Merge pull request #318 from linwinfan/feat/zustand-indexeddb

Merge pull request #333 from Lizhilin/fix/bilibili-cookie-injection
Merge pull request #336 from liang09255/fix-switch-0502
2026-05-06 20:42:52 +08:00 · 2026-05-05 14:34:18 +08:00 · 2026-05-05 14:32:32 +08:00 · 2026-05-05 14:32:22 +08:00 · 2026-05-02 21:51:33 +08:00 · 2026-04-28 23:20:34 +08:00
87 changed files with 23952 additions and 913 deletions
--- a/.dockerignore
+++ b/.dockerignore
@@ -1,321 +1,35 @@
-# Logs
-logs
-*.log
-npm-debug.log*
-yarn-debug.log*
-yarn-error.log*
-lerna-debug.log*
-.pnpm-debug.log*
+# Git 和 IDE
+.git
+.github
+.idea/
+.vscode/
 .DS_Store
-# Diagnostic reports (https://nodejs.org/api/report.html)
-report.[0-9]*.[0-9]*.[0-9]*.[0-9]*.json
-BiliNote/pnpm-lock.yaml
-# Runtime data
-pids
-*.pid
-*.seed
-*.pid.lock

-# Directory for instrumented libs generated by jscoverage/JSCover
-lib-cov
+# Tauri 构建产物（非常大）
+BillNote_frontend/src-tauri/target
+BillNote_frontend/src-tauri/bin

-# Coverage directory used by tools like istanbul
-coverage
-*.lcov
+# 运行时数据
+backend/data
+backend/static
+backend/models
+backend/logs
+backend/uploads
+backend/*.db
+backend/note_results
+backend/bin/

-# nyc test coverage
-.nyc_output
-
-# Grunt intermediate storage (https://gruntjs.com/creating-plugins#storing-task-files)
-.grunt
-
-# Bower dependency directory (https://bower.io/)
-bower_components
-
-# node-waf configuration
-.lock-wscript
-
-# Compiled binary addons (https://nodejs.org/api/addons.html)
-build/Release
-
-# Dependency directories
+# 依赖和构建缓存
 node_modules/
-jspm_packages/
-
-# Snowpack dependency directory (https://snowpack.dev/)
-web_modules/
-
-# TypeScript cache
-*.tsbuildinfo
-
-# Optional npm cache directory
-.npm
-
-# Optional eslint cache
-.eslintcache
-
-# Optional stylelint cache
-.stylelintcache
-
-# Microbundle cache
-.rpt2_cache/
-.rts2_cache_cjs/
-.rts2_cache_es/
-.rts2_cache_umd/
-.BiliNote-dev/*
-# Optional REPL history
-.node_repl_history
-
-# Output of 'npm pack'
-*.tgz
-
-# Yarn Integrity file
-.yarn-integrity
-
-# dotenv environment variable files
-.env
-.env.development.local
-.env.test.local
-.env.production.local
-.env.local
-!.env.example
-# parcel-bundler cache (https://parceljs.org/)
-.cache
-.parcel-cache
-
-# Next.js build output
-.next
-out
-
-# Nuxt.js build / generate output
-.nuxt
-dist
-
-# Gatsby files
-.cache/
-# Comment in the public line in if your project uses Gatsby and not Next.js
-# https://nextjs.org/blog/next-9-1#public-directory-support
-# public
-
-# vuepress build output
-.vuepress/dist
-
-# vuepress v2.x temp and cache directory
-.temp
-.cache
-
-# vitepress build output
-**/.vitepress/dist
-
-# vitepress cache directory
-**/.vitepress/cache
-
-# Docusaurus cache and generated files
-.docusaurus
-
-# Serverless directories
-.serverless/
-
-# FuseBox cache
-.fusebox/
-
-# DynamoDB Local files
-.dynamodb/
-
-# TernJS port file
-.tern-port
-
-# Stores VSCode versions used for testing VSCode extensions
-.vscode-test
-
-# yarn v2
-.yarn/cache
-.yarn/unplugged
-.yarn/build-state.yml
-.yarn/install-state.gz
-.pnp.*
-# Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]
-*$py.class
-
-# C extensions
-*.so
-
-# Distribution / packaging
-.Python
-build/
-develop-eggs/
 dist/
-downloads/
-eggs/
-.eggs/
-
-lib64/
-parts/
-sdist/
-var/
-wheels/
-share/python-wheels/
+build/
+*.tar
 *.egg-info/
-.installed.cfg
-*.egg
-MANIFEST

-# PyInstaller
-#  Usually these files are written by a python script from a template
-#  before PyInstaller builds the exe, so as to inject date/other infos into it.
-*.manifest
-*.spec
-
-# Installer logs
-pip-log.txt
-pip-delete-this-directory.txt
-
-# Unit test / coverage reports
-htmlcov/
-.tox/
-.nox/
-.coverage
-.coverage.*
-.cache
-nosetests.xml
-coverage.xml
-*.cover
-*.py,cover
-.hypothesis/
-.pytest_cache/
-cover/
-
-# Translations
-*.mo
-*.pot
-.idea/
-# Django stuff:
-*.log
-local_settings.py
-db.sqlite3
-db.sqlite3-journal
-
-# Flask stuff:
-instance/
-.webassets-cache
-
-# Scrapy stuff:
-.scrapy
-
-# Sphinx documentation
-docs/_build/
-
-# PyBuilder
-.pybuilder/
-target/
-
-# Jupyter Notebook
-.ipynb_checkpoints
-
-# IPython
-profile_default/
-ipython_config.py
-
-# pyenv
-#   For a library or package, you might want to ignore these files since the code is
-#   intended to run in multiple environments; otherwise, check them in:
-# .python-version
-
-# pipenv
-#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
-#   However, in case of collaboration, if having platform-specific dependencies or dependencies
-#   having no cross-platform support, pipenv may install dependencies that don't work, or not
-#   install all needed dependencies.
-#Pipfile.lock
-
-# UV
-#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#uv.lock
-
-# poetry
-#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
-#   This is especially recommended for binary packages to ensure reproducibility, and is more
-#   commonly ignored for libraries.
-#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
-#poetry.lock
-
-# pdm
-#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
-#pdm.lock
-#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
-#   in version control.
-#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
-.pdm.toml
-.pdm-python
-.pdm-build/
-
-# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
-__pypackages__/
-
-# Celery stuff
-celerybeat-schedule
-celerybeat.pid
-
-# SageMath parsed files
-*.sage.py
-
-# Environments
+# 环境文件
 .env
-.venv
-env/
-venv/
-ENV/
-env.bak/
-venv.bak/
-
-# Spyder project settings
-.spyderproject
-.spyproject
-
-# Rope project settings
-.ropeproject
-
-# mkdocs documentation
-/site
-
-# mypy
-.mypy_cache/
-.dmypy.json
-dmypy.json
-
-# Pyre type checker
-.pyre/
-
-# pytype static type analyzer
-.pytype/
-
-# Cython debug symbols
-cython_debug/
-
-# PyCharm
-#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
-#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
-#  and can be added to the global gitignore or merged into this file.  For a more nuclear
-#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
-#.idea/
-
-# Ruff stuff:
-.ruff_cache/
-
-# PyPI configuration file
-.pypirc
-/backend/data/*
-/backend/static/*
-/backend/note_tasks.db
-/backend/bin/
-/backend/logs/
-/backend/note_results
-/backend/models
-/backend/.idea/*
-/backend/bili_note.db
-/backend/uploads/*
-/BiliNote_frontend/.idea/*
+.env.local
+.env.*.local
+!.env.example
--- a/.env.example
+++ b/.env.example
@@ -4,7 +4,7 @@ FRONTEND_PORT=3015
 BACKEND_HOST=0.0.0.0 # 默认为 0.0.0.0，表示监听所有 IP 地址 不建议动
 APP_PORT= 3015 # docker 部署时用
 # 前端访问后端用 (开发环境使用)
-VITE_API_BASE_URL=http://127.0.0.1:8483
+VITE_API_BASE_URL=http://127.0.0.1:8000
 VITE_SCREENSHOT_BASE_URL=http://127.0.0.1:8483/static/screenshots
 VITE_FRONTEND_PORT=3015
 # 生产环境配置
@@ -19,6 +19,6 @@ FFMPEG_BIN_PATH=

 # transcriber 相关配置
 TRANSCRIBER_TYPE=fast-whisper # fast-whisper/bcut/kuaishou/mlx-whisper(仅Apple平台)/groq
-WHISPER_MODEL_SIZE=base
+WHISPER_MODEL_SIZE=medium

 GROQ_TRANSCRIBER_MODEL=whisper-large-v3-turbo # groq提供的faster-whisper 默认为 whisper-large-v3-turbo
--- a/.github/workflows/docker-build.yml
+++ b/.github/workflows/docker-build.yml
@@ -0,0 +1,73 @@
+name: Build and Publish Docker Image
+
+on:
+  push:
+    tags:
+      - 'v*'
+  workflow_dispatch:
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+
+    steps:
+      - name: Checkout Code
+        uses: actions/checkout@v4
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Log in to GitHub Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Extract metadata (tags, labels)
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+          tags: |
+            type=semver,pattern={{version}}
+            type=semver,pattern={{major}}.{{minor}}
+            type=sha,prefix=
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Build and Push Docker Image
+        uses: docker/build-push-action@v6
+        with:
+          context: .
+          file: Dockerfile.complete
+          push: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}
+          cache-from: type=gha
+          cache-to: type=gha,mode=max
+          platforms: linux/amd64,linux/arm64
+
+      - name: Generate Usage Instructions
+        run: |
+          echo "=========================================="
+          echo "Docker Image Published!"
+          echo "=========================================="
+          echo ""
+          echo "Pull the image:"
+          echo "  docker pull ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest"
+          echo ""
+          echo "Run the container:"
+          echo "  docker run -d -p 80:80 \\"
+          echo "    -v bilinote-data:/app/backend/data \\"
+          echo "    --name bilinote \\"
+          echo "    ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:latest"
+          echo ""
+          echo "Access the application at: http://localhost"
+          echo "=========================================="
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -1,30 +1,45 @@
-# .github/workflows/release.yml
-name: Build Desktop App (Python Backend + Tauri Frontend)
+name: Build & Release Desktop App

 on:
  push:
    tags:
-      - 'v*' # 发布 tag 时触发
+      - 'v*'
  workflow_dispatch:
+
 jobs:
  build:
    strategy:
      matrix:
-        platform: [macos-latest, windows-latest]
+        include:
+          - platform: macos-latest
+            target: universal-apple-darwin
+          - platform: ubuntu-22.04
+            target: x86_64-unknown-linux-gnu
+          - platform: windows-latest
+            target: x86_64-pc-windows-msvc

    runs-on: ${{ matrix.platform }}

    steps:
      - name: Checkout Code
-        uses: actions/checkout@v3
+        uses: actions/checkout@v4

-      # 设置 Python 环境
+      # Linux 系统依赖（Tauri 需要）
+      - name: Install Linux Dependencies
+        if: matrix.platform == 'ubuntu-22.04'
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y libwebkit2gtk-4.1-dev libappindicator3-dev librsvg2-dev patchelf
+
+      # 设置 Python 环境（带 pip 缓存）
      - name: Set up Python
-        uses: actions/setup-python@v4
+        uses: actions/setup-python@v5
        with:
-          python-version: '3.10'
+          python-version: '3.11'
+          cache: 'pip'
+          cache-dependency-path: backend/requirements.txt

-      # 安装 Python 依赖并执行你的 build.sh
+      # 安装 Python 依赖并执行构建
      - name: Install Python dependencies & Build backend
        shell: bash
        run: |
@@ -38,30 +53,108 @@ jobs:
            ./backend/build.sh
          fi

-      # 设置 Node 环境 + 安装前端依赖
+      # 设置 pnpm
+      - name: Setup pnpm
+        uses: pnpm/action-setup@v4
+        with:
+          version: 'latest'
+
+      # 设置 Node 环境
      - name: Set up Node.js
-        uses: actions/setup-node@v3
+        uses: actions/setup-node@v4
        with:
          node-version: '20'

-      - name: Enable Corepack + Install pnpm
+      - name: Install frontend dependencies
        working-directory: BillNote_frontend
-        run: |
-          corepack enable
-          pnpm install 
+        run: pnpm install

      # 设置 Rust 环境
      - name: Set up Rust
        uses: dtolnay/rust-toolchain@stable

+      # Cargo 缓存
+      - name: Cache Cargo
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cargo/bin/
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+            BillNote_frontend/src-tauri/target/
+          key: ${{ runner.os }}-cargo-${{ hashFiles('BillNote_frontend/src-tauri/Cargo.lock') }}
+          restore-keys: ${{ runner.os }}-cargo-
+
      # 打包 Tauri 应用
      - name: Build Tauri App
        working-directory: BillNote_frontend
        run: pnpm tauri build

-      # 可选：上传构建产物
-      - name: Upload Desktop Bundle
+      # 收集产物到统一目录
+      - name: Collect release artifacts
+        shell: bash
+        run: |
+          mkdir -p release-artifacts
+          BUNDLE_DIR="BillNote_frontend/src-tauri/target/release/bundle"
+
+          # macOS: .dmg
+          find "$BUNDLE_DIR" -name "*.dmg" -exec cp {} release-artifacts/ \; 2>/dev/null || true
+          # Windows: .msi, .exe (NSIS)
+          find "$BUNDLE_DIR" -name "*.msi" -exec cp {} release-artifacts/ \; 2>/dev/null || true
+          find "$BUNDLE_DIR/nsis" -name "*.exe" -exec cp {} release-artifacts/ \; 2>/dev/null || true
+          # Linux: .deb, .AppImage
+          find "$BUNDLE_DIR" -name "*.deb" -exec cp {} release-artifacts/ \; 2>/dev/null || true
+          find "$BUNDLE_DIR" -name "*.AppImage" -exec cp {} release-artifacts/ \; 2>/dev/null || true
+
+          echo "=== Collected artifacts ==="
+          ls -lh release-artifacts/
+
+      # 生成 SHA256 校验和
+      - name: Generate checksums
+        shell: bash
+        run: |
+          cd release-artifacts
+          sha256sum * > SHA256SUMS.txt 2>/dev/null || shasum -a 256 * > SHA256SUMS.txt
+          cat SHA256SUMS.txt
+
+      # 上传产物（供 release job 使用）
+      - name: Upload artifacts
        uses: actions/upload-artifact@v4
        with:
-          name: app-${{ matrix.platform }}
-          path: BillNote_frontend/src-tauri/target/release/bundle/
+          name: artifacts-${{ matrix.platform }}
+          path: release-artifacts/
+
+  # 创建 GitHub Release 并上传所有产物
+  release:
+    needs: build
+    runs-on: ubuntu-latest
+    if: startsWith(github.ref, 'refs/tags/v')
+    permissions:
+      contents: write
+
+    steps:
+      - name: Checkout Code
+        uses: actions/checkout@v4
+
+      # 下载所有平台的构建产物
+      - name: Download all artifacts
+        uses: actions/download-artifact@v4
+        with:
+          path: all-artifacts
+          merge-multiple: true
+
+      - name: List all artifacts
+        run: |
+          echo "=== All release artifacts ==="
+          ls -lhR all-artifacts/
+
+      # 创建 Release 并上传产物
+      - name: Create GitHub Release
+        uses: softprops/action-gh-release@v2
+        with:
+          name: BiliNote ${{ github.ref_name }}
+          draft: false
+          prerelease: false
+          generate_release_notes: true
+          files: all-artifacts/*
--- a/.gitignore
+++ b/.gitignore
@@ -320,5 +320,10 @@ cython_debug/
 /backend/uploads/*
 /backend/.idea/*
 /backend/config/*
+/backend/vector_db/
 /BiliNote_frontend/.idea/*
-/BiliNote_frontend/src-tauri/bin/
+/BiliNote_frontend/src-tauri/bin/
+
+# FFmpeg 构建文件（不应该提交到仓库）
+ffmpeg*/
+ffmpg*/
--- a/.vscode/settings.json
+++ b/.vscode/settings.json
@@ -0,0 +1 @@
+{}
--- a/BillNote_frontend/.gitignore
+++ b/BillNote_frontend/.gitignore
@@ -22,5 +22,4 @@ dist-ssr
 *.njsproj
 *.sln
 *.sw?
-/pnpm-lock.yaml
 /src-tauri/bin/
--- a/BillNote_frontend/Dockerfile
+++ b/BillNote_frontend/Dockerfile
@@ -1,25 +1,23 @@
 # === 前端构建阶段 ===
 FROM node:18-alpine AS builder

-# 安装 pnpm
-RUN npm install -g pnpm
+RUN corepack enable && corepack prepare pnpm@latest --activate

