻辑补全中间状态
- 修复 MLX Whisper 静默回退 fast-whisper：移除环境变量门控，macOS 下自动尝试导入
- MLX Whisper 不可用时抛出 RuntimeError 而非静默回退
- 前端展示 MLX Whisper 可用性状态，不可用时禁用保存

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

BillNote_frontend/src/App.tsx

@@ -1,21 +1,23 @@
 import './App.css'
-import { HomePage } from './pages/HomePage/Home.tsx'
+import { lazy, Suspense, useEffect } from 'react'
+import { BrowserRouter, Navigate, Routes, Route } from 'react-router-dom'
 import { useTaskPolling } from '@/hooks/useTaskPolling.ts'
-import SettingPage from './pages/SettingPage/index.tsx'
-import { BrowserRouter, Navigate, Routes } from 'react-router-dom'
-import { Route } from 'react-router-dom'
-import Index from '@/pages/Index.tsx'
-import NotFoundPage from '@/pages/NotFoundPage'
-import Model from '@/pages/SettingPage/Model.tsx'
-import ProviderForm from '@/components/Form/modelForm/Form.tsx'
-import AboutPage from '@/pages/SettingPage/about.tsx'
-import Monitor from '@/pages/SettingPage/Monitor.tsx'
-import Downloader from '@/pages/SettingPage/Downloader.tsx'
-import DownloaderForm from '@/components/Form/DownloaderForm/Form.tsx'
-import { useEffect } from 'react'
-import { systemCheck } from '@/services/system.ts'
 import { useCheckBackend } from '@/hooks/useCheckBackend.ts'
+import { systemCheck } from '@/services/system.ts'
 import BackendInitDialog from '@/components/BackendInitDialog'
+import Index from '@/pages/Index.tsx'
+import { HomePage } from './pages/HomePage/Home.tsx'
+
+// 非首屏页面使用 React.lazy 按需加载
+const SettingPage = lazy(() => import('./pages/SettingPage/index.tsx'))
+const Model = lazy(() => import('@/pages/SettingPage/Model.tsx'))
+const ProviderForm = lazy(() => import('@/components/Form/modelForm/Form.tsx'))
+const AboutPage = lazy(() => import('@/pages/SettingPage/about.tsx'))
+const Monitor = lazy(() => import('@/pages/SettingPage/Monitor.tsx'))
+const Downloader = lazy(() => import('@/pages/SettingPage/Downloader.tsx'))
+const DownloaderForm = lazy(() => import('@/components/Form/DownloaderForm/Form.tsx'))
+const TranscriberPage = lazy(() => import('@/pages/SettingPage/transcriber.tsx'))
+const NotFoundPage = lazy(() => import('@/pages/NotFoundPage'))
 
 function App() {
   useTaskPolling(3000) // 每 3 秒轮询一次
@@ -41,6 +43,7 @@ function App() {
   return (
     <>
       <BrowserRouter>
+        <Suspense fallback={<div className="flex h-screen items-center justify-center">加载中…</div>}>
           <Routes>
             <Route path="/" element={<Index />}>
               <Route index element={<HomePage />} />
@@ -53,6 +56,7 @@ function App() {
                 <Route path="download" element={<Downloader />}>
                   <Route path=":id" element={<DownloaderForm />} />
                 </Route>
+                <Route path="transcriber" element={<TranscriberPage />} />
                 <Route path="monitor" element={<Monitor />}></Route>
                 <Route path="about" element={<AboutPage />}></Route>
                 <Route path="*" element={<NotFoundPage />} />
@@ -60,6 +64,7 @@ function App() {
               <Route path="*" element={<NotFoundPage />} />
             </Route>
           </Routes>
+        </Suspense>
       </BrowserRouter>
     </>
   )

BillNote_frontend/src/components/Form/modelForm/Form.tsx

@@ -27,7 +27,7 @@ import {
 import { ModelSelector } from '@/components/Form/modelForm/ModelSelector.tsx'
 import { Alert, AlertDescription, AlertTitle } from '@/components/ui/alert.tsx'
 import { Tags } from 'lucide-react'
-import { Tag } from 'antd'
+import { X } from 'lucide-react'
 import { useModelStore } from '@/store/modelStore'
 
 // ✅ Provider表单schema
@@ -312,12 +312,12 @@ const ProviderForm = ({ isCreate = false }: { isCreate?: boolean }) => {
             {
               models && models.map(model => {
                 return (
-                  <>
-                    <Tag onClose={()=>{
-                      handelDelete(model.id)
-                    }} key={model.id} closable color={'blue'}>
+                  <span key={model.id} className="inline-flex items-center gap-1 rounded-md bg-blue-100 px-2 py-0.5 text-sm text-blue-700">
                     {model.model_name}
-                    </Tag></>
+                    <button type="button" onClick={() => handelDelete(model.id)} className="hover:text-blue-900">
+                      <X className="h-3 w-3" />
+                    </button>
+                  </span>
 
                 )
               })

BillNote_frontend/src/hooks/useTaskPolling.ts

@@ -22,9 +22,11 @@ export const useTaskPolling = (interval = 3000) => {
         task => task.status != 'SUCCESS' && task.status != 'FAILED'
       )
 
+      // 无活跃任务时跳过轮询
+      if (pendingTasks.length === 0) return
+
       for (const task of pendingTasks) {
         try {
-          console.log('🔄 正在轮询任务：', task.id)
           const res = await get_task_status(task.id)
           const { status } = res
 
@@ -47,9 +49,7 @@ export const useTaskPolling = (interval = 3000) => {
           }
         } catch (e) {
           console.error('❌ 任务轮询失败：', e)
-          // toast.error(`生成失败 ${e.message || e}`)
           updateTaskContent(task.id, { status: 'FAILED' })
-          // removeTask(task.id)
         }
       }
     }, interval)

BillNote_frontend/src/pages/HomePage/Home.tsx

@@ -18,14 +18,15 @@ export const HomePage: FC = () => {
   useEffect(() => {
     if (!currentTask) {
       setStatus('idle')
-    } else if (currentTask.status === 'PENDING') {
-      setStatus('loading')
     } else if (currentTask.status === 'SUCCESS') {
       setStatus('success')
     } else if (currentTask.status === 'FAILED') {
       setStatus('failed')
+    } else {
+      // PENDING、PARSING、DOWNLOADING、TRANSCRIBING、SUMMARIZING 等所有进行中状态
+      setStatus('loading')
     }
-  }, [currentTask])
+  }, [currentTask, currentTask?.status])
 
   // useEffect( () => {
   //     get_task_status('d4e87938-c066-48a0-bbd5-9bec40d53354').then(res=>{

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

@@ -1,4 +1,4 @@
-import { useState, useEffect, useRef } from 'react'
+import { useState, useEffect, useRef, useMemo, memo, FC } from 'react'
 import ReactMarkdown from 'react-markdown'
 import { Button } from '@/components/ui/button.tsx'
 import { Copy, Download, ArrowRight, Play, ExternalLink } from 'lucide-react'
@@ -16,7 +16,6 @@ import remarkMath from 'remark-math'
 import rehypeKatex from 'rehype-katex'
 import 'katex/dist/katex.min.css'
 import 'github-markdown-css/github-markdown-light.css'
-import { FC } from 'react'
 import { ScrollArea } from '@/components/ui/scroll-area.tsx'
 import { useTaskStore } from '@/store/taskStore'
 import { noteStyles } from '@/constant/note.ts'
@@ -45,7 +44,228 @@ const steps = [
   { label: '保存完成', key: 'SUCCESS' },
 ]
 