-# 设置工作目录
 WORKDIR /app

-# 拷贝前端源码
-COPY ./BillNote_frontend /app
+# 先复制 lockfile 利用依赖层缓存
+COPY ./BillNote_frontend/package.json ./BillNote_frontend/pnpm-lock.yaml ./
+RUN pnpm install --frozen-lockfile

-# 安装依赖并构建
-RUN pnpm install && pnpm run build
+# 再复制源代码并构建
+COPY ./BillNote_frontend/ ./
+RUN pnpm run build

 # --- 阶段2：使用 nginx 作为静态服务器 ---
 FROM nginx:1.25-alpine

-# 删除默认配置（可选）
 RUN rm -rf /etc/nginx/conf.d/default.conf
 COPY ./BillNote_frontend/deploy/default.conf /etc/nginx/conf.d/default.conf

-
 # 拷贝构建产物
-COPY --from=builder /app/dist /usr/share/nginx/html
+COPY --from=builder /app/dist /usr/share/nginx/html
--- a/BillNote_frontend/package-lock.json
+++ b/BillNote_frontend/package-lock.json
--- a/BillNote_frontend/package.json
+++ b/BillNote_frontend/package.json
@@ -10,6 +10,7 @@
    "preview": "vite preview"
  },
  "dependencies": {
+    "@ant-design/x": "^2.4.0",
    "@hookform/resolvers": "^5.0.1",
    "@lobehub/icons": "^1.97.1",
    "@lobehub/icons-static-svg": "^1.45.0",
@@ -32,6 +33,8 @@
    "clsx": "^2.1.1",
    "fuse.js": "^7.1.0",
    "github-markdown-css": "^5.8.1",
+    "idb-keyval": "^6.2.2",
+    "jszip": "^3.10.1",
    "katex": "^0.16.22",
    "lottie-react": "^2.4.1",
    "lucide-react": "^0.487.0",
--- a/BillNote_frontend/src-tauri/tauri.conf.json
+++ b/BillNote_frontend/src-tauri/tauri.conf.json
@@ -1,7 +1,7 @@
 {
  "$schema": "../node_modules/@tauri-apps/cli/config.schema.json",
  "productName": "BiliNote",
-  "version": "1.8.1",
+  "version": "2.0.0",
  "identifier": "com.jefferyhuang.bilinote",
  "build": {
    "frontendDist": "../dist",
--- a/BillNote_frontend/src/App.tsx
+++ b/BillNote_frontend/src/App.tsx
@@ -1,24 +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, HashRouter, 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 Transcriber from '@/pages/SettingPage/transcriber.tsx'
-import ProviderForm from '@/components/Form/modelForm/Form.tsx'
-import StepBar from '@/pages/HomePage/components/StepBar.tsx'
-import Downloading from '@/components/Lottie/download.tsx'
-import Prompt from '@/pages/SettingPage/Prompt.tsx'
-import AboutPage from '@/pages/SettingPage/about.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
@@ -43,28 +42,32 @@ function App() {
  // 后端已初始化，渲染主应用
  return (
    <>
-      <HashRouter>
-        <Routes>
-          <Route path="/" element={<Index />}>
-            <Route index element={<HomePage />} />
-            <Route path="settings" element={<SettingPage />}>
-              <Route index element={<Navigate to="model" replace />} />
-              <Route path="model" element={<Model />}>
-                <Route path="new" element={<ProviderForm isCreate />} />
-                <Route path=":id" element={<ProviderForm />} />
+      <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 />} />
+              <Route path="settings" element={<SettingPage />}>
+                <Route index element={<Navigate to="model" replace />} />
+                <Route path="model" element={<Model />}>
+                  <Route path="new" element={<ProviderForm isCreate />} />
+                  <Route path=":id" element={<ProviderForm />} />
+                </Route>
+                <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 />} />
              </Route>
-              <Route path="download" element={<Downloader />}>
-                <Route path=":id" element={<DownloaderForm />} />
-              </Route>
-              <Route path="about" element={<AboutPage />}></Route>
              <Route path="*" element={<NotFoundPage />} />
            </Route>
-            <Route path="*" element={<NotFoundPage />} />
-          </Route>
-        </Routes>
-      </HashRouter>
+          </Routes>
+        </Suspense>
+      </BrowserRouter>
    </>
  )
 }

-export default App
+export default App
--- a/BillNote_frontend/src/components/Form/modelForm/Form.tsx
+++ b/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'}>
-                      {model.model_name}
-                    </Tag></>
+                  <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}
+                    <button type="button" onClick={() => handelDelete(model.id)} className="hover:text-blue-900">
+                      <X className="h-3 w-3" />
+                    </button>
+                  </span>

                )
              })
--- a/BillNote_frontend/src/components/Form/modelForm/components/providerCard.tsx
+++ b/BillNote_frontend/src/components/Form/modelForm/components/providerCard.tsx
@@ -4,47 +4,51 @@ import styles from './index.module.css'
 import { useNavigate, useParams } from 'react-router-dom'
 import AILogo from '@/components/Form/modelForm/Icons'
 import { useProviderStore } from '@/store/providerStore'
+
 export interface IProviderCardProps {
  id: string
  providerName: string
  Icon: string
  enable: number
 }
+
 const ProviderCard: FC<IProviderCardProps> = ({
  providerName,
  Icon,
  id,
-  enable,
 }: IProviderCardProps) => {
  const navigate = useNavigate()
  const updateProvider = useProviderStore(state => state.updateProvider)
-  const handleClick = () => {
-    navigate(`/settings/model/${id}`)
-  }
-  const handleEnable = () => {
-    console.log('enable', enable)
+  const enabled = useProviderStore(state => state.provider.find(p => p.id === id)?.enabled)
+
+  const isChecked = enabled === 1
+
+  const handleToggle = (checked: boolean) => {
+    const allProviders = useProviderStore.getState().provider
+    const provider = allProviders.find(p => p.id === id)
+    if (!provider) return
    updateProvider({
-      id,
-      enabled: enable == 1 ? 0 : 1,
+      ...provider,
+      enabled: checked ? 1 : 0,
    })
  }
-  const rawId = useParams()
-  console.log('rawId', rawId)
+
  // @ts-ignore
  const { id: currentId } = useParams()
  const isActive = currentId === id
+
  return (
    <div
-      onClick={() => {
-        handleClick()
-      }}
      className={
        styles.card +
        ' flex h-14 items-center justify-between rounded border border-[#f3f3f3] p-2' +
        (isActive ? ' bg-[#F0F0F0] font-semibold text-blue-600' : '')
      }
    >
-      <div className="flex items-center text-lg">
+      <div
+        className="flex items-center text-lg"
+        onClick={() => navigate(`/settings/model/${id}`)}
+      >
        <div className="flex h-9 w-9 items-center">
          <AILogo name={Icon} />
        </div>
@@ -53,11 +57,8 @@ const ProviderCard: FC<IProviderCardProps> = ({

      <div>
        <Switch
-          onClick={e => {
-            e.preventDefault()
-            handleEnable()
-          }}
-          checked={enable == 1}
+          checked={isChecked}
+          onCheckedChange={handleToggle}
        />
      </div>
    </div>
--- a/BillNote_frontend/src/components/ui/checkbox.tsx
+++ b/BillNote_frontend/src/components/ui/checkbox.tsx
@@ -1,10 +1,26 @@
 import * as React from 'react'
+import { useState, useEffect } from 'react';
 import * as CheckboxPrimitive from '@radix-ui/react-checkbox'
 import { CheckIcon } from 'lucide-react'

 import { cn } from '@/lib/utils'

-function Checkbox({ className, ...props }: React.ComponentProps<typeof CheckboxPrimitive.Root>) {
+function Checkbox({ className, checked, onChange, ...props }: React.ComponentProps<typeof CheckboxPrimitive.Root>) {
+  const [isChecked, setIsChecked] = useState(checked || false);
+  
+  useEffect(() => {
+    if (checked !== undefined) {
+      setIsChecked(checked);
+    }
+  }, [checked]);
+
+  const handleCheckChange = (newChecked: boolean) => {
+    setIsChecked(newChecked);
+    if (onChange) {
+      onChange({} as React.FormEvent<HTMLButtonElement>);
+    }
+  };
+  
  return (
    <CheckboxPrimitive.Root
      data-slot="checkbox"
@@ -12,6 +28,8 @@ function Checkbox({ className, ...props }: React.ComponentProps<typeof CheckboxP
        'peer border-input dark:bg-input/30 data-[state=checked]:bg-primary data-[state=checked]:text-primary-foreground dark:data-[state=checked]:bg-primary data-[state=checked]:border-primary focus-visible:border-ring focus-visible:ring-ring/50 aria-invalid:ring-destructive/20 dark:aria-invalid:ring-destructive/40 aria-invalid:border-destructive size-4 shrink-0 rounded-[4px] border shadow-xs transition-shadow outline-none focus-visible:ring-[3px] disabled:cursor-not-allowed disabled:opacity-50',
        className
      )}
+      checked={isChecked}
+      onCheckedChange={handleCheckChange}
      {...props}
    >
      <CheckboxPrimitive.Indicator
--- a/BillNote_frontend/src/components/ui/input.tsx
+++ b/BillNote_frontend/src/components/ui/input.tsx
@@ -2,7 +2,7 @@ import * as React from 'react'

 import { cn } from '@/lib/utils'

-function Input({ className, type, ...props }: React.ComponentProps<'input'>) {
+function Input({ className, type, value, onChange, ...props }: React.ComponentProps<'input'>) {
  return (
    <input
      type={type}
@@ -13,6 +13,8 @@ function Input({ className, type, ...props }: React.ComponentProps<'input'>) {
        'aria-invalid:ring-destructive/20 dark:aria-invalid:ring-destructive/40 aria-invalid:border-destructive',
        className
      )}
+      value={value ?? ''}
+      onChange={onChange}
      {...props}
    />
  )
--- a/BillNote_frontend/src/hooks/useTaskPolling.ts
+++ b/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)
--- a/BillNote_frontend/src/layouts/HomeLayout.tsx
+++ b/BillNote_frontend/src/layouts/HomeLayout.tsx
@@ -1,5 +1,5 @@
-import React, { FC } from 'react'
-import { SlidersHorizontal } from 'lucide-react'
+import React, { FC, useRef, useState } from 'react'
+import { SlidersHorizontal, PanelLeftClose, PanelLeftOpen, History as HistoryIcon } from 'lucide-react'
 import {
  Tooltip,
  TooltipContent,
@@ -7,24 +7,39 @@ import {
  TooltipTrigger,
 } from '@/components/ui/tooltip.tsx'

-import { useState } from 'react'
 import { Link } from 'react-router-dom'
 import { ResizablePanel, ResizablePanelGroup, ResizableHandle } from '@/components/ui/resizable'
-import {ScrollArea} from "@/components/ui/scroll-area.tsx";
+import { ScrollArea } from "@/components/ui/scroll-area.tsx"
+import type { ImperativePanelHandle } from 'react-resizable-panels'
 import logo from '@/assets/icon.svg'
+
 interface IProps {
  NoteForm: React.ReactNode
  Preview: React.ReactNode
  History: React.ReactNode
 }
+
 const HomeLayout: FC<IProps> = ({ NoteForm, Preview, History }) => {
  const [, setShowSettings] = useState(false)
+  const [isLeftCollapsed, setIsLeftCollapsed] = useState(false)
+  const [isMiddleCollapsed, setIsMiddleCollapsed] = useState(false)
+  const leftPanelRef = useRef<ImperativePanelHandle>(null)
+  const middlePanelRef = useRef<ImperativePanelHandle>(null)

  return (
    <div className="flex h-screen flex-col overflow-hidden">
      <ResizablePanelGroup direction="horizontal" className="h-full w-full">
        {/* 左边表单 */}
-        <ResizablePanel defaultSize={18} minSize={10} maxSize={35}>
+        <ResizablePanel
+          ref={leftPanelRef}
+          defaultSize={23}
+          minSize={10}
+          maxSize={35}
+          collapsible
+          collapsedSize={0}
+          onCollapse={() => setIsLeftCollapsed(true)}
+          onExpand={() => setIsLeftCollapsed(false)}
+        >
          <aside className="flex h-full flex-col overflow-hidden border-r border-neutral-200 bg-white">
            <header className="flex h-16 items-center justify-between px-6">
              <div className="flex items-center gap-2">
@@ -33,7 +48,22 @@ const HomeLayout: FC<IProps> = ({ NoteForm, Preview, History }) => {
                </div>
                <div className="text-2xl font-bold text-gray-800">BiliNote</div>
              </div>
-              <div>
+              <div className="flex items-center gap-1">
+                <TooltipProvider>
+                  <Tooltip>
+                    <TooltipTrigger asChild>
+                      <button
+                        onClick={() => leftPanelRef.current?.collapse()}
+                        className="text-muted-foreground hover:text-primary cursor-pointer rounded p-1 hover:bg-neutral-100"
+                      >
+                        <PanelLeftClose className="h-5 w-5" />
+                      </button>
+                    </TooltipTrigger>
+                    <TooltipContent>
+                      <span>收起工作区</span>
+                    </TooltipContent>
+                  </Tooltip>
+                </TooltipProvider>
                <TooltipProvider>
                  <Tooltip>
                    <TooltipTrigger onClick={() => setShowSettings(true)}>
@@ -49,26 +79,91 @@ const HomeLayout: FC<IProps> = ({ NoteForm, Preview, History }) => {
              </div>
            </header>
            <ScrollArea className="flex-1 overflow-auto">
-              <div className=' p-4' >{NoteForm}</div>
+              <div className="p-4">{NoteForm}</div>
            </ScrollArea>
          </aside>
        </ResizablePanel>

        <ResizableHandle />

+        {/* 左面板折叠时的展开按钮 */}
+        {isLeftCollapsed && (
+          <TooltipProvider>
+            <Tooltip>
+              <TooltipTrigger asChild>
+                <button
+                  onClick={() => leftPanelRef.current?.expand()}
+                  className="flex h-full w-8 shrink-0 items-center justify-center border-r border-neutral-200 bg-white hover:bg-neutral-50"
+                >
+                  <PanelLeftOpen className="h-4 w-4 text-muted-foreground" />
+                </button>
+              </TooltipTrigger>
+              <TooltipContent side="right">
+                <span>展开工作区</span>
+              </TooltipContent>
+            </Tooltip>
+          </TooltipProvider>
+        )}
+
        {/* 中间历史 */}
-        <ResizablePanel defaultSize={16} minSize={10} maxSize={30}>
+        <ResizablePanel
+          ref={middlePanelRef}
+          defaultSize={16}
+          minSize={10}
+          maxSize={30}
+          collapsible
+          collapsedSize={0}
+          onCollapse={() => setIsMiddleCollapsed(true)}
+          onExpand={() => setIsMiddleCollapsed(false)}
+        >
          <aside className="flex h-full flex-col overflow-hidden border-r border-neutral-200 bg-white">
+            <header className="flex h-10 shrink-0 items-center justify-between border-b border-neutral-100 px-3">
+              <span className="text-sm font-medium text-gray-600">生成历史</span>
+              <TooltipProvider>
+                <Tooltip>
+                  <TooltipTrigger asChild>
+                    <button
+                      onClick={() => middlePanelRef.current?.collapse()}
+                      className="text-muted-foreground hover:text-primary cursor-pointer rounded p-1 hover:bg-neutral-100"
+                    >
+                      <PanelLeftClose className="h-4 w-4" />
+                    </button>
+                  </TooltipTrigger>
+                  <TooltipContent>
+                    <span>收起历史</span>
+                  </TooltipContent>
+                </Tooltip>
+              </TooltipProvider>
+            </header>
            <ScrollArea className="flex-1 overflow-auto">
-            <div className="">{History}</div>
+              <div>{History}</div>
            </ScrollArea>
          </aside>
        </ResizablePanel>

        <ResizableHandle />

+        {/* 中间面板折叠时的展开按钮 */}
+        {isMiddleCollapsed && (
+          <TooltipProvider>
+            <Tooltip>
+              <TooltipTrigger asChild>
+                <button
+                  onClick={() => middlePanelRef.current?.expand()}
+                  className="flex h-full w-8 shrink-0 items-center justify-center border-r border-neutral-200 bg-white hover:bg-neutral-50"
+                >
+                  <HistoryIcon className="h-4 w-4 text-muted-foreground" />
+                </button>
+              </TooltipTrigger>
+              <TooltipContent side="right">
+                <span>展开历史</span>
+              </TooltipContent>
+            </Tooltip>
+          </TooltipProvider>
+        )}
+
        {/* 右边预览 */}
-        <ResizablePanel defaultSize={55} minSize={30}>
+        <ResizablePanel defaultSize={61} minSize={30}>
          <main className="flex h-full flex-col overflow-hidden bg-white p-6">{Preview}</main>
        </ResizablePanel>
      </ResizablePanelGroup>
--- a/BillNote_frontend/src/pages/HomePage/Home.tsx
+++ b/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=>{
--- a/BillNote_frontend/src/pages/HomePage/components/ChatPanel.tsx
+++ b/BillNote_frontend/src/pages/HomePage/components/ChatPanel.tsx
@@ -0,0 +1,294 @@
+import { useState, useEffect, useCallback, useMemo } from 'react'
+import { Bubble, Sender } from '@ant-design/x'
+import ReactMarkdown from 'react-markdown'
+import remarkGfm from 'remark-gfm'
+import { Button } from '@/components/ui/button'
+import { Badge } from '@/components/ui/badge'
+import { Loader2, Trash2, ChevronDown, ChevronUp, BookOpen, UserRound, Bot, Maximize2, Minimize2 } from 'lucide-react'
+import { toast } from 'react-hot-toast'
+import { useChatStore } from '@/store/chatStore'
+import { useTaskStore } from '@/store/taskStore'
+import { askQuestion, getChatStatus, indexTask, type ChatSource, type IndexStatus } from '@/services/chat'
+
+type ChatMode = 'half' | 'full'
+
+interface ChatPanelProps {
+  taskId: string
+  mode: ChatMode
+  onModeChange: (mode: ChatMode) => void
+}
+
+function SourceBadges({ sources }: { sources: ChatSource[] }) {
+  const [expanded, setExpanded] = useState(false)
+
+  if (!sources || sources.length === 0) return null
+
+  return (
+    <div className="mt-1.5">
+      <button
+        onClick={() => setExpanded(!expanded)}
+        className="flex items-center gap-1 text-xs text-neutral-400 hover:text-neutral-600"
+      >
+        <BookOpen className="h-3 w-3" />
+        <span>引用来源 ({sources.length})</span>
+        {expanded ? <ChevronUp className="h-3 w-3" /> : <ChevronDown className="h-3 w-3" />}
+      </button>
+      {expanded && (
+        <div className="mt-1 flex flex-wrap gap-1">
+          {sources.map((s, i) => (
+            <Badge key={i} variant="outline" className="text-xs font-normal">
+              {s.source_type === 'markdown'
+                ? s.section_title || '笔记'
+                : `${(s.start_time ?? 0).toFixed(0)}s ~ ${(s.end_time ?? 0).toFixed(0)}s`}
+            </Badge>
+          ))}
+        </div>
+      )}
+    </div>
+  )
+}
+
+export default function ChatPanel({ taskId, mode, onModeChange }: ChatPanelProps) {
+  const [input, setInput] = useState('')
+  const [loading, setLoading] = useState(false)
+  const [indexStatus, setIndexStatus] = useState<IndexStatus | null>(null)
+
+  const messages = useChatStore(state => state.chatHistory[taskId]) ?? []
+  const addMessage = useChatStore(state => state.addMessage)
+  const clearChat = useChatStore(state => state.clearChat)
+
+  const currentTaskId = useTaskStore(state => state.currentTaskId)
+  const tasks = useTaskStore(state => state.tasks)
+  const currentTask = useMemo(
+    () => tasks.find(t => t.id === currentTaskId) ?? null,
+    [tasks, currentTaskId],
+  )
+
+  // 检查索引状态，未索引时自动触发，indexing 时轮询
+  useEffect(() => {
+    if (!taskId) return
+    let cancelled = false
+    let timer: ReturnType<typeof setTimeout> | null = null
+
+    const poll = async () => {
+      try {
+        const res = await getChatStatus(taskId)
+        if (cancelled) return
+        setIndexStatus(res.status)
+
+        if (res.status === 'idle') {
+          // 未索引，触发后台索引
+          await indexTask(taskId)
+          if (!cancelled) setIndexStatus('indexing')
+        }
+
+        // indexing 状态持续轮询
+        if (res.status === 'indexing' || res.status === 'idle') {
+          timer = setTimeout(poll, 2000)
+        }
+      } catch {
+        if (!cancelled) setIndexStatus('failed')
+      }
+    }
+
+    poll()
+    return () => {
+      cancelled = true
+      if (timer) clearTimeout(timer)
+    }
+  }, [taskId])
+
+  const handleSend = useCallback(
+    async (value: string) => {
+      const question = value.trim()
+      if (!question || loading) return
+
+      const providerId = currentTask?.formData?.provider_id
+      const modelName = currentTask?.formData?.model_name
+      if (!providerId || !modelName) {
+        toast.error('无法获取模型配置，请确认任务已完成')
+        return
+      }
+
+      addMessage(taskId, { role: 'user', content: question })
+      setInput('')
+      setLoading(true)
+
+      try {
+        const history = messages.map(m => ({ role: m.role, content: m.content }))
+        const res = await askQuestion({
+          task_id: taskId,
+          question,
+          history,
+          provider_id: providerId,
+          model_name: modelName,
+        })
+        addMessage(taskId, {
+          role: 'assistant',
+          content: res.answer,
+          sources: res.sources,
+        })
+      } catch {
+        toast.error('问答请求失败')
+      } finally {
+        setLoading(false)
+      }
+    },
+    [loading, taskId, currentTask, messages, addMessage],
+  )
+
+  // 转换为 Bubble.List 的数据格式
+  const bubbleItems = useMemo(() => {
+    const items = messages.map((msg, i) => ({
+      key: `msg-${i}`,
+      role: msg.role === 'user' ? ('user' as const) : ('ai' as const),
+      content: msg.content,
+      footer:
+        msg.role === 'assistant' && msg.sources ? (
+          <SourceBadges sources={msg.sources} />
+        ) : undefined,
+    }))
+
+    if (loading) {
+      items.push({
+        key: 'loading',
+        role: 'ai' as const,
+        content: '思考中...',
+        loading: true,
+      } as any)
+    }
+
+    return items
+  }, [messages, loading])
+
+  // Bubble 角色配置
+  const roles = useMemo(
+    () => ({
+      user: {
+        placement: 'end' as const,
+        avatar: (
+          <div className="flex h-7 w-7 items-center justify-center rounded-full bg-blue-500 text-white">
+            <UserRound className="h-4 w-4" />
+          </div>
+        ),
+        variant: 'filled' as const,
+        styles: { content: { background: '#3b82f6', color: '#fff' } },
+      },
+      ai: {
+        placement: 'start' as const,
+        avatar: (
+          <div className="flex h-7 w-7 items-center justify-center rounded-full bg-neutral-500 text-white">
+            <Bot className="h-4 w-4" />
+          </div>
+        ),
+        variant: 'outlined' as const,
+        contentRender: (content: any) => (
+          <div className="markdown-body prose prose-sm max-w-none prose-p:my-1 prose-li:my-0.5 prose-headings:my-2">
+            <ReactMarkdown remarkPlugins={[remarkGfm]}>
+              {typeof content === 'string' ? content : String(content)}
+            </ReactMarkdown>
+          </div>
+        ),
+      },
+    }),
+    [],
+  )
+
+  if (indexStatus === null || indexStatus === 'indexing' || indexStatus === 'idle') {
+    return (
+      <div className="flex h-full flex-col items-center justify-center gap-3 text-neutral-400">
+        <Loader2 className="h-6 w-6 animate-spin" />
+        <div className="text-center">
+          <p className="text-sm font-medium">正在索引笔记内容...</p>
+          <p className="mt-1 text-xs">首次使用需下载 Embedding 模型（约 80MB），请耐心等待</p>
+        </div>
+      </div>
+    )
+  }
+
+  if (indexStatus === 'failed') {
+    return (
+      <div className="flex h-full flex-col items-center justify-center gap-2 text-neutral-400">
+        <span className="text-sm">索引失败，请重试</span>
+        <Button
+          size="sm"
+          variant="outline"
+          onClick={async () => {
+            setIndexStatus('indexing')
+            try {
+              await indexTask(taskId)
+            } catch {
+              toast.error('索引请求失败')
+              setIndexStatus('failed')
+            }
+          }}
+        >
+          重新索引
+        </Button>
+      </div>
+    )
+  }
+
+  return (
+    <div className="flex h-full flex-col border-l">
+      {/* 头部 */}
+      <div className="flex items-center justify-between border-b px-3 py-2">
+        <span className="text-sm font-medium">AI 问答</span>
+        <div className="flex items-center gap-1">
+          <Button
+            variant="ghost"
+            size="sm"
+            className="h-7 px-2 text-neutral-400 hover:text-neutral-600"
+            onClick={() => onModeChange(mode === 'half' ? 'full' : 'half')}
+            title={mode === 'half' ? '全屏' : '半屏'}
+          >
+            {mode === 'half' ? (
+              <Maximize2 className="h-3.5 w-3.5" />
+            ) : (
+              <Minimize2 className="h-3.5 w-3.5" />
+            )}
+          </Button>
+          {messages.length > 0 && (
+            <Button
+              variant="ghost"
+              size="sm"
+              className="h-7 px-2 text-neutral-400 hover:text-red-500"
+              onClick={() => clearChat(taskId)}
+            >
+              <Trash2 className="h-3.5 w-3.5" />
+            </Button>
+          )}
+        </div>
+      </div>
+
+      {/* 消息列表 */}
+      <div className="flex-1 overflow-hidden">
+        {messages.length === 0 && !loading ? (
+          <div className="flex h-full items-center justify-center text-center text-sm text-neutral-400">
+            <div>
+              <p>针对笔记内容提问</p>
+              <p className="mt-1 text-xs">例如：这个视频的核心观点是什么？</p>
+            </div>
+          </div>
+        ) : (
+          <Bubble.List
+            items={bubbleItems}
+            role={roles}
+            style={{ height: '100%' }}
+          />
+        )}
+      </div>
+
+      {/* 输入区域 */}
+      <div className="border-t px-3 py-2">
+        <Sender
+          value={input}
+          onChange={setInput}
+          onSubmit={handleSend}
+          loading={loading}
+          placeholder="输入你的问题..."
+        />
+      </div>
+    </div>
+  )
+}
--- a/BillNote_frontend/src/pages/HomePage/components/MarkdownHeader.tsx
+++ b/BillNote_frontend/src/pages/HomePage/components/MarkdownHeader.tsx
@@ -1,7 +1,7 @@
 'use client'

 import { useEffect, useState } from 'react'
-import { Copy, Download, BrainCircuit } from 'lucide-react'
+import { Copy, Download, BrainCircuit, MessageSquare } from 'lucide-react'
 import { Button } from '@/components/ui/button'
 import { Select, SelectContent, SelectItem, SelectTrigger } from '@/components/ui/select'
 import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from '@/components/ui/tooltip'
@@ -28,6 +28,8 @@ interface NoteHeaderProps {
  onDownload: () => void
  createAt?: string | Date
  setShowTranscribe: (show: boolean) => void
+  showChat?: false | 'half' | 'full'
+  setShowChat?: (mode: false | 'half' | 'full') => void
 }

 export function MarkdownHeader({
@@ -43,6 +45,8 @@ export function MarkdownHeader({
  createAt,
  showTranscribe,
  setShowTranscribe,
+  showChat,
+  setShowChat,
  viewMode,
  setViewMode,
 }: NoteHeaderProps) {
@@ -183,6 +187,24 @@ export function MarkdownHeader({
            <TooltipContent>原文参照</TooltipContent>
          </Tooltip>
        </TooltipProvider>
+        {setShowChat && (
+          <TooltipProvider>
+            <Tooltip>
+              <TooltipTrigger asChild>
+                <Button
+                  onClick={() => setShowChat(showChat ? false : 'half')}
+                  variant={showChat ? 'default' : 'ghost'}
+                  size="sm"
+                  className="h-8 px-2"
+                >
+                  <MessageSquare className="mr-1.5 h-4 w-4" />
+                  <span className="text-sm">AI 问答</span>
+                </Button>
+              </TooltipTrigger>
+              <TooltipContent>基于笔记内容的 AI 问答</TooltipContent>
+            </Tooltip>
+          </TooltipProvider>
+        )}
      </div>
    </div>
  )
--- a/BillNote_frontend/src/pages/HomePage/components/MarkdownViewer.tsx
+++ b/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'
@@ -16,13 +16,14 @@ 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'
 import { MarkdownHeader } from '@/pages/HomePage/components/MarkdownHeader.tsx'
 import TranscriptViewer from '@/pages/HomePage/components/transcriptViewer.tsx'
 import MarkmapEditor from '@/pages/HomePage/components/MarkmapComponent.tsx'
+import ChatPanel from '@/pages/HomePage/components/ChatPanel.tsx'
+import VideoBanner from '@/pages/HomePage/components/VideoBanner.tsx'

 interface VersionNote {
  ver_id: string
@@ -45,22 +46,249 @@ 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>('')
  const [modelName, setModelName] = useState<string>('')
  const [style, setStyle] = useState<string>('')
  const [createTime, setCreateTime] = useState<string>('')
-  const baseURL = String(import.meta.env.VITE_API_BASE_URL).replace('/api','') || ''
+  // 确保baseURL没有尾部斜杠
+  const baseURL = (String(import.meta.env.VITE_API_BASE_URL || '').replace('/api','') || '').replace(/\/$/, '')
  const getCurrentTask = useTaskStore.getState().getCurrentTask
  const currentTask = useTaskStore(state => state.getCurrentTask())
  const taskStatus = currentTask?.status || 'PENDING'
  const retryTask = useTaskStore.getState().retryTask
  const isMultiVersion = Array.isArray(currentTask?.markdown)
  const [showTranscribe, setShowTranscribe] = useState(false)
+  const [showChat, setShowChat] = useState<false | 'half' | 'full'>(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
@@ -159,7 +387,7 @@ 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>
@@ -197,6 +425,8 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
        createAt={createTime}
        showTranscribe={showTranscribe}
        setShowTranscribe={setShowTranscribe}
+        showChat={showChat}
+        setShowChat={setShowChat}
        viewMode={viewMode}
        setViewMode={setViewMode}
      />
@@ -208,6 +438,7 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
              value={selectedContent}
              onChange={() => {}}
              height="100%" // 根据需求可以设定百分比或固定高度
+              title={currentTask?.audioMeta?.title || '思维导图'}
            />
          </div>
        </div>
@@ -215,253 +446,26 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
        <div className="flex flex-1 overflow-hidden bg-white py-2">
          {selectedContent && selectedContent !== 'loading' && selectedContent !== 'empty' ? (
            <>
-              <ScrollArea className="w-full">
+              {showChat === 'full' && currentTask ? (
+                <div className="h-full w-full">
+                  <ChatPanel taskId={currentTask.id} mode="full" onModeChange={setShowChat} />
+                </div>
+              ) : (
+              <>
+              <ScrollArea className="min-w-0 flex-1">
+                <div className="px-2">
+                  <VideoBanner
+                    audioMeta={currentTask?.audioMeta}
+                    videoUrl={currentTask?.formData?.video_url}
+                  />
+                </div>
                <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 }) =>{
-
-                        let src = baseURL +props.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}
+                    {selectedContent.replace(/^>\s*来源链接：[^\n]*\n*/m, '')}
                  </ReactMarkdown>
                </div>
              </ScrollArea>
@@ -470,6 +474,14 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
                  <TranscriptViewer />
                </div>
              )}
+              {/* 侧边问答模式：markdown + ChatPanel 各占一半 */}
+              {showChat === 'half' && currentTask && (
+                <div className="ml-2 h-full w-1/2 shrink-0">
+                  <ChatPanel taskId={currentTask.id} mode="half" onModeChange={setShowChat} />
+                </div>
+              )}
+              </>
+              )}
            </>
          ) : (
            <div className="flex h-full w-full items-center justify-center">
@@ -486,6 +498,8 @@ const MarkdownViewer: FC<MarkdownViewerProps> = ({ status }) => {
      )}
    </div>
  )
-}
+})
+
+MarkdownViewer.displayName = 'MarkdownViewer'

 export default MarkdownViewer
--- a/BillNote_frontend/src/pages/HomePage/components/MarkmapComponent.tsx
+++ b/BillNote_frontend/src/pages/HomePage/components/MarkmapComponent.tsx
@@ -1,8 +1,9 @@
-import React, { useEffect, useRef, useState } from 'react'
+import { useEffect, useRef, useState } from 'react'
 import { Markmap } from 'markmap-view'
 import { transformer } from '@/lib/markmap.ts'
-import { Toolbar, ToolbarButton } from 'markmap-toolbar'
+import { Toolbar } from 'markmap-toolbar'
 import 'markmap-toolbar/dist/style.css'
+import JSZip from 'jszip'

 export interface MarkmapEditorProps {
  /** 要渲染的 Markdown 文本 */
@@ -12,9 +13,11 @@ export interface MarkmapEditorProps {
  /** Toolbar 上要展示的 item id 列表，默认使用 Toolbar.defaultItems */
  toolbarItems?: string[]
  /** 自定义按钮列表，会依次注册 */
-  customButtons?: ToolbarButton[]
+  customButtons?: any[]
  /** 容器 SVG 的高度，默认为 600px */
  height?: string
+  /** 文档标题，用于导出HTML时的文件名 */
+  title?: string
 }

 export default function MarkmapEditor({
@@ -23,9 +26,10 @@ export default function MarkmapEditor({
  toolbarItems,
  customButtons = [],
  height = '600px',
+  title = 'mindmap',
 }: MarkmapEditorProps) {
  const svgRef = useRef<SVGSVGElement>(null)
-  const mmRef = useRef<Markmap>()
+  const mmRef = useRef<Markmap | undefined>()
  const toolbarRef = useRef<HTMLDivElement>(null)