-const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
+const remarkPlugins = [gfm, remarkMath]
+const rehypePlugins = [rehypeKatex]
+
+/**
+ * 构建 ReactMarkdown components 对象，baseURL 用于修正图片路径。
+ * 使用函数 + useMemo 避免每次渲染都创建新的函数实例。
+ */
+function createMarkdownComponents(baseURL: string) {
+  return {
+    h1: ({ children, ...props }: any) => (
+      <h1
+        className="text-primary my-6 scroll-m-20 text-3xl font-extrabold tracking-tight lg:text-4xl"
+        {...props}
+      >
+        {children}
+      </h1>
+    ),
+    h2: ({ children, ...props }: any) => (
+      <h2
+        className="text-primary mt-10 mb-4 scroll-m-20 border-b pb-2 text-2xl font-semibold tracking-tight first:mt-0"
+        {...props}
+      >
+        {children}
+      </h2>
+    ),
+    h3: ({ children, ...props }: any) => (
+      <h3
+        className="text-primary mt-8 mb-4 scroll-m-20 text-xl font-semibold tracking-tight"
+        {...props}
+      >
+        {children}
+      </h3>
+    ),
+    h4: ({ children, ...props }: any) => (
+      <h4
+        className="text-primary mt-6 mb-2 scroll-m-20 text-lg font-semibold tracking-tight"
+        {...props}
+      >
+        {children}
+      </h4>
+    ),
+    p: ({ children, ...props }: any) => (
+      <p className="leading-7 [&:not(:first-child)]:mt-6" {...props}>
+        {children}
+      </p>
+    ),
+    a: ({ href, children, ...props }: any) => {
+      const isOriginLink =
+        typeof children[0] === 'string' &&
+        (children[0] as string).startsWith('原片 @')
+
+      if (isOriginLink) {
+        const timeMatch = (children[0] as string).match(/原片 @ (\d{2}:\d{2})/)
+        const timeText = timeMatch ? timeMatch[1] : '原片'
+
+        return (
+          <span className="origin-link my-2 inline-flex">
+            <a
+              href={href}
+              target="_blank"
+              rel="noopener noreferrer"
+              className="inline-flex items-center gap-1.5 rounded-full bg-blue-50 px-3 py-1 text-sm font-medium text-blue-700 transition-colors hover:bg-blue-100"
+              {...props}
+            >
+              <Play className="h-3.5 w-3.5" />
+              <span>原片（{timeText}）</span>
+            </a>
+          </span>
+        )
+      }
+
+      return (
+        <a
+          href={href}
+          target="_blank"
+          rel="noopener noreferrer"
+          className="text-primary hover:text-primary/80 inline-flex items-center gap-0.5 font-medium underline underline-offset-4"
+          {...props}
+        >
+          {children}
+          {href?.startsWith('http') && (
+            <ExternalLink className="ml-0.5 inline-block h-3 w-3" />
+          )}
+        </a>
+      )
+    },
+    img: ({ node, ...props }: any) => {
+      let src = props.src
+      if (src.startsWith('/')) {
+        src = baseURL + src
+      }
+      props.src = src
+
+      return (
+        <div className="my-8 flex justify-center">
+          <Zoom>
+            <img
+              {...props}
+              className="max-w-full cursor-zoom-in rounded-lg object-cover shadow-md transition-all hover:shadow-lg"
+              style={{ maxHeight: '500px' }}
+            />
+          </Zoom>
+        </div>
+      )
+    },
+    strong: ({ children, ...props }: any) => (
+      <strong className="text-primary font-bold" {...props}>
+        {children}
+      </strong>
+    ),
+    li: ({ children, ...props }: any) => {
+      const rawText = String(children)
+      const isFakeHeading = /^(\*\*.+\*\*)$/.test(rawText.trim())
+
+      if (isFakeHeading) {
+        return (
+          <div className="text-primary my-4 text-lg font-bold">{children}</div>
+        )
+      }
+
+      return (
+        <li className="my-1" {...props}>
+          {children}
+        </li>
+      )
+    },
+    ul: ({ children, ...props }: any) => (
+      <ul className="my-6 ml-6 list-disc [&>li]:mt-2" {...props}>
+        {children}
+      </ul>
+    ),
+    ol: ({ children, ...props }: any) => (
+      <ol className="my-6 ml-6 list-decimal [&>li]:mt-2" {...props}>
+        {children}
+      </ol>
+    ),
+    blockquote: ({ children, ...props }: any) => (
+      <blockquote
+        className="border-primary/20 text-muted-foreground mt-6 border-l-4 pl-4 italic"
+        {...props}
+      >
+        {children}
+      </blockquote>
+    ),
+    code: ({ inline, className, children, ...props }: any) => {
+      const match = /language-(\w+)/.exec(className || '')
+      const codeContent = String(children).replace(/\n$/, '')
+
+      if (!inline && match) {
+        return (
+          <div className="group bg-muted relative my-6 overflow-hidden rounded-lg border shadow-sm">
+            <div className="bg-muted text-muted-foreground flex items-center justify-between px-4 py-1.5 text-sm font-medium">
+              <div>{match[1].toUpperCase()}</div>
+              <button
+                onClick={() => {
+                  navigator.clipboard.writeText(codeContent)
+                  toast.success('代码已复制')
+                }}
+                className="bg-background/80 hover:bg-background flex items-center gap-1 rounded-md px-2 py-1 text-xs font-medium transition-colors"
+              >
+                <Copy className="h-3.5 w-3.5" />
+                复制
+              </button>
+            </div>
+            <SyntaxHighlighter
+              style={codeStyle}
+              language={match[1]}
+              PreTag="div"
+              className="!bg-muted !m-0 !p-0"
+              customStyle={{
+                margin: 0,
+                padding: '1rem',
+                background: 'transparent',
+                fontSize: '0.9rem',
+              }}
+              {...props}
+            >
+              {codeContent}
+            </SyntaxHighlighter>
+          </div>
+        )
+      }
+
+      return (
+        <code
+          className="bg-muted relative rounded px-[0.3rem] py-[0.2rem] font-mono text-sm"
+          {...props}
+        >
+          {children}
+        </code>
+      )
+    },
+    table: ({ children, ...props }: any) => (
+      <div className="my-6 w-full overflow-y-auto">
+        <table className="w-full border-collapse text-sm" {...props}>
+          {children}
+        </table>
+      </div>
+    ),
+    th: ({ children, ...props }: any) => (
+      <th
+        className="border-muted-foreground/20 border px-4 py-2 text-left font-medium [&[align=center]]:text-center [&[align=right]]:text-right"
+        {...props}
+      >
+        {children}
+      </th>
+    ),
+    td: ({ children, ...props }: any) => (
+      <td
+        className="border-muted-foreground/20 border px-4 py-2 text-left [&[align=center]]:text-center [&[align=right]]:text-right"
+        {...props}
+      >
+        {children}
+      </td>
+    ),
+    hr: ({ ...props }: any) => (
+      <hr className="border-muted-foreground/20 my-8" {...props} />
+    ),
+  }
+}
+
+const MarkdownViewer: FC<MarkdownViewerProps> = memo(({ status }) => {
   const [copied, setCopied] = useState(false)
   const [currentVerId, setCurrentVerId] = useState<string>('')
   const [selectedContent, setSelectedContent] = useState<string>('')
@@ -62,6 +282,10 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
   const [showTranscribe, setShowTranscribe] = useState(false)
   const [viewMode, setViewMode] = useState<'map' | 'preview'>('preview')
   const svgRef = useRef<SVGSVGElement>(null)
+
+  // 缓存 ReactMarkdown components，仅在 baseURL 变化时重建
+  const markdownComponents = useMemo(() => createMarkdownComponents(baseURL), [baseURL])
+
   // 多版本内容处理
   useEffect(() => {
     if (!currentTask) return
@@ -160,7 +384,7 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
       <div className="flex h-screen w-full flex-col items-center justify-center space-y-3 text-neutral-500">
         <Idle />
         <div className="text-center">
-          <p className="text-lg font-bold">输入视频链接并点击“生成笔记”</p>
+          <p className="text-lg font-bold">输入视频链接并点击"生成笔记"</p>
           <p className="mt-2 text-xs text-neutral-500">支持哔哩哔哩、YouTube 、抖音等视频平台</p>
         </div>
       </div>
@@ -220,248 +444,9 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
               <ScrollArea className="w-full">
                 <div className={'markdown-body w-full px-2'}>
                   <ReactMarkdown
-                    remarkPlugins={[gfm, remarkMath]}
-                    rehypePlugins={[rehypeKatex]}
-                    components={{
-                      // Headings with improved styling and anchor links
-                      h1: ({ children, ...props }) => (
-                        <h1
-                          className="text-primary my-6 scroll-m-20 text-3xl font-extrabold tracking-tight lg:text-4xl"
-                          {...props}
-                        >
-                          {children}
-                        </h1>
-                      ),
-                      h2: ({ children, ...props }) => (
-                        <h2
-                          className="text-primary mt-10 mb-4 scroll-m-20 border-b pb-2 text-2xl font-semibold tracking-tight first:mt-0"
-                          {...props}
-                        >
-                          {children}
-                        </h2>
-                      ),
-                      h3: ({ children, ...props }) => (
-                        <h3
-                          className="text-primary mt-8 mb-4 scroll-m-20 text-xl font-semibold tracking-tight"
-                          {...props}
-                        >
-                          {children}
-                        </h3>
-                      ),
-                      h4: ({ children, ...props }) => (
-                        <h4
-                          className="text-primary mt-6 mb-2 scroll-m-20 text-lg font-semibold tracking-tight"
-                          {...props}
-                        >
-                          {children}
-                        </h4>
-                      ),
-
-                      // Paragraphs with better line height
-                      p: ({ children, ...props }) => (
-                        <p className="leading-7 [&:not(:first-child)]:mt-6" {...props}>
-                          {children}
-                        </p>
-                      ),
-
-                      // Enhanced links with special handling for "原片" links
-                      a: ({ href, children, ...props }) => {
-                        const isOriginLink =
-                          typeof children[0] === 'string' &&
-                          (children[0] as string).startsWith('原片 @')
-
-                        if (isOriginLink) {
-                          const timeMatch = (children[0] as string).match(/原片 @ (\d{2}:\d{2})/)
-                          const timeText = timeMatch ? timeMatch[1] : '原片'
-
-                          return (
-                            <span className="origin-link my-2 inline-flex">
-                              <a
-                                href={href}
-                                target="_blank"
-                                rel="noopener noreferrer"
-                                className="inline-flex items-center gap-1.5 rounded-full bg-blue-50 px-3 py-1 text-sm font-medium text-blue-700 transition-colors hover:bg-blue-100"
-                                {...props}
-                              >
-                                <Play className="h-3.5 w-3.5" />
-                                <span>原片（{timeText}）</span>
-                              </a>
-                            </span>
-                          )
-                        }
-
-                        // Default link styling with external indicator
-                        return (
-                          <a
-                            href={href}
-                            target="_blank"
-                            rel="noopener noreferrer"
-                            className="text-primary hover:text-primary/80 inline-flex items-center gap-0.5 font-medium underline underline-offset-4"
-                            {...props}
-                          >
-                            {children}
-                            {href?.startsWith('http') && (
-                              <ExternalLink className="ml-0.5 inline-block h-3 w-3" />
-                            )}
-                          </a>
-                        )
-                      },
-
-                      // Enhanced image with zoom capability
-                      img: ({ node, ...props }) =>{
-                        // Fix the URL by removing the 'undefined' prefix if it exists
-                        let src = props.src
-                        if (src.startsWith('/')) {
-                          src = baseURL + src
-                        }
-                        props.src = src
-
-                     return(
-                      <div className="my-8 flex justify-center">
-                          <Zoom>
-                            <img
-                              {...props}
-                              className="max-w-full cursor-zoom-in rounded-lg object-cover shadow-md transition-all hover:shadow-lg"
-                              style={{ maxHeight: '500px' }}
-                            />
-                          </Zoom>
-                        </div>
-                      )},
-
-                      // Better strong/bold text
-                      strong: ({ children, ...props }) => (
-                        <strong className="text-primary font-bold" {...props}>
-                          {children}
-                        </strong>
-                      ),
-
-                      // Enhanced list items with support for "fake headings"
-                      li: ({ children, ...props }) => {
-                        const rawText = String(children)
-                        const isFakeHeading = /^(\*\*.+\*\*)$/.test(rawText.trim())
-
-                        if (isFakeHeading) {
-                          return (
-                            <div className="text-primary my-4 text-lg font-bold">{children}</div>
-                          )
-                        }
-
-                        return (
-                          <li className="my-1" {...props}>
-                            {children}
-                          </li>
-                        )
-                      },
-
-                      // Enhanced unordered lists
-                      ul: ({ children, ...props }) => (
-                        <ul className="my-6 ml-6 list-disc [&>li]:mt-2" {...props}>
-                          {children}
-                        </ul>
-                      ),
-
-                      // Enhanced ordered lists
-                      ol: ({ children, ...props }) => (
-                        <ol className="my-6 ml-6 list-decimal [&>li]:mt-2" {...props}>
-                          {children}
-                        </ol>
-                      ),
-
-                      // Enhanced blockquotes
-                      blockquote: ({ children, ...props }) => (
-                        <blockquote
-                          className="border-primary/20 text-muted-foreground mt-6 border-l-4 pl-4 italic"
-                          {...props}
-                        >
-                          {children}
-                        </blockquote>
-                      ),
-
-                      // Enhanced code blocks with syntax highlighting and copy button
-                      code: ({ inline, className, children, ...props }) => {
-                        const match = /language-(\w+)/.exec(className || '')
-                        const codeContent = String(children).replace(/\n$/, '')
-
-                        if (!inline && match) {
-                          return (
-                            <div className="group bg-muted relative my-6 overflow-hidden rounded-lg border shadow-sm">
-                              <div className="bg-muted text-muted-foreground flex items-center justify-between px-4 py-1.5 text-sm font-medium">
-                                <div>{match[1].toUpperCase()}</div>
-                                <button
-                                  onClick={() => {
-                                    navigator.clipboard.writeText(codeContent)
-                                    toast.success('代码已复制')
-                                  }}
-                                  className="bg-background/80 hover:bg-background flex items-center gap-1 rounded-md px-2 py-1 text-xs font-medium transition-colors"
-                                >
-                                  <Copy className="h-3.5 w-3.5" />
-                                  复制
-                                </button>
-                              </div>
-                              <SyntaxHighlighter
-                                style={codeStyle}
-                                language={match[1]}
-                                PreTag="div"
-                                className="!bg-muted !m-0 !p-0"
-                                customStyle={{
-                                  margin: 0,
-                                  padding: '1rem',
-                                  background: 'transparent',
-                                  fontSize: '0.9rem',
-                                }}
-                                {...props}
-                              >
-                                {codeContent}
-                              </SyntaxHighlighter>
-                            </div>
-                          )
-                        }
-
-                        // Inline code styling
-                        return (
-                          <code
-                            className="bg-muted relative rounded px-[0.3rem] py-[0.2rem] font-mono text-sm"
-                            {...props}
-                          >
-                            {children}
-                          </code>
-                        )
-                      },
-
-                      // Enhanced tables
-                      table: ({ children, ...props }) => (
-                        <div className="my-6 w-full overflow-y-auto">
-                          <table className="w-full border-collapse text-sm" {...props}>
-                            {children}
-                          </table>
-                        </div>
-                      ),
-
-                      // Table headers
-                      th: ({ children, ...props }) => (
-                        <th
-                          className="border-muted-foreground/20 border px-4 py-2 text-left font-medium [&[align=center]]:text-center [&[align=right]]:text-right"
-                          {...props}
-                        >
-                          {children}
-                        </th>
-                      ),
-
-                      // Table cells
-                      td: ({ children, ...props }) => (
-                        <td
-                          className="border-muted-foreground/20 border px-4 py-2 text-left [&[align=center]]:text-center [&[align=right]]:text-right"
-                          {...props}
-                        >
-                          {children}
-                        </td>
-                      ),
-
-                      // Horizontal rule
-                      hr: ({ ...props }) => (
-                        <hr className="border-muted-foreground/20 my-8" {...props} />
-                      ),
-                    }}
+                    remarkPlugins={remarkPlugins}
+                    rehypePlugins={rehypePlugins}
+                    components={markdownComponents}
                   >
                     {selectedContent}
                   </ReactMarkdown>
@@ -488,6 +473,8 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
       )}
     </div>
   )
-}
+})
+
+MarkdownViewer.displayName = 'MarkdownViewer'
 