  // 用于跟踪是否处于全屏状态
@@ -56,6 +60,353 @@ export default function MarkmapEditor({
      document.exitFullscreen()
    }
  }
+  
+  // 导出HTML思维导图
+  const exportHtml = () => {
+    try {
+      const { root } = transformer.transform(value)
+      const data = JSON.stringify(root)
+      
+      // 创建HTML内容
+      const html = `<!DOCTYPE html>
+<html>
+<head>
+  <meta charset="UTF-8">
+  <meta name="viewport" content="width=device-width, initial-scale=1.0">
+  <title>${title || 'BiliNote思维导图'}</title>
+  <style>
+  body {
+    margin: 0;
+    padding: 0;
+    font-family: sans-serif;
+  }
+  #mindmap {
+    display: block;
+    width: 100%;
+    height: 100vh;
+  }
+  </style>
+  <script src="https://cdn.jsdelivr.net/npm/d3@7"></script>
+  <script src="https://cdn.jsdelivr.net/npm/markmap-view@0.18.10"></script>
+</head>
+<body>
+  <svg id="mindmap"></svg>
+  <script>
+  (async () => {
+    const { markmap } = window;
+    const { Markmap } = markmap;
+    const mm = Markmap.create(document.getElementById('mindmap'));
+    mm.setData(${data});
+    mm.fit();
+  })();
+  </script>
+</body>
+</html>`;
+      
+      const blob = new Blob([html], { type: 'text/html;charset=utf-8' });
+      const url = URL.createObjectURL(blob);
+      const a = document.createElement('a');
+      a.href = url;
+      a.download = `${title || 'mindmap'}.html`;
+      document.body.appendChild(a);
+      a.click();
+      document.body.removeChild(a);
+      URL.revokeObjectURL(url);
+    } catch (error) {
+      console.error('导出HTML失败:', error);
+    }
+  };
+
+  // 导出SVG思维导图（矢量图）
+  const exportSvg = async () => {
+    try {
+      if (!svgRef.current || !mmRef.current) return;
+
+      const svgEl = svgRef.current;
+      const mm = mmRef.current;
+
+      // 先调用fit()确保显示完整的思维导图内容
+      await mm.fit();
+      // 等待渲染完成
+      await new Promise(resolve => setTimeout(resolve, 100));
+
+      // 克隆SVG以避免修改原始SVG
+      const clonedSvg = svgEl.cloneNode(true) as SVGSVGElement;
+
+      // 获取SVG内容的实际边界框
+      const gElement = svgEl.querySelector('g');
+      if (gElement) {
+        const bbox = gElement.getBBox();
+        // 添加一些边距
+        const padding = 50;
+        const viewBoxX = bbox.x - padding;
+        const viewBoxY = bbox.y - padding;
+        const viewBoxWidth = bbox.width + padding * 2;
+        const viewBoxHeight = bbox.height + padding * 2;
+
+        // 设置viewBox以确保SVG可以无限缩放
+        clonedSvg.setAttribute('viewBox', `${viewBoxX} ${viewBoxY} ${viewBoxWidth} ${viewBoxHeight}`);
+        // 移除固定尺寸，让SVG根据viewBox自适应
+        clonedSvg.removeAttribute('width');
+        clonedSvg.removeAttribute('height');
+        // 设置默认尺寸为100%，可以在任何容器中自适应
+        clonedSvg.setAttribute('width', '100%');
+        clonedSvg.setAttribute('height', '100%');
+        // 保持宽高比
+        clonedSvg.setAttribute('preserveAspectRatio', 'xMidYMid meet');
+      }
+
+      // 设置SVG的背景为白色
+      const style = document.createElementNS('http://www.w3.org/2000/svg', 'style');
+      style.textContent = 'svg { background-color: white; }';
+      clonedSvg.insertBefore(style, clonedSvg.firstChild);
+
+      // 添加白色背景矩形（确保背景在所有查看器中都是白色）
+      const bgRect = document.createElementNS('http://www.w3.org/2000/svg', 'rect');
+      const viewBox = clonedSvg.getAttribute('viewBox')?.split(' ').map(Number) || [0, 0, 800, 600];
+      bgRect.setAttribute('x', viewBox[0].toString());
+      bgRect.setAttribute('y', viewBox[1].toString());
+      bgRect.setAttribute('width', viewBox[2].toString());
+      bgRect.setAttribute('height', viewBox[3].toString());
+      bgRect.setAttribute('fill', 'white');
+      // 插入到最前面作为背景
+      const firstG = clonedSvg.querySelector('g');
+      if (firstG) {
+        clonedSvg.insertBefore(bgRect, firstG);
+      } else {
+        clonedSvg.insertBefore(bgRect, clonedSvg.firstChild);
+      }
+
+      // 确保SVG有正确的命名空间
+      clonedSvg.setAttribute('xmlns', 'http://www.w3.org/2000/svg');
+      clonedSvg.setAttribute('xmlns:xlink', 'http://www.w3.org/1999/xlink');
+
+      // 序列化SVG
+      const svgData = new XMLSerializer().serializeToString(clonedSvg);
+
+      // 创建下载
+      const blob = new Blob([svgData], { type: 'image/svg+xml;charset=utf-8' });
+      const url = URL.createObjectURL(blob);
+      const a = document.createElement('a');
+      a.href = url;
+      a.download = `${title || 'mindmap'}.svg`;
+      document.body.appendChild(a);
+      a.click();
+      document.body.removeChild(a);
+      URL.revokeObjectURL(url);
+    } catch (error) {
+      console.error('导出SVG失败:', error);
+    }
+  };
+
+  // 导出XMind格式思维导图
+  const exportXMind = async () => {
+    try {
+      const { root } = transformer.transform(value);
+
+      // 生成唯一ID
+      const generateId = () => Math.random().toString(36).substring(2, 15);
+
+      // 解码HTML实体（如 &#x5b9e; -> 实，&#12345; -> 对应字符）
+      const decodeHtmlEntities = (text: string): string => {
+        if (!text) return text;
+
+        // 首先手动处理十六进制数字实体 &#xHHHH;
+        let decoded = text.replace(/&#x([0-9a-fA-F]+);/g, (_, hex) => {
+          return String.fromCodePoint(parseInt(hex, 16));
+        });
+
+        // 处理十进制数字实体 &#DDDD;
+        decoded = decoded.replace(/&#(\d+);/g, (_, dec) => {
+          return String.fromCodePoint(parseInt(dec, 10));
+        });
+
+        // 使用textarea处理命名实体（如 &amp; &lt; &gt; 等）
+        const textarea = document.createElement('textarea');
+        textarea.innerHTML = decoded;
+        return textarea.value;
+      };
+
+      // 清理HTML标签，只保留纯文本
+      const stripHtml = (html: string): string => {
+        if (!html) return html;
+        // 先解码HTML实体
+        let text = decodeHtmlEntities(html);
+        // 移除HTML标签
+        const div = document.createElement('div');
+        div.innerHTML = text;
+        return div.textContent || div.innerText || text;
+      };
+
+      // 将 markmap 节点转换为 XMind 节点格式
+      const convertToXMindNode = (node: any, isRoot = false): any => {
+        const rawTitle = node.content || node.payload?.content || '未命名';
+        const xmindNode: any = {
+          id: generateId(),
+          class: isRoot ? 'topic' : 'topic',
+          title: stripHtml(rawTitle),
+        };
+
+        if (node.children && node.children.length > 0) {
+          xmindNode.children = {
+            attached: node.children.map((child: any) => convertToXMindNode(child, false))
+          };
+        }
+
+        return xmindNode;
+      };
+
+      const rootTopic = convertToXMindNode(root, true);
+      const sheetId = generateId();
+
+      // XMind content.json 结构
+      const content = [{
+        id: sheetId,
+        class: 'sheet',
+        title: stripHtml(title) || '思维导图',
+        rootTopic: rootTopic,
+        topicPositioning: 'fixed'
+      }];
+
+      // XMind metadata.json
+      const metadata = {
+        creator: {
+          name: 'BiliNote',
+          version: '1.0.0'
+        }
+      };
+
+      // XMind manifest.json
+      const manifest = {
+        'file-entries': {
+          'content.json': {},
+          'metadata.json': {}
+        }
+      };
+
+      // 使用 JSZip 创建 .xmind 文件
+      // 直接传入字符串，JSZip会自动处理UTF-8编码
+      const zip = new JSZip();
+      zip.file('content.json', JSON.stringify(content, null, 2));
+      zip.file('metadata.json', JSON.stringify(metadata, null, 2));
+      zip.file('manifest.json', JSON.stringify(manifest, null, 2));
+
+      // 生成 ZIP 并下载
+      const blob = await zip.generateAsync({ type: 'blob' });
+      const url = URL.createObjectURL(blob);
+      const a = document.createElement('a');
+      a.href = url;
+      a.download = `${title || 'mindmap'}.xmind`;
+      document.body.appendChild(a);
+      a.click();
+      document.body.removeChild(a);
+      URL.revokeObjectURL(url);
+    } catch (error) {
+      console.error('导出XMind失败:', error);
+    }
+  };
+
+  // 导出PNG思维导图
+  const exportPng = async () => {
+    try {
+      if (!svgRef.current || !mmRef.current) return;
+
+      const svgEl = svgRef.current;
+      const mm = mmRef.current;
+
+      // 先调用fit()确保显示完整的思维导图内容
+      await mm.fit();
+      // 等待渲染完成
+      await new Promise(resolve => setTimeout(resolve, 100));
+      
+      // 获取SVG实际尺寸
+      const svgWidth = svgEl.width.baseVal.value || svgEl.clientWidth || 800;
+      const svgHeight = svgEl.height.baseVal.value || svgEl.clientHeight || 600;
+      
+      // 设置足够大的缩放比例以确保高清输出
+      const scale = 3;
+      
+      // 克隆SVG以避免修改原始SVG
+      const clonedSvg = svgEl.cloneNode(true) as SVGSVGElement;
+      
+      // 设置SVG的背景为白色
+      const style = document.createElementNS('http://www.w3.org/2000/svg', 'style');
+      style.textContent = 'svg { background-color: white; }';
+      clonedSvg.insertBefore(style, clonedSvg.firstChild);
+      
+      // 确保SVG有正确的命名空间
+      clonedSvg.setAttribute('xmlns', 'http://www.w3.org/2000/svg');
+      clonedSvg.setAttribute('width', svgWidth.toString());
+      clonedSvg.setAttribute('height', svgHeight.toString());
+      
+      // 将SVG转换为Data URI (避免使用Blob URL来解决跨域问题)
+      const svgData = new XMLSerializer().serializeToString(clonedSvg);
+      const svgBase64 = btoa(unescape(encodeURIComponent(svgData)));
+      const dataUri = `data:image/svg+xml;base64,${svgBase64}`;
+      
+      // 创建Canvas
+      const canvas = document.createElement('canvas');
+      canvas.width = svgWidth * scale;
+      canvas.height = svgHeight * scale;
+      
+      // 获取上下文并设置白色背景
+      const ctx = canvas.getContext('2d');
+      if (!ctx) {
+        throw new Error('无法获取Canvas上下文');
+      }
+      
+      // 设置白色背景
+      ctx.fillStyle = '#FFFFFF';
+      ctx.fillRect(0, 0, canvas.width, canvas.height);
+      
+      // 创建Image对象
+      const img = new Image();
+      
+      // 当图片加载完成后，在Canvas上绘制并导出
+      img.onload = () => {
+        try {
+          // 应用缩放
+          ctx.setTransform(scale, 0, 0, scale, 0, 0);
+          
+          // 绘制SVG
+          ctx.drawImage(img, 0, 0);
+          
+          // 重置变换
+          ctx.setTransform(1, 0, 0, 1, 0, 0);
+          
+          // 将Canvas转换为PNG Blob
+          canvas.toBlob((blob) => {
+            if (blob) {
+              // 创建下载链接
+              const url = URL.createObjectURL(blob);
+              const a = document.createElement('a');
+              a.href = url;
+              a.download = `${title || 'mindmap'}.png`;
+              document.body.appendChild(a);
+              a.click();
+              document.body.removeChild(a);
+              URL.revokeObjectURL(url);
+            } else {
+              console.error('无法创建Blob对象');
+            }
+          }, 'image/png');
+        } catch (err) {
+          console.error('Canvas处理失败:', err);
+        }
+      };
+      
+      // 设置图片加载错误处理
+      img.onerror = (error) => {
+        console.error('导出PNG失败（图片加载错误）:', error);
+      };
+      
+      // 开始加载SVG图像 (使用Data URI而不是Blob URL)
+      img.src = dataUri;
+      
+    } catch (error) {
+      console.error('导出PNG失败:', error);
+    }
+  };

  // 初始化 Markmap 实例 + Toolbar
  useEffect(() => {
@@ -82,14 +433,42 @@ export default function MarkmapEditor({
  }, [value])

  // 文本输入变化回调（如果你自行添加 textarea 编辑区）
-  const handleChange = (e: React.ChangeEvent<HTMLTextAreaElement>) => {
-    onChange(e.target.value)
-  }
+  // const handleChange = (e: React.ChangeEvent<HTMLTextAreaElement>) => {
+  //   onChange(e.target.value)
+  // }

  return (
    <div className="relative flex h-full flex-col bg-white">
      {/* 全屏/退出全屏 按钮 */}
      <div className="absolute top-2 right-2 z-20 flex space-x-2">
+        <button
+          onClick={exportXMind}
+          className="rounded p-1 hover:bg-gray-200"
+          title="导出XMind格式"
+        >
+          🧠
+        </button>
+        <button
+          onClick={exportSvg}
+          className="rounded p-1 hover:bg-gray-200"
+          title="导出SVG矢量图（可无限放大）"
+        >
+          📐
+        </button>
+        <button
+          onClick={exportPng}
+          className="rounded p-1 hover:bg-gray-200"
+          title="导出PNG图片"
+        >
+          🖼️
+        </button>
+        <button
+          onClick={exportHtml}
+          className="rounded p-1 hover:bg-gray-200"
+          title="导出HTML（可交互）"
+        >
+          💾
+        </button>
        {isFullscreen ? (
          <button
            onClick={exitFullscreen}
--- a/BillNote_frontend/src/pages/HomePage/components/NoteForm.tsx
+++ b/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'
@@ -43,7 +43,7 @@ import { useNavigate } from 'react-router-dom'
 /* -------------------- 校验 Schema -------------------- */
 const formSchema = z
  .object({
-    video_url: z.string(),
+    video_url: z.string().optional(),
    platform: z.string().nonempty('请选择平台'),
    quality: z.enum(['fast', 'medium', 'slow']),
    screenshot: z.boolean().optional(),
@@ -53,28 +53,36 @@ const formSchema = z
    style: z.string().nonempty('请选择笔记生成风格'),
    extras: z.string().optional(),
    video_understanding: z.boolean().optional(),
-    video_interval: z.coerce.number().min(1).max(30).default(4).optional(),
+    video_interval: z.coerce.number().min(1).max(30).default(6).optional(),
    grid_size: z
      .tuple([z.coerce.number().min(1).max(10), z.coerce.number().min(1).max(10)])
-      .default([3, 3])
+      .default([2, 2])
      .optional(),
  })
  .superRefine(({ video_url, platform }, ctx) => {
-    if (platform === 'local' || platform === 'douyin') {
+    if (platform === 'local') {
      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 (!['http:', 'https:'].includes(url.protocol)) throw new Error()
-      } catch {
-        ctx.addIssue({ code: 'custom', message: '请输入正确的视频链接', path: ['video_url'] })
+    }
+    else {
+      if (!video_url) {
+        ctx.addIssue({ code: 'custom', message: '视频链接不能为空', path: ['video_url'] })
+      }
+      else {
+        try {
+          const url = new URL(video_url)
+          if (!['http:', 'https:'].includes(url.protocol))
+            throw new Error()
+        }
+        catch {
+          ctx.addIssue({ code: 'custom', message: '请输入正确的视频链接', path: ['video_url'] })
+        }
      }
    }
  })