 export default MarkdownViewer

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

@@ -13,7 +13,7 @@ import { zodResolver } from '@hookform/resolvers/zod'
 import { z } from 'zod'
 
 import { Info, Loader2, Plus } from 'lucide-react'
-import { message, Alert } from 'antd'
+import { Alert, AlertDescription } from '@/components/ui/alert.tsx'
 import { generateNote } from '@/services/note.ts'
 import { uploadFile } from '@/services/upload.ts'
 import { useTaskStore } from '@/store/taskStore'
@@ -513,17 +513,11 @@ const NoteForm = () => {
                 )}
               />
             </div>
-            <Alert
-              closable
-              type="error"
-              message={
-                <div>
-                  <strong>提示：</strong>
-                  <p>视频理解功能必须使用多模态模型。</p>
-                </div>
-              }
-              className="text-sm"
-            />
+            <Alert variant="warning" className="text-sm">
+              <AlertDescription>
+                <strong>提示：</strong>视频理解功能必须使用多模态模型。
+              </AlertDescription>
+            </Alert>
           </div>
 
           {/* 笔记格式 */}

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

@@ -14,7 +14,7 @@ import {
   TooltipTrigger,
 } from '@/components/ui/tooltip.tsx'
 import LazyImage from "@/components/LazyImage.tsx";
-import {FC, useState ,useEffect } from 'react'
+import {FC, useState, useEffect, useMemo} from 'react'
 
 interface NoteHistoryProps {
   onSelect: (taskId: string) => void
@@ -28,10 +28,10 @@ const NoteHistory: FC<NoteHistoryProps> = ({ onSelect, selectedId }) => {
   const baseURL = (String(import.meta.env.VITE_API_BASE_URL || 'api')).replace(/\/$/, '')
   const [rawSearch, setRawSearch] = useState('')
   const [search, setSearch] = useState('')
-  const fuse = new Fuse(tasks, {
+  const fuse = useMemo(() => new Fuse(tasks, {
     keys: ['audioMeta.title'],
     threshold: 0.4 // 匹配精度（越低越严格）
-  })
+  }), [tasks])
   useEffect(() => {
     const timer = setTimeout(() => {
       if (rawSearch === '') return

BillNote_frontend/src/pages/SettingPage/Menu.tsx

@@ -1,5 +1,6 @@
 import {
   BotMessageSquare,
+  Captions,
   HardDriveDownload,
   Info,
   Activity,
@@ -14,14 +15,12 @@ const Menu = () => {
       icon: <BotMessageSquare />,
       path: '/settings/model',
     },
-    // TODO ：下一版本升级优化
-    // {
-    //   id: ' transcriber',
-    //   name: '音频转译配置',
-    //   icon: <Captions />,
-    //   path: '/settings/transcriber',
-    // },
-    // //下载配置
+    {
+      id: 'transcriber',
+      name: '音频转写配置',
+      icon: <Captions />,
+      path: '/settings/transcriber',
+    },
     {
       id: 'download',
       name: '下载配置',

BillNote_frontend/src/pages/SettingPage/transcriber.tsx

@@ -1,8 +1,255 @@
-const Transcriber = () => {
+import { useState, useEffect, useCallback } from 'react'
+import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card'
+import { Button } from '@/components/ui/button'
+import { Badge } from '@/components/ui/badge'
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from '@/components/ui/select'
+import { Alert, AlertDescription } from '@/components/ui/alert'
+import { AudioLines, AlertTriangle, CheckCircle2, Download, Loader2, Save, XCircle } from 'lucide-react'
+import { toast } from 'react-hot-toast'
+import {
+  getTranscriberConfig,
+  updateTranscriberConfig,
+  getModelsStatus,
+  downloadModel,
+  TranscriberConfig,
+  ModelStatus,
+} from '@/services/transcriber'
+
+const isWhisperType = (type: string) =>
+  type === 'fast-whisper' || type === 'mlx-whisper'
+
+export default function Transcriber() {
+  const [config, setConfig] = useState<TranscriberConfig | null>(null)
+  const [loading, setLoading] = useState(true)
+  const [saving, setSaving] = useState(false)
+  const [selectedType, setSelectedType] = useState('')
+  const [selectedModelSize, setSelectedModelSize] = useState('')
+  const [modelStatuses, setModelStatuses] = useState<ModelStatus[]>([])
+  const [mlxModelStatuses, setMlxModelStatuses] = useState<ModelStatus[]>([])
+  const [mlxAvailable, setMlxAvailable] = useState(false)
+
+  const fetchModelsStatus = useCallback(async () => {
+    try {
+      const data = await getModelsStatus()
+      setModelStatuses(data.whisper)
+      setMlxModelStatuses(data.mlx_whisper)
+      setMlxAvailable(data.mlx_available)
+    } catch {
+      // 静默失败，不阻塞主流程
+    }
+  }, [])
+
+  useEffect(() => {
+    const load = async () => {
+      try {
+        const data = await getTranscriberConfig()
+        setConfig(data)
+        setSelectedType(data.transcriber_type)
+        setSelectedModelSize(data.whisper_model_size)
+      } catch {
+        toast.error('获取转写器配置失败')
+      } finally {
+        setLoading(false)
+      }
+    }
+    load()
+    fetchModelsStatus()
+  }, [fetchModelsStatus])
+
+  // 有下载中的模型时自动轮询状态
+  useEffect(() => {
+    const hasDownloading =
+      modelStatuses.some(m => m.downloading) || mlxModelStatuses.some(m => m.downloading)
+    if (!hasDownloading) return
+
+    const timer = setInterval(fetchModelsStatus, 3000)
+    return () => clearInterval(timer)
+  }, [modelStatuses, mlxModelStatuses, fetchModelsStatus])
+
+  const handleSave = async () => {
+    setSaving(true)
+    try {
+      const payload: { transcriber_type: string; whisper_model_size?: string } = {
+        transcriber_type: selectedType,
+      }
+      if (isWhisperType(selectedType)) {
+        payload.whisper_model_size = selectedModelSize
+      }
+      await updateTranscriberConfig(payload)
+      toast.success('转写器配置已保存')
+    } catch {
+      toast.error('保存失败')
+    } finally {
+      setSaving(false)
+    }
+  }
+
+  const handleDownload = async (modelSize: string, transcriberType: string) => {
+    try {
+      await downloadModel({ model_size: modelSize, transcriber_type: transcriberType })
+      toast.success(`模型 ${modelSize} 开始下载`)
+      // 立即刷新状态
+      setTimeout(fetchModelsStatus, 1000)
+    } catch {
+      toast.error('下载请求失败')
+    }
+  }
+
+  if (loading) {
     return (
-    <div className="flex h-screen w-full flex-col items-center justify-center">
-      <h1 className="text-center text-4xl font-bold">Transcriber is under development</h1>
+      <div className="flex h-64 items-center justify-center">
+        <Loader2 className="h-6 w-6 animate-spin text-neutral-400" />
+      </div>
+    )
+  }
+
+  if (!config) {
+    return <div className="p-6 text-center text-neutral-500">无法加载配置</div>
+  }
+
+  const currentModels = selectedType === 'mlx-whisper' ? mlxModelStatuses : modelStatuses
+
+  return (
+    <div className="space-y-6 p-6">
+      <div>
+        <h2 className="text-2xl font-semibold">音频转写配置</h2>
+        <p className="mt-1 text-sm text-neutral-500">
+          选择视频音频转写为文字所使用的引擎，保存后对新任务立即生效
+        </p>
+      </div>
+
+      {/* 转写引擎选择 */}
+      <Card>
+        <CardHeader>
+          <CardTitle className="flex items-center gap-2 text-lg">
+            <AudioLines className="h-5 w-5" />
+            转写引擎
+          </CardTitle>
+        </CardHeader>
+        <CardContent className="space-y-4">
+          <div className="space-y-2">
+            <label className="text-sm font-medium">转写器类型</label>
+            <Select value={selectedType} onValueChange={setSelectedType}>
+              <SelectTrigger className="w-full max-w-xs">
+                <SelectValue />
+              </SelectTrigger>
+              <SelectContent>
+                {config.available_types.map(t => (
+                  <SelectItem key={t.value} value={t.value}>
+                    {t.label}
+                  </SelectItem>
+                ))}
+              </SelectContent>
+            </Select>
+          </div>
+
+          {isWhisperType(selectedType) && (
+            <div className="space-y-2">
+              <label className="text-sm font-medium">Whisper 模型大小</label>
+              <Select value={selectedModelSize} onValueChange={setSelectedModelSize}>
+                <SelectTrigger className="w-full max-w-xs">
+                  <SelectValue />
+                </SelectTrigger>
+                <SelectContent>
+                  {config.whisper_model_sizes.map(size => {
+                    const status = currentModels.find(m => m.model_size === size)
+                    return (
+                      <SelectItem key={size} value={size}>
+                        <span className="flex items-center gap-2">
+                          {size}
+                          {status?.downloaded && (
+                            <CheckCircle2 className="h-3 w-3 text-green-500" />
+                          )}
+                        </span>
+                      </SelectItem>
+                    )
+                  })}
+                </SelectContent>
+              </Select>
+              <p className="text-xs text-neutral-400">
+                模型越大精度越高，但速度更慢、占用更多显存
+              </p>
+            </div>
+          )}
+
+          {selectedType === 'mlx-whisper' && !config.mlx_whisper_available && (
+            <Alert variant="warning" className="text-sm">
+              <AlertTriangle className="h-4 w-4" />
+              <AlertDescription>
+                MLX Whisper 当前不可用。需要 macOS 平台并安装{' '}
+                <code className="rounded bg-neutral-100 px-1">pip install mlx_whisper</code>，
+                安装后重启后端生效。
+              </AlertDescription>
+            </Alert>
+          )}
+
+          <Button onClick={handleSave} disabled={saving || (selectedType === 'mlx-whisper' && !config.mlx_whisper_available)} className="mt-2">
+            {saving ? (
+              <Loader2 className="mr-2 h-4 w-4 animate-spin" />
+            ) : (
+              <Save className="mr-2 h-4 w-4" />
+            )}
+            保存配置
+          </Button>
+        </CardContent>
+      </Card>
+
+      {/* Whisper 模型管理 */}
+      {isWhisperType(selectedType) && currentModels.length > 0 && (
+        <Card>
+          <CardHeader>
+            <CardTitle className="flex items-center gap-2 text-lg">
+              <Download className="h-5 w-5" />
+              模型管理
+              <span className="text-sm font-normal text-neutral-400">
+                {selectedType === 'mlx-whisper' ? 'MLX Whisper' : 'Faster Whisper'}
+              </span>
+            </CardTitle>
+          </CardHeader>
+          <CardContent>
+            <div className="space-y-3">
+              {currentModels.map(model => (
+                <div
+                  key={model.model_size}
+                  className="flex items-center justify-between rounded-md border px-4 py-3"
+                >
+                  <div className="flex items-center gap-3">
+                    <span className="font-medium">{model.model_size}</span>
+                    {model.downloaded ? (
+                      <Badge variant="default" className="bg-green-500 hover:bg-green-600">
+                        已下载
+                      </Badge>
+                    ) : model.downloading ? (
+                      <Badge variant="secondary" className="flex items-center gap-1">
+                        <Loader2 className="h-3 w-3 animate-spin" />
+                        下载中
+                      </Badge>
+                    ) : (
+                      <Badge variant="outline">未下载</Badge>
+                    )}
+                  </div>
+                  {!model.downloaded && !model.downloading && (
+                    <Button
+                      size="sm"
+                      variant="outline"
+                      onClick={() => handleDownload(model.model_size, selectedType)}
+                    >
+                      <Download className="mr-1 h-4 w-4" />
+                      下载
+                    </Button>
+                  )}
+                </div>
+              ))}
+            </div>
+          </CardContent>
+        </Card>
+      )}
     </div>
   )
 }
-export default Transcriber