-type NoteFormValues = z.infer<typeof formSchema>
+export type NoteFormValues = z.infer<typeof formSchema>

 /* -------------------- 可复用子组件 -------------------- */
 const SectionHeader = ({ title, tip }: { title: string; tip?: string }) => (
@@ -136,8 +144,8 @@ const NoteForm = () => {
      quality: 'medium',
      model_name: modelList[0]?.model_name || '',
      style: 'minimal',
-      video_interval: 4,
-      grid_size: [3, 3],
+      video_interval: 6,
+      grid_size: [2, 2],
      format: [],
    },
  })
@@ -173,8 +181,8 @@ const NoteForm = () => {
      screenshot: formData.screenshot ?? false,
      link: formData.link ?? false,
      video_understanding: formData.video_understanding ?? false,
-      video_interval: formData.video_interval ?? 4,
-      grid_size: formData.grid_size ?? [3, 3],
+      video_interval: formData.video_interval ?? 6,
+      grid_size: formData.grid_size ?? [2, 2],
      format: formData.format ?? [],
    })
  }, [
@@ -202,7 +210,7 @@ const NoteForm = () => {
        setUploadSuccess(true)
    } catch (err) {
      console.error('上传失败:', err)
-      message.error('上传失败，请重试')
+      // message.error('上传失败，请重试')
    } finally {
      setIsUploading(false)
    }
@@ -220,13 +228,13 @@ const NoteForm = () => {
      return
    }

-    message.success('已提交任务')
+    // message.success('已提交任务')
    const  data  = await generateNote(payload)
    addPendingTask(data.task_id, values.platform, payload)
  }
  const onInvalid = (errors: FieldErrors<NoteFormValues>) => {
    console.warn('表单校验失败：', errors)
-    message.error('请完善所有必填项后再提交')
+    // message.error('请完善所有必填项后再提交')
  }
  const handleCreateNew = () => {
    // 🔁 这里清空当前任务状态
@@ -297,7 +305,7 @@ const NoteForm = () => {
                      ))}
                    </SelectContent>
                  </Select>
-                  <FormMessage />
+                  <FormMessage style={{ display: 'none' }} />
                </FormItem>
              )}
            />
@@ -314,7 +322,7 @@ const NoteForm = () => {
                  ) : (
                    <Input disabled={!!editing} placeholder="请输入视频网站链接" {...field} />
                  )}
-                  <FormMessage />
+                  <FormMessage style={{ display: 'none' }} />
                </FormItem>
              )}
            />
@@ -505,17 +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 */}
--- a/BillNote_frontend/src/pages/HomePage/components/NoteHistory.tsx
+++ b/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
@@ -24,13 +24,14 @@ interface NoteHistoryProps {
 const NoteHistory: FC<NoteHistoryProps> = ({ onSelect, selectedId }) => {
  const tasks = useTaskStore(state => state.tasks)
  const removeTask = useTaskStore(state => state.removeTask)
-  const baseURL = import.meta.env.VITE_API_BASE_URL || 'api/'
+  // 确保baseURL没有尾部斜杠
+  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
@@ -77,16 +78,15 @@ const NoteHistory: FC<NoteHistoryProps> = ({ onSelect, selectedId }) => {
      <div className="flex flex-col gap-2 overflow-hidden">
        {filteredTasks.map(task => (
          <div
-              onClick={() => onSelect(task.id)}
+            key={task.id}
+            onClick={() => onSelect(task.id)}
            className={cn(
              'flex cursor-pointer flex-col rounded-md border border-neutral-200 p-3',
              selectedId === task.id && 'border-primary bg-primary-light'
            )}
          >
            <div
-              key={task.id}
              className={cn('flex items-center gap-4')}
-
            >
              {/* 封面图 */}
              {task.platform === 'local' ? (
@@ -102,7 +102,7 @@ const NoteHistory: FC<NoteHistoryProps> = ({ onSelect, selectedId }) => {

                      src={
                        task.audioMeta.cover_url
-                            ? baseURL+`/image_proxy?url=${encodeURIComponent(task.audioMeta.cover_url)}`
+                            ? `${baseURL}/image_proxy?url=${encodeURIComponent(task.audioMeta.cover_url)}`
                            : '/placeholder.png'
                      }
                      alt="封面"
--- a/BillNote_frontend/src/pages/HomePage/components/VideoBanner.tsx
+++ b/BillNote_frontend/src/pages/HomePage/components/VideoBanner.tsx
@@ -0,0 +1,86 @@
+import { ExternalLink } from 'lucide-react'
+import type { AudioMeta } from '@/store/taskStore'
+
+interface VideoBannerProps {
+  audioMeta?: AudioMeta
+  videoUrl?: string
+}
+
+/** 平台 label 映射 */
+const platformLabel: Record<string, string> = {
+  bilibili: '哔哩哔哩',
+  youtube: 'YouTube',
+  douyin: '抖音',
+  xiaohongshu: '小红书',
+}
+
+export default function VideoBanner({ audioMeta, videoUrl }: VideoBannerProps) {
+  if (!audioMeta) return null
+
+  const rawCover = audioMeta.cover_url
+  // 通过后端代理加载封面，避免跨域/Referrer 限制
+  const apiBase = String(import.meta.env.VITE_API_BASE_URL || 'api').replace(/\/$/, '')
+  const coverUrl = rawCover
+    ? `${apiBase}/image_proxy?url=${encodeURIComponent(rawCover)}`
+    : ''
+  const title = audioMeta.title
+  const uploader = audioMeta.raw_info?.uploader || ''
+  const platform = platformLabel[audioMeta.platform] || audioMeta.platform || ''
+  const originalUrl = videoUrl || audioMeta.raw_info?.webpage_url || ''
+
+  return (
+    <div className="relative mb-4 overflow-hidden rounded-lg">
+      {/* 模糊背景封面 */}
+      <div className="absolute inset-0">
+        {coverUrl ? (
+          <img
+            src={coverUrl}
+            alt=""
+            referrerPolicy="no-referrer"
+            className="h-full w-full object-cover blur-md brightness-[0.4] scale-110"
+          />
+        ) : (
+          <div className="h-full w-full bg-gradient-to-r from-blue-600 to-indigo-700" />
+        )}
+      </div>
+
+      {/* 内容层 */}
+      <div className="relative flex items-center gap-4 px-5 py-4">
+        {/* 封面缩略图 */}
+        {coverUrl && (
+          <img
+            src={coverUrl}
+            alt={title}
+            referrerPolicy="no-referrer"
+            className="h-16 w-28 shrink-0 rounded-md object-cover shadow-md"
+          />
+        )}
+
+        {/* 文字信息 */}
+        <div className="min-w-0 flex-1">
+          <h2 className="truncate text-base font-bold text-white" title={title}>
+            {title}
+          </h2>
+          <div className="mt-1 flex flex-wrap items-center gap-2 text-sm text-white/70">
+            {uploader && <span>{uploader}</span>}
+            {uploader && platform && <span className="text-white/40">·</span>}
+            {platform && <span>{platform}</span>}
+          </div>
+        </div>
+
+        {/* 跳转原视频 */}
+        {originalUrl && (
+          <a
+            href={originalUrl}
+            target="_blank"
+            rel="noopener noreferrer"
+            className="flex shrink-0 items-center gap-1.5 rounded-full bg-white/15 px-3 py-1.5 text-xs font-medium text-white backdrop-blur-sm transition-colors hover:bg-white/25"
+          >
+            <ExternalLink className="h-3.5 w-3.5" />
+            <span>原视频</span>
+          </a>
+        )}
+      </div>
+    </div>
+  )
+}
--- a/BillNote_frontend/src/pages/SettingPage/Menu.tsx
+++ b/BillNote_frontend/src/pages/SettingPage/Menu.tsx
@@ -1,10 +1,9 @@
 import {
  BotMessageSquare,
-  SquareChevronRight,
  Captions,
  HardDriveDownload,
-  Wrench,
  Info,
+  Activity,
 } from 'lucide-react'
 import MenuBar, { IMenuProps } from '@/pages/SettingPage/components/menuBar.tsx'

@@ -16,14 +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',
@@ -37,6 +34,12 @@ const Menu = () => {
    //   icon: <SquareChevronRight />,
    //   path: '/settings/prompt',
    // },
+    {
+      id: 'monitor',
+      name: '部署监控',
+      icon: <Activity />,
+      path: '/settings/monitor',
+    },
    {
      id: 'about',
      name: '关于',
--- a/BillNote_frontend/src/pages/SettingPage/Monitor.tsx
+++ b/BillNote_frontend/src/pages/SettingPage/Monitor.tsx
@@ -0,0 +1,241 @@
+import { Card, CardContent, CardHeader, CardTitle } from '@/components/ui/card'
+import { Button } from '@/components/ui/button'
+import { Badge } from '@/components/ui/badge'
+import { ScrollArea } from '@/components/ui/scroll-area'
+import {
+    Server,
+    Cpu,
+    AudioLines,
+    Film,
+    RefreshCw,
+    CheckCircle2,
+    XCircle,
+    Loader2
+} from 'lucide-react'
+import { useState, useEffect, useCallback } from 'react'
+import { getDeployStatus, DeployStatus } from '@/services/system'
+
+export default function Monitor() {
+    const [status, setStatus] = useState<DeployStatus | null>(null)
+    const [loading, setLoading] = useState(true)
+    const [error, setError] = useState<string | null>(null)
+    const [lastUpdated, setLastUpdated] = useState<Date | null>(null)
+
+    const fetchStatus = useCallback(async () => {
+        try {
+            setLoading(true)
+            setError(null)
+            const data = await getDeployStatus()
+            setStatus(data)
+            setLastUpdated(new Date())
+        } catch (err) {
+            setError('无法连接到后端服务')
+            setStatus(null)
+        } finally {
+            setLoading(false)
+        }
+    }, [])
+
+    useEffect(() => {
+        fetchStatus()
+        // 自动刷新（每 30 秒）
+        const interval = setInterval(fetchStatus, 30000)
+        return () => clearInterval(interval)
+    }, [fetchStatus])
+
+    const StatusBadge = ({ ok, label }: { ok: boolean; label?: string }) => (
+        <Badge
+            variant={ok ? 'default' : 'destructive'}
+            className={ok ? 'bg-green-500 hover:bg-green-600' : ''}
+        >
+            {ok ? (
+                <><CheckCircle2 className="mr-1 h-3 w-3" />{label || '正常'}</>
+            ) : (
+                <><XCircle className="mr-1 h-3 w-3" />{label || '异常'}</>
+            )}
+        </Badge>
+    )
+
+    return (
+        <ScrollArea className="h-full overflow-y-auto bg-white">
+            <div className="container mx-auto px-4 py-8">
+                {/* Header */}
+                <div className="mb-8 flex items-center justify-between">
+                    <div>
+                        <h1 className="text-2xl font-bold">部署监控</h1>
+                        <p className="text-muted-foreground text-sm">
+                            实时监控系统各组件运行状态
+                        </p>
+                    </div>
+                    <div className="flex items-center gap-4">
+                        {lastUpdated && (
+                            <span className="text-muted-foreground text-xs">
+                                最后更新: {lastUpdated.toLocaleTimeString()}
+                            </span>
+                        )}
+                        <Button
+                            variant="outline"
+                            size="sm"
+                            onClick={fetchStatus}
+                            disabled={loading}
+                        >
+                            {loading ? (
+                                <Loader2 className="mr-2 h-4 w-4 animate-spin" />
+                            ) : (
+                                <RefreshCw className="mr-2 h-4 w-4" />
+                            )}
+                            刷新
+                        </Button>
+                    </div>
+                </div>
+
+                {error && (
+                    <div className="mb-6 rounded-lg border border-red-200 bg-red-50 p-4 text-red-700">
+                        {error}
+                    </div>
+                )}
+
+                {/* Status Cards */}
+                <div className="grid grid-cols-1 gap-6 md:grid-cols-2">
+                    {/* Backend FastAPI */}
+                    <Card>
+                        <CardHeader className="flex flex-row items-center justify-between space-y-0 pb-2">
+                            <CardTitle className="text-lg font-medium">
+                                <Server className="mr-2 inline h-5 w-5 text-blue-500" />
+                                后端 FastAPI
+                            </CardTitle>
+                            {status && <StatusBadge ok={status.backend.status === 'running'} label="运行中" />}
+                        </CardHeader>
+                        <CardContent>
+                            {loading && !status ? (
+                                <div className="flex items-center gap-2 text-gray-500">
+                                    <Loader2 className="h-4 w-4 animate-spin" />
+                                    加载中...
+                                </div>
+                            ) : status ? (
+                                <div className="space-y-2 text-sm">
+                                    <div className="flex justify-between">
+                                        <span className="text-muted-foreground">状态:</span>
+                                        <span className={status.backend.status === 'running' ? 'font-medium text-green-600' : 'font-medium text-red-600'}>
+                                            {status.backend.status === 'running' ? '运行中' : status.backend.status}
+                                        </span>
+                                    </div>
+                                    <div className="flex justify-between">
+                                        <span className="text-muted-foreground">端口:</span>
+                                        <span className="font-mono">{status.backend.port}</span>
+                                    </div>
+                                </div>
+                            ) : null}
+                        </CardContent>
+                    </Card>
+
+                    {/* CUDA GPU */}
+                    <Card>
+                        <CardHeader className="flex flex-row items-center justify-between space-y-0 pb-2">
+                            <CardTitle className="text-lg font-medium">
+                                <Cpu className="mr-2 inline h-5 w-5 text-green-500" />
+                                CUDA GPU
+                            </CardTitle>
+                            {status && <StatusBadge ok={status.cuda.available} label={status.cuda.available ? '已启用' : '未启用'} />}
+                        </CardHeader>
+                        <CardContent>
+                            {loading && !status ? (
+                                <div className="flex items-center gap-2 text-gray-500">
+                                    <Loader2 className="h-4 w-4 animate-spin" />
+                                    加载中...
+                                </div>
+                            ) : status ? (
+                                <div className="space-y-2 text-sm">
+                                    {status.cuda.available ? (
+                                        <>
+                                            <div className="flex justify-between">
+                                                <span className="text-muted-foreground">GPU:</span>
+                                                <span className="font-medium">{status.cuda.gpu_name}</span>
+                                            </div>
+                                            <div className="flex justify-between">
+                                                <span className="text-muted-foreground">CUDA 版本:</span>
+                                                <span className="font-mono">{status.cuda.version}</span>
+                                            </div>
+                                        </>
+                                    ) : (
+                                        <div className="text-muted-foreground">
+                                            CUDA 不可用，将使用 CPU 模式
+                                        </div>
+                                    )}
+                                </div>
+                            ) : null}
+                        </CardContent>
+                    </Card>
+
+                    {/* Whisper Model */}
+                    <Card>
+                        <CardHeader className="flex flex-row items-center justify-between space-y-0 pb-2">
+                            <CardTitle className="text-lg font-medium">
+                                <AudioLines className="mr-2 inline h-5 w-5 text-purple-500" />
+                                Whisper 模型
+                            </CardTitle>
+                            {status && <StatusBadge ok={true} label="已配置" />}
+                        </CardHeader>
+                        <CardContent>
+                            {loading && !status ? (
+                                <div className="flex items-center gap-2 text-gray-500">
+                                    <Loader2 className="h-4 w-4 animate-spin" />
+                                    加载中...
+                                </div>
+                            ) : status ? (
+                                <div className="space-y-2 text-sm">
+                                    <div className="flex justify-between">
+                                        <span className="text-muted-foreground">模型大小:</span>
+                                        <span className="font-medium">{status.whisper.model_size}</span>
+                                    </div>
+                                    <div className="flex justify-between">
+                                        <span className="text-muted-foreground">转写引擎:</span>
+                                        <span className="font-mono">{status.whisper.transcriber_type}</span>
+                                    </div>
+                                </div>
+                            ) : null}
+                        </CardContent>
+                    </Card>
+
+                    {/* FFmpeg */}
+                    <Card>
+                        <CardHeader className="flex flex-row items-center justify-between space-y-0 pb-2">
+                            <CardTitle className="text-lg font-medium">
+                                <Film className="mr-2 inline h-5 w-5 text-orange-500" />
+                                FFmpeg
+                            </CardTitle>
+                            {status && <StatusBadge ok={status.ffmpeg.available} label={status.ffmpeg.available ? '可用' : '不可用'} />}
+                        </CardHeader>
+                        <CardContent>
+                            {loading && !status ? (
+                                <div className="flex items-center gap-2 text-gray-500">
+                                    <Loader2 className="h-4 w-4 animate-spin" />
+                                    加载中...
+                                </div>
+                            ) : status ? (
+                                <div className="space-y-2 text-sm">
+                                    <div className="flex justify-between">
+                                        <span className="text-muted-foreground">状态:</span>
+                                        <span className={status.ffmpeg.available ? 'font-medium text-green-600' : 'font-medium text-red-600'}>
+                                            {status.ffmpeg.available ? '已安装' : '未安装'}
+                                        </span>
+                                    </div>
+                                    {!status.ffmpeg.available && (
+                                        <div className="text-xs text-red-500">
+                                            请安装 FFmpeg 并添加到系统 PATH
+                                        </div>
+                                    )}
+                                </div>
+                            ) : null}
+                        </CardContent>
+                    </Card>
+                </div>
+
+                {/* Footer Info */}
+                <div className="mt-8 text-center text-xs text-gray-400">
+                    状态每 30 秒自动刷新
+                </div>
+            </div>
+        </ScrollArea>
+    )
+}
--- a/BillNote_frontend/src/pages/SettingPage/about.tsx
+++ b/BillNote_frontend/src/pages/SettingPage/about.tsx
@@ -26,7 +26,7 @@ export default function AboutPage() {
              height={50}
              className="rounded-lg"
            />
-            <h1 className="text-4xl font-bold">BiliNote v1.8.1</h1>
+            <h1 className="text-4xl font-bold">BiliNote v2.0.0</h1>
          </div>
          <p className="text-muted-foreground mb-6 text-xl italic">
            AI 视频笔记生成工具 让 AI 为你的视频做笔记
--- a/BillNote_frontend/src/pages/SettingPage/transcriber.tsx
+++ b/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-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="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="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
--- a/BillNote_frontend/src/services/chat.ts
+++ b/BillNote_frontend/src/services/chat.ts
@@ -0,0 +1,44 @@
+import request from '@/utils/request'
+
+export interface ChatMessage {
+  role: 'user' | 'assistant'
+  content: string
+}
+
+export interface ChatSource {
+  text: string
+  source_type: 'markdown' | 'transcript'
+  section_title?: string
+  start_time?: number
+  end_time?: number
+}
+
+export interface AskResponse {
+  answer: string
+  sources: ChatSource[]
+}
+
+export type IndexStatus = 'idle' | 'indexing' | 'indexed' | 'failed'
+
+export interface ChatStatusResponse {
+  indexed: boolean
+  status: IndexStatus
+}
+
+export const indexTask = async (taskId: string): Promise<void> => {
+  return await request.post('/chat/index', { task_id: taskId })
+}
+
+export const askQuestion = async (data: {
+  task_id: string
+  question: string
+  history: ChatMessage[]
+  provider_id: string
+  model_name: string
+}): Promise<AskResponse> => {
+  return await request.post('/chat/ask', data, { timeout: 60000 })
+}
+
+export const getChatStatus = async (taskId: string): Promise<ChatStatusResponse> => {
+  return await request.get(`/chat/status?task_id=${taskId}`)
+}
--- a/BillNote_frontend/src/services/system.ts
+++ b/BillNote_frontend/src/services/system.ts
@@ -1,5 +1,29 @@
 import request from '@/utils/request'

-export const systemCheck=async()=>{
+export const systemCheck = async () => {
  return await request.get('/sys_health')
 }
+
+export interface DeployStatus {
+  backend: {
+    status: string
+    port: number
+  }
+  cuda: {
+    available: boolean
+    version: string | null
+    gpu_name: string | null
+  }
+  whisper: {
+    model_size: string
+    transcriber_type: string
+  }
+  ffmpeg: {
+    available: boolean
+  }
+}
+
+export const getDeployStatus = async (): Promise<DeployStatus> => {
+  return await request.get('/deploy_status')
+}
+
--- a/BillNote_frontend/src/services/transcriber.ts
+++ b/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)
+}
--- a/BillNote_frontend/src/store/chatStore/index.ts
+++ b/BillNote_frontend/src/store/chatStore/index.ts
@@ -0,0 +1,43 @@
+import { create } from 'zustand'
+import { persist } from 'zustand/middleware'
+import type { ChatSource } from '@/services/chat'
+
+export interface ChatMessage {
+  role: 'user' | 'assistant'
+  content: string
+  sources?: ChatSource[]
+}
+
+interface ChatState {
+  chatHistory: Record<string, ChatMessage[]>
+  addMessage: (taskId: string, msg: ChatMessage) => void
+  clearChat: (taskId: string) => void
+  getMessages: (taskId: string) => ChatMessage[]
+}
+
+export const useChatStore = create<ChatState>()(
+  persist(
+    (set, get) => ({
+      chatHistory: {},
+
+      addMessage: (taskId, msg) =>
+        set(state => ({
+          chatHistory: {
+            ...state.chatHistory,
+            [taskId]: [...(state.chatHistory[taskId] || []), msg],
+          },
+        })),
+
+      clearChat: (taskId) =>
+        set(state => {
+          const { [taskId]: _, ...rest } = state.chatHistory
+          return { chatHistory: rest }
+        }),
+
+      getMessages: (taskId) => get().chatHistory[taskId] || [],
+    }),
+    {
+      name: 'bilinote-chat-storage',
+    },
+  ),
+)
--- a/BillNote_frontend/src/store/providerStore/index.ts
+++ b/BillNote_frontend/src/store/providerStore/index.ts
@@ -75,19 +75,19 @@ export const useProviderStore = create<ProviderStore>((set, get) => ({
  getProviderById: id => get().provider.find(p => p.id === id),
  updateProvider: async (provider: IProvider) => {
    try {
+      const existing = get().provider.find(p => p.id === provider.id)
+      const merged = { ...existing, ...provider }
+
      const data = {
-        ...provider,
-        api_key: provider.apiKey,
-        base_url: provider.baseUrl,
-      }
-      const res = await updateProviderById(data)
-      if (res.data.code === 0) {
-        const item = res.data.data
-        console.log('Provider ', item)
-        await get().fetchProviderList()
+        ...merged,
+        api_key: merged.apiKey,
+        base_url: merged.baseUrl,
      }
+      // 拦截器已解包：成功时直接返回 data 部分
+      await updateProviderById(data)
+      await get().fetchProviderList()
    } catch (error) {
-      console.error('Error fetching provider:', error)
+      console.error('Error updating provider:', error)
    }
  },
  getProviderList: () => get().provider,
--- a/BillNote_frontend/src/store/taskStore/index.ts
+++ b/BillNote_frontend/src/store/taskStore/index.ts
@@ -1,8 +1,9 @@
 import { create } from 'zustand'
-import { persist } from 'zustand/middleware'
+import { persist, createJSONStorage } from 'zustand/middleware'
 import { delete_task, generateNote } from '@/services/note.ts'
 import { v4 as uuidv4 } from 'uuid'
 import toast from 'react-hot-toast'
+import { get, set, del } from 'idb-keyval'


 export type TaskStatus = 'PENDING' | 'RUNNING' | 'SUCCESS' | 'FAILD'
@@ -211,6 +212,18 @@ export const useTaskStore = create<TaskStore>()(
    }),
    {
      name: 'task-storage',
+      storage: createJSONStorage(() => ({
+        getItem: async (name: string): Promise<string | null> => {
+          const value = await get(name)
+          return value ?? null
+        },
+        setItem: async (name: string, value: string): Promise<void> => {
+          await set(name, value)
+        },
+        removeItem: async (name: string): Promise<void> => {
+          await del(name)
+        },
+      })),
    }
  )
 )
--- a/BillNote_frontend/vite.config.ts
+++ b/BillNote_frontend/vite.config.ts
@@ -1,14 +1,19 @@
 import { defineConfig, loadEnv } from 'vite'
 import react from '@vitejs/plugin-react'
 import path from 'path'
+import { fileURLToPath } from 'url'
 import tailwindcss from '@tailwindcss/vite'

+const __dirname = path.dirname(fileURLToPath(import.meta.url))
+
 // https://vitejs.dev/config/
 export default defineConfig(({ mode }) => {
-  const env = loadEnv(mode, process.cwd() + '/../')
+  // 在 Docker 环境中，父目录可能没有 .env 文件，使用当前目录
+  const envDir = process.env.DOCKER_BUILD ? __dirname : path.resolve(__dirname, '../')
+  const env = loadEnv(mode, envDir)

-  const apiBaseUrl = env.VITE_API_BASE_URL || 'http://localhost:8000'
-  const port = env.VITE_FRONTEND_PORT || 3015
+  const apiBaseUrl = env.VITE_API_BASE_URL || 'http://127.0.0.1:8483'
+  const port = parseInt(env.VITE_FRONTEND_PORT || '3015', 10)

  return {
    base: './',
@@ -18,9 +23,21 @@ 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,
+      allowedHosts: true, // 允许任意域名访问
      proxy: {
        '/api': {
          target: apiBaseUrl,
--- a/CLAUDE.md
+++ b/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").
--- a/Dockerfile.complete
+++ b/Dockerfile.complete
@@ -0,0 +1,112 @@
+# === 阶段1：构建 Backend ===
+FROM python:3.11-slim AS backend-builder
+
+ARG APT_MIRROR=mirrors.tuna.tsinghua.edu.cn
+ARG PIP_INDEX=https://pypi.tuna.tsinghua.edu.cn/simple
+
+RUN set -ex && \
+    rm -f /etc/apt/sources.list && \
+    rm -rf /etc/apt/sources.list.d/* && \
+    echo "deb http://${APT_MIRROR}/debian bookworm main contrib non-free non-free-firmware" > /etc/apt/sources.list && \
+    echo "deb http://${APT_MIRROR}/debian bookworm-updates main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    echo "deb http://${APT_MIRROR}/debian-security bookworm-security main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    apt-get update && \
+    apt-get install -y --no-install-recommends ffmpeg && \
+    rm -rf /var/lib/apt/lists/*
+
+ENV PATH="/usr/bin:${PATH}"
+ENV HF_ENDPOINT=https://hf-mirror.com
+
+WORKDIR /tmp/backend
+
+# 先复制 requirements.txt 利用层缓存
+COPY ./backend/requirements.txt /tmp/backend/requirements.txt
+RUN pip install --no-cache-dir -i ${PIP_INDEX} -r requirements.txt
+
+COPY ./backend /tmp/backend
+
+# === 阶段2：构建 Frontend ===
+FROM node:18-alpine AS frontend-builder
+
+RUN corepack enable && corepack prepare pnpm@latest --activate
+
+WORKDIR /tmp/frontend
+
+# 先复制 package.json 利用依赖层缓存
+COPY ./BillNote_frontend/package.json ./
+RUN pnpm install
+
+COPY ./BillNote_frontend /tmp/frontend
+
+# 设置环境变量，告诉 vite.config.ts 这是 Docker 构建
+ENV DOCKER_BUILD=1
+RUN pnpm run build
+
+# === 阶段3：完整应用镜像 ===
+FROM python:3.11-slim
+
+ARG APT_MIRROR=mirrors.tuna.tsinghua.edu.cn
+
+# 安装必要的运行时依赖
+RUN set -ex && \
+    rm -f /etc/apt/sources.list && \
+    rm -rf /etc/apt/sources.list.d/* && \
+    echo "deb http://${APT_MIRROR}/debian bookworm main contrib non-free non-free-firmware" > /etc/apt/sources.list && \
+    echo "deb http://${APT_MIRROR}/debian bookworm-updates main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    echo "deb http://${APT_MIRROR}/debian-security bookworm-security main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    apt-get update && \
+    apt-get install -y --no-install-recommends ffmpeg nginx supervisor procps && \
+    rm -rf /var/lib/apt/lists/*
+
+ENV PATH="/usr/bin:${PATH}"
+ENV HF_ENDPOINT=https://hf-mirror.com
+ENV PYTHONUNBUFFERED=1
+
+# 复制 Python 依赖
+COPY --from=backend-builder /usr/local/lib/python3.11/site-packages /usr/local/lib/python3.11/site-packages
+COPY --from=backend-builder /usr/local/bin /usr/local/bin
+
+# 复制 backend 代码
+COPY ./backend /app/backend
+WORKDIR /app/backend
+
+# 复制前端静态文件到 nginx
+COPY --from=frontend-builder /tmp/frontend/dist /usr/share/nginx/html
+
+# 配置 nginx
+RUN rm -rf /etc/nginx/conf.d/default.conf
+COPY ./nginx/default.conf /etc/nginx/conf.d/default.conf
+
+# 创建 supervisor 配置
+RUN mkdir -p /var/log/supervisor
+COPY <<EOF /etc/supervisor/conf.d/supervisord.conf
+[supervisord]
+nodaemon=true
+user=root
+logfile=/var/log/supervisor/supervisord.log
+pidfile=/var/run/supervisord.pid
+
+[program:nginx]
+command=nginx -g "daemon off;"
+stdout_logfile=/var/log/supervisor/nginx.log
+stderr_logfile=/var/log/supervisor/nginx.log
+autorestart=true
+priority=10
+
+[program:backend]
+command=python main.py
+directory=/app/backend
+stdout_logfile=/var/log/supervisor/backend.log
+stderr_logfile=/var/log/supervisor/backend.log
+autorestart=true
+priority=20
+environment=BACKEND_PORT="8483",BACKEND_HOST="0.0.0.0"
+EOF
+
+# 修改 nginx 配置以使用本地 backend
+RUN sed -i 's/proxy_pass http:\/\/backend:8483/proxy_pass http:\/\/127.0.0.1:8483/g' /etc/nginx/conf.d/default.conf && \
+    sed -i 's/proxy_pass http:\/\/frontend:80/proxy_pass http:\/\/127.0.0.1:8080/g' /etc/nginx/conf.d/default.conf
+
+# 启动 supervisor
+EXPOSE 80
+CMD ["/usr/bin/supervisord", "-c", "/etc/supervisor/conf.d/supervisord.conf"]
--- a/README.md
+++ b/README.md
@@ -3,17 +3,17 @@
    <p align="center">
  <img src="./doc/icon.svg" alt="BiliNote Banner" width="50" height="50"  />
 </p>
-<h1 align="center" > BiliNote v1.8.1</h1>
+<h1 align="center" > BiliNote v2.0.0</h1>
 </div>

 <p align="center"><i>AI 视频笔记生成工具 让 AI 为你的视频做笔记</i></p>

 <p align="center">
  <img src="https://img.shields.io/badge/license-MIT-blue.svg" />
-  <img src="https://img.shields.io/badge/frontend-react-blue" />
+  <img src="https://img.shields.io/badge/frontend-react%2019-blue" />
  <img src="https://img.shields.io/badge/backend-fastapi-green" />
  <img src="https://img.shields.io/badge/GPT-openai%20%7C%20deepseek%20%7C%20qwen-ff69b4" />
-  <img src="https://img.shields.io/badge/docker-compose-blue" />
+  <img src="https://img.shields.io/badge/docker-ghcr.io-blue" />
  <img src="https://img.shields.io/badge/status-active-success" />
  <img src="https://img.shields.io/github/stars/jefferyhcool/BiliNote?style=social" />
 </p>
@@ -22,30 +22,48 @@

 ## ✨ 项目简介

-BiliNote 是一个开源的 AI 视频笔记助手，支持通过哔哩哔哩、YouTube、抖音等视频链接，自动提取内容并生成结构清晰、重点明确的 Markdown 格式笔记。支持插入截图、原片跳转等功能。
+BiliNote 是一个开源的 AI 视频笔记助手，支持通过哔哩哔哩、YouTube、抖音等视频链接，自动提取内容并生成结构清晰、重点明确的 Markdown 格式笔记。支持插入截图、原片跳转、AI 问答等功能。
+
 ## 📝 使用文档
 详细文档可以查看[这里](https://docs.bilinote.app/)

 ## 体验地址
 可以通过访问 [这里](https://www.bilinote.app/) 进行体验，速度略慢，不支持长视频。
-## 📦 Windows 打包版
-本项目提供了 Windows 系统的 exe 文件，可在[release](https://github.com/JefferyHcool/BiliNote/releases/tag/v1.1.1)进行下载。**注意一定要在没有中文路径的环境下运行。**

+## 📦 桌面版下载
+本项目提供了 Windows 和 macOS 桌面客户端，可在 [Releases](https://github.com/JefferyHcool/BiliNote/releases) 页面下载最新版本。
+
+> Windows 用户请注意：一定要在没有中文路径的环境下运行。