BillNote_frontend/src/services/transcriber.ts

@@ -0,0 +1,43 @@
+import request from '@/utils/request'
+
+export interface TranscriberConfig {
+  transcriber_type: string
+  whisper_model_size: string
+  available_types: { value: string; label: string }[]
+  whisper_model_sizes: string[]
+  mlx_whisper_available: boolean
+}
+
+export interface ModelStatus {
+  model_size: string
+  downloaded: boolean
+  downloading: boolean
+}
+
+export interface ModelsStatusResponse {
+  whisper: ModelStatus[]
+  mlx_whisper: ModelStatus[]
+  mlx_available: boolean
+}
+
+export const getTranscriberConfig = async (): Promise<TranscriberConfig> => {
+  return await request.get('/transcriber_config')
+}
+
+export const updateTranscriberConfig = async (data: {
+  transcriber_type: string
+  whisper_model_size?: string
+}) => {
+  return await request.post('/transcriber_config', data)
+}
+
+export const getModelsStatus = async (): Promise<ModelsStatusResponse> => {
+  return await request.get('/transcriber_models_status')
+}
+
+export const downloadModel = async (data: {
+  model_size: string
+  transcriber_type?: string
+}) => {
+  return await request.post('/transcriber_download', data)
+}

BillNote_frontend/vite.config.ts

@@ -18,6 +18,17 @@ export default defineConfig(({ mode }) => {
         '@': path.resolve(__dirname, './src'),
       },
     },
+    build: {
+      rollupOptions: {
+        output: {
+          manualChunks: {
+            markdown: ['react-markdown', 'react-syntax-highlighter', 'remark-gfm', 'remark-math', 'rehype-katex'],
+            markmap: ['markmap-lib', 'markmap-view', 'markmap-toolbar', 'markmap-common'],
+            vendor: ['react', 'react-dom', 'react-router-dom'],
+          },
+        },
+      },
+    },
     server: {
       host: '0.0.0.0',
       port: port,

CLAUDE.md

@@ -0,0 +1,74 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project Overview
+
+BiliNote is an AI video note generation tool. It extracts content from video links (Bilibili, YouTube, Douyin, Kuaishou, local files) and generates structured Markdown notes using LLM models. Full-stack app with a FastAPI backend, React frontend, and optional Tauri desktop packaging.
+
+## Development Commands
+
+### Backend (Python 3.11 + FastAPI)
+```bash
+cd backend
+pip install -r requirements.txt
+python main.py                    # Starts on 0.0.0.0:8483
+```
+
+### Frontend (React 19 + Vite + TypeScript)
+```bash
+cd BillNote_frontend
+pnpm install
+pnpm dev          # Dev server on port 3015, proxies /api to backend
+pnpm build        # Production build
+pnpm lint         # ESLint
+```
+
+### Docker
+```bash
+docker-compose up                              # Web stack (backend + frontend + nginx)
+docker-compose -f docker-compose.gpu.yml up    # GPU variant
+```
+
+### Desktop (Tauri)
+```bash
+cd backend && ./build.sh          # Build PyInstaller backend binary
+cd BillNote_frontend && pnpm tauri build
+```
+
+## Architecture
+
+**Backend** (`backend/`) — FastAPI app, entry point `main.py`:
+- `app/routers/` — API routes: `note.py` (generation), `provider.py`, `model.py`, `config.py`
+- `app/services/` — Business logic: `note.py` (NoteGenerator orchestrates the full pipeline), `task_serial_executor.py` (task queue)
+- `app/downloaders/` — Platform adapters (bilibili, youtube, douyin, kuaishou, local) with shared `base.py` interface
+- `app/transcriber/` — Speech-to-text engines (fast-whisper, groq, bcut, kuaishou, mlx-whisper) with factory in `transcriber_provider.py`
+- `app/gpt/` — LLM integration with factory pattern (`gpt_factory.py`), prompt templates (`prompt.py`, `prompt_builder.py`), and `request_chunker.py` for long transcripts
+- `app/db/` — SQLite + SQLAlchemy: DAO pattern (`provider_dao.py`, `model_dao.py`, `video_task_dao.py`), models in `models/`
+- `app/utils/` — `response.py` (ResponseWrapper for consistent JSON), `video_helper.py` (screenshots via FFmpeg), `export.py` (PDF/DOCX)
+- `events/` (root level) — Blinker signal system for post-processing (e.g., temp file cleanup after transcription)
+
+**Frontend** (`BillNote_frontend/src/`) — React 19 + Vite + Tailwind + shadcn/ui:
+- `pages/HomePage/` — Main note generation UI: `NoteForm.tsx` (input), `MarkdownViewer.tsx` (preview), `MarkmapComponent.tsx` (mind map)
+- `pages/SettingPage/` — LLM provider management, system monitoring, transcriber config
+- `store/` — Zustand stores: `taskStore`, `modelStore`, `configStore`, `providerStore`
+- `services/` — Axios API clients matching backend routes
+- `hooks/useTaskPolling.ts` — Polls task status every 3 seconds
+- `components/ui/` — shadcn/ui (Radix-based) components
+- Path alias: `@` → `./src`
+
+**Core Workflow**: User submits URL → task queued → download video → extract audio (FFmpeg) → transcribe (Whisper/Groq/etc) → generate notes (LLM) → frontend polls for completion → display Markdown + mind map.
+
+## Key Configuration
+
+- **Ports**: Backend 8483, Frontend dev 3015, Docker maps 3015→80
+- **Environment**: Root `.env` (copy from `.env.example`). LLM API keys are configured through the UI, not env vars.
+- **Database**: SQLite at `backend/app/db/bili_note.db`, auto-initialized on first run
+- **FFmpeg**: Required system dependency for video/audio processing
+- **Vite proxy**: Dev server proxies `/api` and `/static` to backend (configured in `vite.config.ts`, reads env from parent dir)
+
+## Code Style
+
+- **Frontend**: ESLint + Prettier (2 spaces, single quotes, 100 char width, Tailwind plugin). TypeScript strict mode.
+- **Backend**: Python with type hints. No configured linter. Uses Pydantic models for validation.
+- **Note**: The frontend directory is named `BillNote_frontend` (not "Bili").