 ## 🔧 功能特性

- 支持多平台：Bilibili、YouTube、本地视频、抖音（后续会加入更多平台）
+- 支持多平台：Bilibili、YouTube、本地视频、抖音、快手
 - 支持返回笔记格式选择
 - 支持笔记风格选择
 - 支持多模态视频理解
 - 支持多版本记录保留
- 支持自行配置 GPT 大模型
- 本地模型音频转写（支持 Fast-Whisper）
+- 支持自行配置 GPT 大模型（OpenAI、DeepSeek、Qwen 等）
+- 本地模型音频转写（支持 Fast-Whisper、MLX-Whisper、Groq、BCut）
 - GPT 大模型总结视频内容
 - 自动生成结构化 Markdown 笔记
 - 可选插入截图（自动截取）
 - 可选内容跳转链接（关联原视频）
 - 任务记录与历史回看
+- 基于 RAG 的笔记内容 AI 问答（支持 Function Calling）
+- 笔记顶部视频封面 Banner 展示
+- 工作区和生成历史面板支持折叠/展开
+
+### v2.0.0 新增
+
+- 基于 RAG 的笔记内容 AI 问答功能，支持半屏/全屏模式
+- AI 问答支持 Function Calling，模型可主动查询原文数据
+- RAG 索引支持视频元信息（标题、作者、简介、标签等）
+- AI 回复支持 Markdown 渲染
+- 笔记顶部新增视频封面 Banner
+- 工作区和生成历史面板支持折叠/展开
+- 笔记开头添加来源链接功能
+- YouTube 字幕优先获取，有字幕时跳过音频下载
+- 性能优化与转写器配置改进

 ## 📸 截图预览
 ![screenshot](./doc/image1.png)
@@ -56,7 +74,34 @@ BiliNote 是一个开源的 AI 视频笔记助手，支持通过哔哩哔哩、Y

 ## 🚀 快速开始

-### 1. 克隆仓库
+### 方式一：Docker 部署（推荐）
+
+确保已安装 Docker，直接拉取预构建镜像运行：
+
+```bash
+docker pull ghcr.io/jefferyhcool/bilinote:latest
+
+docker run -d -p 80:80 \
+  -v bilinote-data:/app/backend/data \
+  --name bilinote \
+  ghcr.io/jefferyhcool/bilinote:latest
+```
+
+访问：`http://localhost`
+
+也可以使用 docker-compose 本地构建：
+
+```bash
+# 标准部署
+docker-compose up -d
+
+# GPU 加速部署（需要 NVIDIA GPU）
+docker-compose -f docker-compose.gpu.yml up -d
+```
+
+### 方式二：源码部署
+
+#### 1. 克隆仓库

 ```bash
 git clone https://github.com/JefferyHcool/BiliNote.git
@@ -64,7 +109,7 @@ cd BiliNote
 mv .env.example .env
 ```

-### 2. 启动后端（FastAPI）
+#### 2. 启动后端（FastAPI）

 ```bash
 cd backend
@@ -72,7 +117,7 @@ pip install -r requirements.txt
 python main.py
 ```

-### 3. 启动前端（Vite + React）
+#### 3. 启动前端（Vite + React）

 ```bash
 cd BillNote_frontend
@@ -80,11 +125,12 @@ pnpm install
 pnpm dev
 ```

-访问：`http://localhost:5173`
+访问：`http://localhost:3015`

 ## ⚙️ 依赖说明
+
 ### 🎬 FFmpeg
-本项目依赖 ffmpeg 用于音频处理与转码，必须安装：
+本项目依赖 ffmpeg 用于音频处理与转码，源码部署时必须安装：
 ```bash
 # Mac (brew)
 brew install ffmpeg
@@ -96,6 +142,8 @@ sudo apt install ffmpeg
 # 请从官网下载安装：https://ffmpeg.org/download.html
 ```
 > ⚠️ 若系统无法识别 ffmpeg，请将其加入系统环境变量 PATH
+>
+> Docker 部署已内置 FFmpeg，无需额外安装。

 ### 🚀 CUDA 加速（可选）
 若你希望更快地执行音频转写任务，可使用具备 NVIDIA GPU 的机器，并启用 fast-whisper + CUDA 加速版本：
@@ -104,24 +152,43 @@ sudo apt install ffmpeg

 ### 🐳 使用 Docker 一键部署

-确保你已安装 Docker 和 Docker Compose：
+确保你已安装 Docker，然后直接拉取预构建镜像运行：

-[docker 部署](https://github.com/JefferyHcool/bilinote-deploy/blob/master/README.md)
+```bash
+# 拉取最新镜像
+docker pull ghcr.io/jefferyhcool/bilinote:latest
+
+# 运行容器
+docker run -d -p 80:80 \
+  -v bilinote-data:/app/backend/data \
+  --name bilinote \
+  ghcr.io/jefferyhcool/bilinote:latest
+```
+
+访问：`http://localhost`
+
+也可以使用 docker-compose 本地构建：
+
+```bash
+# 标准部署
+docker-compose up -d
+
+# GPU 加速部署（需要 NVIDIA GPU）
+docker-compose -f docker-compose.gpu.yml up -d
+```

 ## 🧠 TODO

 - [x] 支持抖音及快手等视频平台
 - [x] 支持前端设置切换 AI 模型切换、语音转文字模型
 - [x] AI 摘要风格自定义（学术风、口语风、重点提取等）
- [ ] 笔记导出为 PDF / Word / Notion
 - [x] 加入更多模型支持
 - [x] 加入更多音频转文本模型支持
+- [x] 基于 RAG 的笔记内容 AI 问答
+- [ ] 笔记导出为 PDF / Word / Notion

 ### Contact and Join-联系和加入社区
- BiliNote 交流QQ群：785367111
- BiliNote 交流微信群:
-  
-  <img src="doc/wechat.png" alt="wechat" style="zoom:33%;" />
+年会恢复更新以后放出最新社区地址



--- a/backend/Dockerfile
+++ b/backend/Dockerfile
@@ -1,23 +1,27 @@
 FROM python:3.11-slim

+ARG APT_MIRROR=mirrors.tuna.tsinghua.edu.cn
+ARG PIP_INDEX=https://pypi.tuna.tsinghua.edu.cn/simple

 RUN rm -f /etc/apt/sources.list && \
    rm -rf /etc/apt/sources.list.d/* && \
-    echo "deb https://mirrors.tuna.tsinghua.edu.cn/debian bookworm main contrib non-free non-free-firmware" > /etc/apt/sources.list && \
-    echo "deb https://mirrors.tuna.tsinghua.edu.cn/debian bookworm-updates main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
-    echo "deb https://mirrors.tuna.tsinghua.edu.cn/debian-security bookworm-security main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    echo "deb https://${APT_MIRROR}/debian bookworm main contrib non-free non-free-firmware" > /etc/apt/sources.list && \
+    echo "deb https://${APT_MIRROR}/debian bookworm-updates main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
+    echo "deb https://${APT_MIRROR}/debian-security bookworm-security main contrib non-free non-free-firmware" >> /etc/apt/sources.list && \
    apt-get update && \
-    apt-get install -y ffmpeg && \
+    apt-get install -y --no-install-recommends ffmpeg curl && \
    rm -rf /var/lib/apt/lists/*

-# 确保 PATH 中包含 ffmpeg 路径（可选）
 ENV PATH="/usr/bin:${PATH}"
-
-# 设置 Hugging Face 镜像源环境变量
 ENV HF_ENDPOINT=https://hf-mirror.com

 WORKDIR /app
+
+# 先复制 requirements.txt 利用层缓存
+COPY ./backend/requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -i ${PIP_INDEX} -r requirements.txt
+
+# 再复制应用代码（频繁变动不影响 pip 缓存层）
 COPY ./backend /app
-RUN pip install --no-cache-dir -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt

 CMD ["python", "main.py"]
--- a/backend/Dockerfile.gpu
+++ b/backend/Dockerfile.gpu
@@ -0,0 +1,27 @@
+FROM nvidia/cuda:12.4.1-cudnn-runtime-ubuntu22.04
+
+ARG APT_MIRROR=mirrors.tuna.tsinghua.edu.cn
+ARG PIP_INDEX=https://pypi.tuna.tsinghua.edu.cn/simple
+
+RUN rm -f /etc/apt/sources.list && \
+    rm -rf /etc/apt/sources.list.d/* && \
+    echo "deb https://${APT_MIRROR}/ubuntu jammy main restricted universe multiverse" > /etc/apt/sources.list && \
+    echo "deb https://${APT_MIRROR}/ubuntu jammy-updates main restricted universe multiverse" >> /etc/apt/sources.list && \
+    echo "deb https://${APT_MIRROR}/ubuntu jammy-security main restricted universe multiverse" >> /etc/apt/sources.list && \
+    apt-get update && \
+    apt-get install -y --no-install-recommends ffmpeg python3-pip curl && \
+    rm -rf /var/lib/apt/lists/*
+
+ENV HF_ENDPOINT=https://hf-mirror.com
+
+WORKDIR /app
+
+# 先复制 requirements.txt 利用层缓存
+COPY ./backend/requirements.txt /app/requirements.txt
+RUN pip install --no-cache-dir -i ${PIP_INDEX} -r requirements.txt && \
+    pip install --no-cache-dir -i ${PIP_INDEX} 'transformers[torch]>=4.23'
+
+# 再复制应用代码
+COPY ./backend /app
+
+CMD ["python3", "main.py"]
--- a/backend/app/init.py
+++ b/backend/app/init.py
@@ -1,6 +1,6 @@
 from fastapi import FastAPI

-from .routers import note, provider, model, config
+from .routers import note, provider, model, config, chat



@@ -10,5 +10,6 @@ def create_app(lifespan) -> FastAPI:
    app.include_router(provider.router, prefix="/api")
    app.include_router(model.router,prefix="/api")
    app.include_router(config.router,  prefix="/api")
+    app.include_router(chat.router, prefix="/api")

    return app
--- a/backend/app/db/engine.py
+++ b/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)
--- a/backend/app/db/model_dao.py
+++ b/backend/app/db/model_dao.py
@@ -1,5 +1,6 @@
 from app.db.engine import get_db
 from app.db.models.models import Model
+from app.db.models.providers import Provider


 def get_model_by_provider_and_name(provider_id: int, model_name: str):
@@ -58,7 +59,8 @@ def delete_model(model_id: int):
 def get_all_models():
    db = next(get_db())
    try:
-        models = db.query(Model).all()
+        # 只查询启用状态供应商的模型
+        models = db.query(Model).join(Provider, Model.provider_id == Provider.id).filter(Provider.enabled == 1).all()
        return [
            {"id": m.id, "provider_id": m.provider_id, "model_name": m.model_name}
            for m in models
--- a/backend/app/downloaders/base.py
+++ b/backend/app/downloaders/base.py
@@ -5,6 +5,7 @@ from typing import Optional, Union

 from app.enmus.note_enums import DownloadQuality
 from app.models.notes_model import AudioDownloadResult
+from app.models.transcriber_model import TranscriptResult
 from os import getenv
 QUALITY_MAP = {
    "fast": "32",
@@ -21,7 +22,8 @@ class Downloader(ABC):

    @abstractmethod
    def download(self, video_url: str, output_dir: str = None,
-                 quality: DownloadQuality = "fast", need_video: Optional[bool] = False) -> AudioDownloadResult:
+                 quality: DownloadQuality = "fast", need_video: Optional[bool] = False,
+                 skip_download: bool = False) -> AudioDownloadResult:
        '''

        :param need_video:
@@ -36,3 +38,15 @@ class Downloader(ABC):
    def download_video(self, video_url: str,
                       output_dir: Union[str, None] = None) -> str:
        pass
+
+    def download_subtitles(self, video_url: str, output_dir: str = None,
+                           langs: list = None) -> Optional[TranscriptResult]:
+        '''
+        尝试获取平台字幕（人工字幕或自动生成字幕）
+
+        :param video_url: 视频链接
+        :param output_dir: 输出路径
+        :param langs: 优先语言列表，如 ['zh-Hans', 'zh', 'en']
+        :return: TranscriptResult 或 None（无字幕时）
+        '''
+        return None
--- a/backend/app/downloaders/bilibili_downloader.py
+++ b/backend/app/downloaders/bilibili_downloader.py
@@ -1,18 +1,44 @@
 import os
+import json
+import logging
+import tempfile
 from abc import ABC
-from typing import Union, Optional
+from typing import Union, Optional, List

 import yt_dlp

 from app.downloaders.base import Downloader, DownloadQuality, QUALITY_MAP
 from app.models.notes_model import AudioDownloadResult
+from app.models.transcriber_model import TranscriptResult, TranscriptSegment
 from app.utils.path_helper import get_data_dir
 from app.utils.url_parser import extract_video_id
+from app.services.cookie_manager import CookieConfigManager
+
+logger = logging.getLogger(__name__)


 class BilibiliDownloader(Downloader, ABC):
    def __init__(self):
        super().__init__()
+        self._cookie_mgr = CookieConfigManager()
+        self._cookie = self._cookie_mgr.get('bilibili')
+        self._cookiefile = self._write_netscape_cookie_file()
+
+    def _write_netscape_cookie_file(self) -> Optional[str]:
+        """将 Cookie 写入 Netscape 格式临时文件，返回文件路径（供 yt-dlp cookiefile 使用）"""
+        if not self._cookie:
+            logger.warning("B站 Cookie 未配置，下载可能失败")
+            return None
+        lines = ["# Netscape HTTP Cookie File\n"]
+        for pair in self._cookie.split("; "):
+            if "=" in pair:
+                key, value = pair.split("=", 1)
+                lines.append(f".bilibili.com\tTRUE\t/\tFALSE\t0\t{key}\t{value}\n")
+        tmp = tempfile.NamedTemporaryFile(mode='w', suffix='.txt', delete=False, encoding='utf-8')
+        tmp.writelines(lines)
+        tmp.close()
+        logger.info("已生成 B站 Netscape Cookie 文件: %s (条目: %d)", tmp.name, len(lines) - 1)
+        return tmp.name

    def download(
        self,
@@ -32,6 +58,7 @@ class BilibiliDownloader(Downloader, ABC):
        ydl_opts = {
            'format': 'bestaudio[ext=m4a]/bestaudio/best',
            'outtmpl': output_path,
+            'http_headers': {'Referer': 'https://www.bilibili.com'},
            'postprocessors': [
                {
                    'key': 'FFmpegExtractAudio',
@@ -42,6 +69,8 @@ class BilibiliDownloader(Downloader, ABC):
            'noplaylist': True,
            'quiet': False,
        }
+        if self._cookiefile:
+            ydl_opts['cookiefile'] = self._cookiefile

        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            info = ydl.extract_info(video_url, download=True)
@@ -88,10 +117,13 @@ class BilibiliDownloader(Downloader, ABC):
        ydl_opts = {
            'format': 'bv*[ext=mp4]/bestvideo+bestaudio/best',
            'outtmpl': output_path,
+            'http_headers': {'Referer': 'https://www.bilibili.com'},
            'noplaylist': True,
            'quiet': False,
            'merge_output_format': 'mp4',  # 确保合并成 mp4
        }
+        if self._cookiefile:
+            ydl_opts['cookiefile'] = self._cookiefile