backend/app/db/engine.py

@@ -13,10 +13,19 @@ engine_args = {}
 if DATABASE_URL.startswith("sqlite"):
     engine_args["connect_args"] = {"check_same_thread": False}
 
+_pool_args = {}
+if not DATABASE_URL.startswith("sqlite"):
+    _pool_args = {
+        "pool_size": int(os.getenv("DB_POOL_SIZE", "10")),
+        "max_overflow": int(os.getenv("DB_MAX_OVERFLOW", "20")),
+        "pool_pre_ping": True,
+    }
+
 engine = create_engine(
     DATABASE_URL,
     echo=os.getenv("SQLALCHEMY_ECHO", "false").lower() == "true",
-    **engine_args
+    **engine_args,
+    **_pool_args,
 )
 
 SessionLocal = sessionmaker(autocommit=False, autoflush=False, bind=engine)

backend/app/gpt/universal_gpt.py

@@ -26,6 +26,9 @@ class UniversalGPT(GPT):
         self.max_request_bytes = int(os.getenv("OPENAI_MAX_REQUEST_BYTES", str(45 * 1024 * 1024)))
         self.checkpoint_dir = Path(os.getenv("NOTE_OUTPUT_DIR", "note_results"))
         self.checkpoint_dir.mkdir(parents=True, exist_ok=True)
+        # 初始化时缓存重试配置，避免每次请求重复读取环境变量
+        self._max_retry_attempts = max(1, int(os.getenv("OPENAI_RETRY_ATTEMPTS", "3")))
+        self._retry_base_backoff = float(os.getenv("OPENAI_RETRY_BACKOFF_SECONDS", "1.5"))
 
     def _format_time(self, seconds: float) -> str:
         return str(timedelta(seconds=int(seconds)))[2:]
@@ -176,11 +179,8 @@ class UniversalGPT(GPT):
         return status in {408, 409, 429, 500, 502, 503, 504, 524}
 
     def _chat_completion_create(self, messages: list):
-        max_attempts = max(1, int(os.getenv("OPENAI_RETRY_ATTEMPTS", "3")))
-        base_backoff = float(os.getenv("OPENAI_RETRY_BACKOFF_SECONDS", "1.5"))
-
         last_exc = None
-        for attempt in range(max_attempts):
+        for attempt in range(self._max_retry_attempts):
             try:
                 return self.client.chat.completions.create(
                     model=self.model,
@@ -189,9 +189,9 @@ class UniversalGPT(GPT):
                 )
             except Exception as exc:
                 last_exc = exc
-                if attempt == max_attempts - 1 or not self._is_retryable_error(exc):
+                if attempt == self._max_retry_attempts - 1 or not self._is_retryable_error(exc):
                     raise
-                sleep_seconds = base_backoff * (2 ** attempt)
+                sleep_seconds = self._retry_base_backoff * (2 ** attempt)
                 time.sleep(sleep_seconds)
 
         if last_exc is not None:

backend/app/routers/config.py

@@ -1,13 +1,23 @@
-from fastapi import APIRouter, HTTPException
+import os
+import platform
+from pathlib import Path
+
+from fastapi import APIRouter, HTTPException, BackgroundTasks
 from pydantic import BaseModel
 from typing import Optional
 from app.utils.response import ResponseWrapper as R
+from app.utils.logger import get_logger
+from app.utils.path_helper import get_model_dir
 
 from app.services.cookie_manager import CookieConfigManager
+from app.services.transcriber_config_manager import TranscriberConfigManager
 from ffmpeg_helper import ensure_ffmpeg_or_raise
 
+logger = get_logger(__name__)
+
 router = APIRouter()
 cookie_manager = CookieConfigManager()
+transcriber_config_manager = TranscriberConfigManager()
 
 
 class CookieUpdateRequest(BaseModel):
@@ -32,6 +42,165 @@ def update_cookie(data: CookieUpdateRequest):
 
     )
 
+class TranscriberConfigRequest(BaseModel):
+    transcriber_type: str
+    whisper_model_size: Optional[str] = None
+
+
+AVAILABLE_TRANSCRIBER_TYPES = [
+    {"value": "fast-whisper", "label": "Faster Whisper（本地）"},
+    {"value": "bcut", "label": "必剪（在线）"},
+    {"value": "kuaishou", "label": "快手（在线）"},
+    {"value": "groq", "label": "Groq（在线）"},
+    {"value": "mlx-whisper", "label": "MLX Whisper（仅macOS）"},
+]
+
+WHISPER_MODEL_SIZES = ["tiny", "base", "small", "medium", "large-v3", "large-v3-turbo"]
+
+
+@router.get("/transcriber_config")
+def get_transcriber_config():
+    from app.transcriber.transcriber_provider import MLX_WHISPER_AVAILABLE
+
+    config = transcriber_config_manager.get_config()
+    return R.success(data={
+        **config,
+        "available_types": AVAILABLE_TRANSCRIBER_TYPES,
+        "whisper_model_sizes": WHISPER_MODEL_SIZES,
+        "mlx_whisper_available": MLX_WHISPER_AVAILABLE,
+    })
+
+
+@router.post("/transcriber_config")
+def update_transcriber_config(data: TranscriberConfigRequest):
+    config = transcriber_config_manager.update_config(
+        transcriber_type=data.transcriber_type,
+        whisper_model_size=data.whisper_model_size,
+    )
+    return R.success(data=config)
+
+
+# ---- Whisper 模型下载状态 & 下载触发 ----
+
+# 用于跟踪正在进行的下载任务
+_downloading: dict[str, str] = {}  # model_size -> status ("downloading" | "done" | "failed")
+
+
+def _check_whisper_model_exists(model_size: str, subdir: str = "whisper") -> bool:
+    """检查指定 whisper 模型是否已下载到本地。"""
+    model_dir = get_model_dir(subdir)
+    model_path = os.path.join(model_dir, f"whisper-{model_size}")
+    return Path(model_path).exists()
+
+
+@router.get("/transcriber_models_status")
+def get_transcriber_models_status():
+    """返回所有 whisper 模型的下载状态。"""
+    statuses = []
+    for size in WHISPER_MODEL_SIZES:
+        downloaded = _check_whisper_model_exists(size, "whisper")
+        download_status = _downloading.get(size)
+        statuses.append({
+            "model_size": size,
+            "downloaded": downloaded,
+            "downloading": download_status == "downloading",
+        })
+
+    # 也检查 mlx-whisper（仅 macOS）
+    mlx_available = platform.system() == "Darwin"
+    mlx_statuses = []
+    if mlx_available:
+        for size in WHISPER_MODEL_SIZES:
+            mlx_key = f"mlx-{size}"
+            model_dir = get_model_dir("mlx-whisper")
+            model_path = os.path.join(model_dir, f"mlx-community/whisper-{size}")
+            downloaded = Path(model_path).exists()
+            mlx_statuses.append({
+                "model_size": size,
+                "downloaded": downloaded,
+                "downloading": _downloading.get(mlx_key) == "downloading",
+            })
+
+    return R.success(data={
+        "whisper": statuses,
+        "mlx_whisper": mlx_statuses,
+        "mlx_available": mlx_available,
+    })
+
+
+class ModelDownloadRequest(BaseModel):
+    model_size: str
+    transcriber_type: str = "fast-whisper"  # "fast-whisper" 或 "mlx-whisper"
+
+
+def _do_download_whisper(model_size: str):
+    """后台下载 faster-whisper 模型。"""
+    from app.transcriber.whisper import MODEL_MAP
+    from modelscope import snapshot_download
+
+    try:
+        _downloading[model_size] = "downloading"
+        model_dir = get_model_dir("whisper")
+        model_path = os.path.join(model_dir, f"whisper-{model_size}")
+        if Path(model_path).exists():
+            _downloading[model_size] = "done"
+            return
+        repo_id = MODEL_MAP.get(model_size)
+        if not repo_id:
+            _downloading[model_size] = "failed"
+            return
+        logger.info(f"开始下载 whisper 模型: {model_size}")
+        snapshot_download(repo_id, local_dir=model_path)
+        logger.info(f"whisper 模型下载完成: {model_size}")
+        _downloading[model_size] = "done"
+    except Exception as e:
+        logger.error(f"whisper 模型下载失败: {model_size}, {e}")
+        _downloading[model_size] = "failed"
+
+
+def _do_download_mlx_whisper(model_size: str):
+    """后台下载 mlx-whisper 模型。"""
+    key = f"mlx-{model_size}"
+    try:
+        _downloading[key] = "downloading"
+        from huggingface_hub import snapshot_download as hf_download
+
+        model_dir = get_model_dir("mlx-whisper")
+        model_name = f"mlx-community/whisper-{model_size}"
+        model_path = os.path.join(model_dir, model_name)
+        if Path(model_path).exists():
+            _downloading[key] = "done"
+            return
+        logger.info(f"开始下载 mlx-whisper 模型: {model_size}")
+        hf_download(model_name, local_dir=model_path, local_dir_use_symlinks=False)
+        logger.info(f"mlx-whisper 模型下载完成: {model_size}")
+        _downloading[key] = "done"
+    except Exception as e:
+        logger.error(f"mlx-whisper 模型下载失败: {model_size}, {e}")
+        _downloading[key] = "failed"
+
+
+@router.post("/transcriber_download")
+def download_transcriber_model(data: ModelDownloadRequest, background_tasks: BackgroundTasks):
+    """触发后台下载指定的 whisper 模型。"""
+    if data.model_size not in WHISPER_MODEL_SIZES:
+        return R.error(msg=f"不支持的模型大小: {data.model_size}")
+
+    if data.transcriber_type == "mlx-whisper":
+        if platform.system() != "Darwin":
+            return R.error(msg="MLX Whisper 仅支持 macOS")
+        key = f"mlx-{data.model_size}"
+        if _downloading.get(key) == "downloading":
+            return R.success(msg="模型正在下载中")
+        background_tasks.add_task(_do_download_mlx_whisper, data.model_size)
+    else:
+        if _downloading.get(data.model_size) == "downloading":
+            return R.success(msg="模型正在下载中")
+        background_tasks.add_task(_do_download_whisper, data.model_size)
+
+    return R.success(msg="模型下载已开始")
+
+
 @router.get("/sys_health")
 async def sys_health():
     try:
@@ -59,9 +228,10 @@ async def deploy_status():
         "gpu_name": torch.cuda.get_device_name(0) if cuda_available else None,
     }
     
-    # Whisper 模型状态
-    model_size = os.getenv("WHISPER_MODEL_SIZE", "base")
-    transcriber_type = os.getenv("TRANSCRIBER_TYPE", "fast-whisper")
+    # Whisper 模型状态（从配置文件读取，与前端设置同步）
+    transcriber_cfg = transcriber_config_manager.get_config()
+    model_size = transcriber_cfg["whisper_model_size"]
+    transcriber_type = transcriber_cfg["transcriber_type"]
     
     # FFmpeg 状态
     try:

backend/app/routers/note.py

@@ -101,7 +101,7 @@ def run_note_task(task_id: str, video_url: str, platform: str, quality: Download
             grid_size=grid_size,
         )
 
-    logger.info(f"任务进入串行队列，等待执行 (task_id={task_id})")
+    logger.info(f"任务进入执行队列 (task_id={task_id})")
     note = task_serial_executor.run(_execute_note_task)
     logger.info(f"Note generated: {task_id}")
     if not note or not note.markdown:

backend/app/services/note.py

@@ -66,9 +66,11 @@ class NoteGenerator:
     """
 
     def __init__(self):
-        self.model_size: str = "base"
+        from app.services.transcriber_config_manager import TranscriberConfigManager
+        config_manager = TranscriberConfigManager()
+        self.model_size: str = config_manager.get_whisper_model_size()
         self.device: Optional[str] = None
-        self.transcriber_type: str = os.getenv("TRANSCRIBER_TYPE", "fast-whisper")
+        self.transcriber_type: str = config_manager.get_transcriber_type()
         self.transcriber: Transcriber = self._init_transcriber()
         self.video_path: Optional[Path] = None
         self.video_img_urls=[]

backend/app/services/task_serial_executor.py

@@ -1,14 +1,23 @@
-import threading
+import os
+from concurrent.futures import ThreadPoolExecutor, Future
 from typing import Any, Callable
 
 
-class SerialTaskExecutor:
-    def __init__(self):
-        self._lock = threading.Lock()
+class ConcurrentTaskExecutor:
+    """使用线程池并发执行任务，替代原来的串行锁。"""
+
+    def __init__(self, max_workers: int | None = None):
+        self._max_workers = max_workers or int(os.getenv("TASK_MAX_WORKERS", "3"))
+        self._pool = ThreadPoolExecutor(max_workers=self._max_workers)
 
     def run(self, fn: Callable[..., Any], *args: Any, **kwargs: Any) -> Any:
-        with self._lock:
-            return fn(*args, **kwargs)
+        future: Future = self._pool.submit(fn, *args, **kwargs)
+        return future.result()
+
+    def shutdown(self, wait: bool = True):
+        self._pool.shutdown(wait=wait)
 
 
-task_serial_executor = SerialTaskExecutor()
+# 保持向后兼容的导出名
+SerialTaskExecutor = ConcurrentTaskExecutor
+task_serial_executor = ConcurrentTaskExecutor()

backend/app/services/transcriber_config_manager.py

@@ -0,0 +1,58 @@
+import json
+import os
+from pathlib import Path
+from typing import Optional, Dict, Any
+
+
+class TranscriberConfigManager:
+    """管理转写器配置，存储在 JSON 文件中，支持前端动态修改。"""
+
+    def __init__(self, filepath: str = "config/transcriber.json"):
+        self.path = Path(filepath)
+        self.path.parent.mkdir(parents=True, exist_ok=True)
+
+    def _read(self) -> Dict[str, Any]:
+        if not self.path.exists():
+            return {}
+        try:
+            with self.path.open("r", encoding="utf-8") as f:
+                return json.load(f)
+        except Exception:
+            return {}
+
+    def _write(self, data: Dict[str, Any]):
+        with self.path.open("w", encoding="utf-8") as f:
+            json.dump(data, f, ensure_ascii=False, indent=2)
+
+    def get_config(self) -> Dict[str, Any]:
+        """获取当前转写器配置，fallback 到环境变量默认值。"""
+        data = self._read()
+        return {
+            "transcriber_type": data.get(
+                "transcriber_type",
+                os.getenv("TRANSCRIBER_TYPE", "fast-whisper"),
+            ),
+            "whisper_model_size": data.get(
+                "whisper_model_size",
+                os.getenv("WHISPER_MODEL_SIZE", "medium"),
+            ),
+        }
+
+    def update_config(
+        self,
+        transcriber_type: str,
+        whisper_model_size: Optional[str] = None,
+    ) -> Dict[str, Any]:
+        """更新转写器配置并持久化。"""
+        data = self._read()
+        data["transcriber_type"] = transcriber_type
+        if whisper_model_size is not None:
+            data["whisper_model_size"] = whisper_model_size
+        self._write(data)
+        return self.get_config()
+
+    def get_transcriber_type(self) -> str:
+        return self.get_config()["transcriber_type"]
+
+    def get_whisper_model_size(self) -> str:
+        return self.get_config()["whisper_model_size"]

backend/app/transcriber/transcriber_provider.py

@@ -17,15 +17,15 @@ class TranscriberType(str, Enum):
     KUAISHOU = "kuaishou"
     GROQ = "groq"
 
-# 仅在 Apple 平台启用 MLX Whisper
+# 在 Apple 平台尝试导入 MLX Whisper（不再依赖环境变量，支持前端动态切换）
 MLX_WHISPER_AVAILABLE = False
-if platform.system() == "Darwin" and os.environ.get("TRANSCRIBER_TYPE") == "mlx-whisper":
+if platform.system() == "Darwin":
     try:
         from app.transcriber.mlx_whisper_transcriber import MLXWhisperTranscriber
         MLX_WHISPER_AVAILABLE = True
         logger.info("MLX Whisper 可用，已导入")
     except ImportError:
-        logger.warning("MLX Whisper 导入失败，可能未安装或平台不支持")
+        logger.warning("MLX Whisper 导入失败，可能未安装 mlx_whisper")
 
 logger.info('初始化转录服务提供器')
 
@@ -97,8 +97,10 @@ def get_transcriber(transcriber_type="fast-whisper", model_size="base", device="
 
     elif transcriber_enum == TranscriberType.MLX_WHISPER:
         if not MLX_WHISPER_AVAILABLE:
-            logger.warning("MLX Whisper 不可用，回退到 fast-whisper")
-            return get_whisper_transcriber(whisper_model_size, device=device)
+            raise RuntimeError(
+                "MLX Whisper 不可用：需要 macOS 平台并安装 mlx_whisper 包 (pip install mlx_whisper)。"
+                "请在「音频转写配置」页面切换到其他转写引擎。"
+            )
         return get_mlx_whisper_transcriber(whisper_model_size)
 
     elif transcriber_enum == TranscriberType.BCUT:

backend/app/utils/video_reader.py

@@ -3,6 +3,7 @@ import hashlib
 import os
 import re
 import subprocess
+from concurrent.futures import ThreadPoolExecutor, as_completed
 import ffmpeg
 from PIL import Image, ImageDraw, ImageFont
 
@@ -54,6 +55,18 @@ class VideoReader:
             return mm * 60 + ss
         return float('inf')
 
+    def _extract_single_frame(self, ts: int) -> str | None:
+        """提取单帧，返回输出路径或 None（失败时）。"""
+        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"]
+        try:
+            subprocess.run(cmd, check=True)
+            return output_path
+        except subprocess.CalledProcessError:
+            return None
+
     def extract_frames(self, max_frames=1000) -> list[str]:
 
         try:
@@ -61,14 +74,22 @@ class VideoReader:
             duration = float(ffmpeg.probe(self.video_path)["format"]["duration"])
             timestamps = [i for i in range(0, int(duration), self.frame_interval)][:max_frames]
 
+            # 并行提取帧
+            max_workers = min(os.cpu_count() or 4, 8, len(timestamps))
+            frame_results: dict[int, str | None] = {}
+            with ThreadPoolExecutor(max_workers=max_workers) as pool:
+                futures = {pool.submit(self._extract_single_frame, ts): ts for ts in timestamps}
+                for future in as_completed(futures):
+                    ts = futures[future]
+                    frame_results[ts] = future.result()
+
+            # 按时间戳顺序整理结果，并进行去重
             image_paths = []
             last_hash = None
             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)
+                output_path = frame_results.get(ts)
+                if not output_path or not os.path.exists(output_path):
+                    continue
 
                 if self.dedupe_enabled:
                     frame_hash = self._calculate_file_md5(output_path)

backend/main.py

@@ -4,6 +4,7 @@ from contextlib import asynccontextmanager
 import uvicorn
 from fastapi import FastAPI
 from starlette.middleware.cors import CORSMiddleware
+from starlette.middleware.gzip import GZipMiddleware
 from starlette.staticfiles import StaticFiles
 from dotenv import load_dotenv
 
@@ -14,7 +15,7 @@ from app.exceptions.exception_handlers import register_exception_handlers
 # from app.db.provider_dao import init_provider_table
 from app.utils.logger import get_logger
 from app import create_app
-from app.transcriber.transcriber_provider import get_transcriber
+from app.services.transcriber_config_manager import TranscriberConfigManager
 from events import register_handler
 from ffmpeg_helper import ensure_ffmpeg_or_raise
 
@@ -40,7 +41,10 @@ if not os.path.exists(out_dir):
 async def lifespan(app: FastAPI):
     register_handler()
     init_db()
-    get_transcriber(transcriber_type=os.getenv("TRANSCRIBER_TYPE", "fast-whisper"))
+    # 转写器不再在启动时强制初始化，而是在首次生成笔记时按需创建
+    # 如果配置了不可用的类型（如 mlx-whisper 未安装），会在使用时报错而非静默回退
+    _cfg = TranscriberConfigManager().get_config()
+    logger.info(f"当前转写器配置: type={_cfg['transcriber_type']}, model_size={_cfg['whisper_model_size']}")
     seed_default_providers()
     yield
 
@@ -58,6 +62,7 @@ app.add_middleware(
     allow_methods=["*"],
     allow_headers=["*"],
 )
+app.add_middleware(GZipMiddleware, minimum_size=1000)
 register_exception_handlers(app)
 app.mount(static_path, StaticFiles(directory=static_dir), name="static")
 app.mount("/uploads", StaticFiles(directory=uploads_dir), name="uploads")

docker-compose.yml

@@ -14,6 +14,14 @@ services:
       - ./backend:/app
     expose:
       - "${BACKEND_PORT}"  # 不再对外暴露，用于 nginx 内部通信
+    restart: on-failure:3
+    healthcheck:
+      test: ["CMD", "curl", "-f", "http://localhost:${BACKEND_PORT}/api/sys_health"]
+      interval: 30s
+      timeout: 10s
+      retries: 3
+      start_period: 15s
+    mem_limit: 4g
 
   frontend:
     container_name: bilinote-frontend
@@ -24,6 +32,8 @@ services:
       - .env
     expose:
       - "80"  # 不暴露给宿主机，只供 nginx 访问
+    restart: on-failure:3
+    mem_limit: 512m
 
   nginx:
     container_name: bilinote-nginx
@@ -33,5 +43,9 @@ services:
     volumes:
       - ./nginx/default.conf:/etc/nginx/conf.d/default.conf
     depends_on:
-      - backend
-      - frontend
+      backend:
+        condition: service_healthy
+      frontend:
+        condition: service_started
+    restart: on-failure:3
+    mem_limit: 256m

nginx/default.conf

@@ -1,6 +1,14 @@
 server {
   listen 80;
   client_max_body_size 10G;
+
+  # gzip 压缩
+  gzip on;
+  gzip_vary on;
+  gzip_min_length 1024;
+  gzip_proxied any;
+  gzip_types text/plain text/css application/json application/javascript text/xml application/xml application/xml+rss text/javascript image/svg+xml;
+
   # 所有非 /api 请求全部代理给 frontend 容器
   location / {
     proxy_pass http://frontend:80;
@@ -17,6 +25,7 @@ server {
     proxy_pass http://backend:8483/static/;
     proxy_set_header Host $host;
     proxy_set_header X-Real-IP $remote_addr;
-
+    expires 7d;
+    add_header Cache-Control "public, immutable";
   }
 }