        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
            info = ydl.extract_info(video_url, download=True)
@@ -111,4 +143,191 @@ class BilibiliDownloader(Downloader, ABC):
            os.remove(video_path)
            return f"视频文件已删除: {video_path}"
        else:
-            return f"视频文件未找到: {video_path}"
+            return f"视频文件未找到: {video_path}"
+
+    def download_subtitles(self, video_url: str, output_dir: str = None,
+                           langs: List[str] = None) -> Optional[TranscriptResult]:
+        """
+        尝试获取B站视频字幕
+
+        :param video_url: 视频链接
+        :param output_dir: 输出路径
+        :param langs: 优先语言列表
+        :return: TranscriptResult 或 None
+        """
+        if output_dir is None:
+            output_dir = get_data_dir()
+        if not output_dir:
+            output_dir = self.cache_data
+        os.makedirs(output_dir, exist_ok=True)
+
+        if langs is None:
+            langs = ['zh-Hans', 'zh', 'zh-CN', 'ai-zh', 'en', 'en-US']
+
+        video_id = extract_video_id(video_url, "bilibili")
+
+        ydl_opts = {
+            'writesubtitles': True,
+            'writeautomaticsub': True,
+            'subtitleslangs': langs,
+            'subtitlesformat': 'srt/json3/best',  # 支持多种格式
+            'skip_download': True,
+            'outtmpl': os.path.join(output_dir, f'{video_id}.%(ext)s'),
+            'quiet': True,
+        }
+
+        # 通过 CookieConfigManager 注入 B站 Cookie（Netscape cookiefile）
+        if self._cookiefile:
+            ydl_opts['cookiefile'] = self._cookiefile
+            ydl_opts['http_headers'] = {'Referer': 'https://www.bilibili.com'}
+
+        try:
+            with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+                info = ydl.extract_info(video_url, download=True)
+
+                # 查找下载的字幕文件
+                subtitles = info.get('requested_subtitles') or {}
+                if not subtitles:
+                    logger.info(f"B站视频 {video_id} 没有可用字幕")
+                    return None
+
+                # 按优先级查找字幕
+                detected_lang = None
+                sub_info = None
+                for lang in langs:
+                    if lang in subtitles:
+                        detected_lang = lang
+                        sub_info = subtitles[lang]
+                        break
+
+                # 如果按优先级没找到，取第一个可用的（排除弹幕）
+                if not detected_lang:
+                    for lang, info_item in subtitles.items():
+                        if lang != 'danmaku':  # 排除弹幕
+                            detected_lang = lang
+                            sub_info = info_item
+                            break
+
+                if not sub_info:
+                    logger.info(f"B站视频 {video_id} 没有可用字幕（排除弹幕）")
+                    return None
+
+                # 检查是否有内嵌数据（yt-dlp 有时直接返回字幕内容）
+                if 'data' in sub_info and sub_info['data']:
+                    logger.info(f"直接从返回数据解析字幕: {detected_lang}")
+                    return self._parse_srt_content(sub_info['data'], detected_lang)
+
+                # 查找字幕文件
+                ext = sub_info.get('ext', 'srt')
+                subtitle_file = os.path.join(output_dir, f"{video_id}.{detected_lang}.{ext}")
+
+                if not os.path.exists(subtitle_file):
+                    logger.info(f"字幕文件不存在: {subtitle_file}")
+                    return None
+
+                # 根据格式解析字幕文件
+                if ext == 'json3':
+                    return self._parse_json3_subtitle(subtitle_file, detected_lang)
+                else:
+                    with open(subtitle_file, 'r', encoding='utf-8') as f:
+                        return self._parse_srt_content(f.read(), detected_lang)
+
+        except Exception as e:
+            logger.warning(f"获取B站字幕失败: {e}")
+            return None
+
+    def _parse_srt_content(self, srt_content: str, language: str) -> Optional[TranscriptResult]:
+        """
+        解析 SRT 格式字幕内容
+
+        :param srt_content: SRT 字幕文本内容
+        :param language: 语言代码
+        :return: TranscriptResult
+        """
+        import re
+        try:
+            segments = []
+            # SRT 格式: 序号\n时间戳\n文本\n\n
+            pattern = r'(\d+)\n(\d{2}:\d{2}:\d{2},\d{3})\s*-->\s*(\d{2}:\d{2}:\d{2},\d{3})\n(.*?)(?=\n\n|\n\d+\n|$)'
+            matches = re.findall(pattern, srt_content, re.DOTALL)
+
+            for match in matches:
+                idx, start_time, end_time, text = match
+                text = text.strip()
+                if not text:
+                    continue
+
+                # 转换时间格式 00:00:00,000 -> 秒
+                def time_to_seconds(t):
+                    parts = t.replace(',', '.').split(':')
+                    return float(parts[0]) * 3600 + float(parts[1]) * 60 + float(parts[2])
+
+                segments.append(TranscriptSegment(
+                    start=time_to_seconds(start_time),
+                    end=time_to_seconds(end_time),
+                    text=text
+                ))
+
+            if not segments:
+                return None
+
+            full_text = ' '.join(seg.text for seg in segments)
+            logger.info(f"成功解析B站SRT字幕，共 {len(segments)} 段")
+            return TranscriptResult(
+                language=language,
+                full_text=full_text,
+                segments=segments,
+                raw={'source': 'bilibili_subtitle', 'format': 'srt'}
+            )
+
+        except Exception as e:
+            logger.warning(f"解析SRT字幕失败: {e}")
+            return None
+
+    def _parse_json3_subtitle(self, subtitle_file: str, language: str) -> Optional[TranscriptResult]:
+        """
+        解析 json3 格式字幕文件
+
+        :param subtitle_file: 字幕文件路径
+        :param language: 语言代码
+        :return: TranscriptResult
+        """
+        try:
+            with open(subtitle_file, 'r', encoding='utf-8') as f:
+                data = json.load(f)
+
+            segments = []
+            events = data.get('events', [])
+
+            for event in events:
+                # json3 格式中时间单位是毫秒
+                start_ms = event.get('tStartMs', 0)
+                duration_ms = event.get('dDurationMs', 0)
+
+                # 提取文本
+                segs = event.get('segs', [])
+                text = ''.join(seg.get('utf8', '') for seg in segs).strip()
+
+                if text:  # 只添加非空文本
+                    segments.append(TranscriptSegment(
+                        start=start_ms / 1000.0,
+                        end=(start_ms + duration_ms) / 1000.0,
+                        text=text
+                    ))
+
+            if not segments:
+                return None
+
+            full_text = ' '.join(seg.text for seg in segments)
+
+            logger.info(f"成功解析B站字幕，共 {len(segments)} 段")
+            return TranscriptResult(
+                language=language,
+                full_text=full_text,
+                segments=segments,
+                raw={'source': 'bilibili_subtitle', 'file': subtitle_file}
+            )
+
+        except Exception as e:
+            logger.warning(f"解析字幕文件失败: {e}")
+            return None
--- a/backend/app/downloaders/youtube_downloader.py
+++ b/backend/app/downloaders/youtube_downloader.py
@@ -1,14 +1,19 @@
 import os
+import logging
 from abc import ABC
-from typing import Union, Optional
+from typing import Union, Optional, List

 import yt_dlp

 from app.downloaders.base import Downloader, DownloadQuality
+from app.downloaders.youtube_subtitle import YouTubeSubtitleFetcher
 from app.models.notes_model import AudioDownloadResult
+from app.models.transcriber_model import TranscriptResult
 from app.utils.path_helper import get_data_dir
 from app.utils.url_parser import extract_video_id

+logger = logging.getLogger(__name__)
+

 class YoutubeDownloader(Downloader, ABC):
    def __init__(self):
@@ -20,12 +25,13 @@ class YoutubeDownloader(Downloader, ABC):
        video_url: str,
        output_dir: Union[str, None] = None,
        quality: DownloadQuality = "fast",
-        need_video:Optional[bool]=False
+        need_video: Optional[bool] = False,
+        skip_download: bool = False,
    ) -> AudioDownloadResult:
        if output_dir is None:
            output_dir = get_data_dir()
        if not output_dir:
-            output_dir=self.cache_data
+            output_dir = self.cache_data
        os.makedirs(output_dir, exist_ok=True)

        output_path = os.path.join(output_dir, "%(id)s.%(ext)s")
@@ -37,15 +43,17 @@ class YoutubeDownloader(Downloader, ABC):
            'quiet': False,
        }

+        if skip_download:
+            ydl_opts['skip_download'] = True
+
        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
-            info = ydl.extract_info(video_url, download=True)
+            info = ydl.extract_info(video_url, download=not skip_download)
            video_id = info.get("id")
            title = info.get("title")
            duration = info.get("duration", 0)
            cover_url = info.get("thumbnail")
-            ext = info.get("ext", "m4a")  # 兜底用 m4a
+            ext = info.get("ext", "m4a")
            audio_path = os.path.join(output_dir, f"{video_id}.{ext}")
-        print('os.path.join(output_dir, f"{video_id}.{ext}")',os.path.join(output_dir, f"{video_id}.{ext}"))

        return AudioDownloadResult(
            file_path=audio_path,
@@ -54,8 +62,8 @@ class YoutubeDownloader(Downloader, ABC):
            cover_url=cover_url,
            platform="youtube",
            video_id=video_id,
-            raw_info={'tags':info.get('tags')}, #全部返回会报错
-            video_path=None  # ❗音频下载不包含视频路径
+            raw_info={'tags': info.get('tags')},
+            video_path=None,
        )

    def download_video(
@@ -92,3 +100,24 @@ class YoutubeDownloader(Downloader, ABC):
            raise FileNotFoundError(f"视频文件未找到: {video_path}")

        return video_path
+
+    def download_subtitles(self, video_url: str, output_dir: str = None,
+                           langs: List[str] = None) -> Optional[TranscriptResult]:
+        """
+        通过 YouTube InnerTube API 直接获取字幕（优先人工字幕，其次自动生成）。
+        比 yt_dlp 方式更轻量，无需写临时文件到磁盘。
+
+        :param video_url: 视频链接
+        :param output_dir: 未使用（保留接口兼容）
+        :param langs: 优先语言列表
+        :return: TranscriptResult 或 None
+        """
+        if langs is None:
+            langs = ['zh-Hans', 'zh', 'zh-CN', 'zh-TW', 'en', 'en-US', 'ja']
+
+        video_id = extract_video_id(video_url, "youtube")
+        fetcher = YouTubeSubtitleFetcher()
+        print(
+            f"尝试获取字幕，video_id={video_id}, langs={langs}"
+        )
+        return fetcher.fetch_subtitles(video_id, langs)
--- a/backend/app/downloaders/youtube_subtitle.py
+++ b/backend/app/downloaders/youtube_subtitle.py
@@ -0,0 +1,98 @@
+"""
+通过 youtube-transcript-api 获取 YouTube 字幕。
+优先人工字幕，其次自动生成字幕。不依赖 yt_dlp，无需下载任何文件。
+"""
+
+from typing import Optional, List
+
+from youtube_transcript_api import YouTubeTranscriptApi
+
+from app.models.transcriber_model import TranscriptResult, TranscriptSegment
+from app.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+
+class YouTubeSubtitleFetcher:
+    """通过 youtube-transcript-api 获取 YouTube 字幕。"""
+
+    def __init__(self):
+        self._api = YouTubeTranscriptApi()
+
+    def fetch_subtitles(
+        self,
+        video_id: str,
+        langs: Optional[List[str]] = None,
+    ) -> Optional[TranscriptResult]:
+        if langs is None:
+            langs = ["zh-Hans", "zh", "zh-CN", "zh-TW", "en", "en-US", "ja"]
+
+        try:
+            # 1. 列出所有可用字幕
+            transcript_list = self._api.list(video_id)
+
+            available = []
+            for t in transcript_list:
+                available.append(
+                    f"{t.language_code}({'auto' if t.is_generated else 'manual'})"
+                )
+            logger.info(f"可用字幕轨道: {', '.join(available)}")
+
+            # 2. 按优先级查找：先人工字幕，再自动字幕
+            transcript = None
+            try:
+                transcript = transcript_list.find_manually_created_transcript(langs)
+                logger.info(f"选中人工字幕: {transcript.language_code} ({transcript.language})")
+            except Exception:
+                try:
+                    transcript = transcript_list.find_generated_transcript(langs)
+                    logger.info(f"选中自动字幕: {transcript.language_code} ({transcript.language})")
+                except Exception:
+                    # 都没匹配，取第一个可用的
+                    for t in transcript_list:
+                        transcript = t
+                        source = "auto" if t.is_generated else "manual"
+                        logger.info(f"使用首个可用字幕: {t.language_code} ({source})")
+                        break
+
+            if not transcript:
+                logger.info(f"YouTube 视频 {video_id} 没有任何可用字幕")
+                return None
+
+            # 3. 获取字幕内容
+            fetched = transcript.fetch()
+            segments = []
+            for snippet in fetched:
+                text = snippet.get("text", "").strip() if isinstance(snippet, dict) else str(snippet).strip()
+                if not text:
+                    continue
+                start = snippet.get("start", 0) if isinstance(snippet, dict) else 0
+                duration = snippet.get("duration", 0) if isinstance(snippet, dict) else 0
+                segments.append(TranscriptSegment(
+                    start=float(start),
+                    end=float(start) + float(duration),
+                    text=text,
+                ))
+
+            if not segments:
+                logger.warning(f"YouTube 字幕内容为空: {video_id}")
+                return None
+
+            full_text = " ".join(seg.text for seg in segments)
+            logger.info(f"成功获取 YouTube 字幕，共 {len(segments)} 段")
+
+            return TranscriptResult(
+                language=transcript.language_code,
+                full_text=full_text,
+                segments=segments,
+                raw={
+                    "source": "youtube_transcript_api",
+                    "language": transcript.language,
+                    "language_code": transcript.language_code,
+                    "is_generated": transcript.is_generated,
+                },
+            )
+
+        except Exception as e:
+            logger.warning(f"YouTube 字幕获取失败: {e}")
+            return None
--- a/backend/app/gpt/prompt.py
+++ b/backend/app/gpt/prompt.py
@@ -18,12 +18,12 @@ BASE_PROMPT = '''
 - **不要**将输出包裹在代码块中（例如：```` ```markdown ````，```` ``` ````）。
 请注意，在生成 Markdown 时，避免将编号标题（如“1. **内容**”）写成有序列表的格式，以免解析错误。

- 如果要加粗并保留编号，应使用 `1\. **内容**`（加反斜杠），防止被误解析为有序列表。
+- 如果要加粗并保留编号，应使用 `1\\. **内容**`（加反斜杠），防止被误解析为有序列表。
 - 或者使用 `## 1. 内容` 的形式作为标题。

 请确保以下格式 **不会出现误渲染**：
 `1. **xxx**`
- `1\. **xxx**` 或 `## 1. xxx`
+ `1\\. **xxx**` 或 `## 1. xxx`

 视频分段（格式：开始时间 - 内容）：

@@ -66,4 +66,13 @@ SCREENSHOT='''
 8. **Screenshot placeholders**: If a section involves **visual demonstrations, code walkthroughs, UI interactions**, or any content where visuals aid understanding, insert a screenshot cue at the end of that section:
   - Format: `*Screenshot-[mm:ss]`
   - Only use it when truly helpful.
-'''
+'''
+
+MERGE_PROMPT = '''
+你将收到多个来自同一视频的 Markdown 笔记片段，请合并成一份完整笔记：
+- 只做合并与去重，不要发明新内容
+- 保持原有标题层级与 Markdown 结构
+- 保留所有 *Content-[mm:ss] 与 *Screenshot-[mm:ss] 标记
+- 保持中文输出，专有名词保留英文
+- 不要使用代码块包裹输出
+'''
--- a/backend/app/gpt/request_chunker.py
+++ b/backend/app/gpt/request_chunker.py
@@ -0,0 +1,161 @@
+from dataclasses import dataclass
+from typing import Callable, List, Optional
+
+
+@dataclass
+class ChunkPayload:
+    segments: list
+    image_urls: list
+
+
+class RequestChunker:
+    def __init__(self, message_builder: Callable, max_bytes: int, size_estimator: Optional[Callable] = None):
+        self.message_builder = message_builder
+        self.max_bytes = max_bytes
+        self.size_estimator = size_estimator
+
+    def estimate(self, messages) -> int:
+        if self.size_estimator:
+            return self.size_estimator(messages)
+        import json
+        return len(json.dumps(messages, ensure_ascii=False).encode("utf-8"))
+
+    def _messages_size(self, segments, image_urls, **kwargs) -> int:
+        messages = self.message_builder(segments, image_urls, **kwargs)
+        return self.estimate(messages)
+
+    def _get_text(self, segment) -> str:
+        if isinstance(segment, dict):
+            return segment.get("text", "")
+        return getattr(segment, "text", "")
+
+    def _make_segment(self, segment, text: str):
+        if isinstance(segment, dict):
+            new_seg = dict(segment)
+            new_seg["text"] = text
+            return new_seg
+        if hasattr(segment, "__dict__"):
+            data = dict(segment.__dict__)
+            data["text"] = text
+            return type(segment)(**data)
+        return type(segment)(segment.start, segment.end, text)
+
+    def _split_segment_to_fit(self, segment, **kwargs):
+        text = self._get_text(segment)
+        if not text:
+            raise ValueError("empty segment cannot be split")
+        lo, hi = 1, len(text)
+        best = None
+        while lo <= hi:
+            mid = (lo + hi) // 2
+            candidate = self._make_segment(segment, text[:mid])
+            size = self._messages_size([candidate], [], **kwargs)
+            if size <= self.max_bytes:
+                best = mid
+                lo = mid + 1
+            else:
+                hi = mid - 1
+        if best is None:
+            raise ValueError("single segment too large to fit request")
+        head = self._make_segment(segment, text[:best])
+        tail = self._make_segment(segment, text[best:])
+        return head, tail
+
+    def chunk(self, segments: list, image_urls: list, **kwargs) -> List[ChunkPayload]:
+        segments = list(segments or [])
+        image_urls = list(image_urls or [])
+        if not segments and not image_urls:
+            return []
+
+        chunks: List[ChunkPayload] = []
+        seg_idx = 0
+
+        while seg_idx < len(segments):
+            batch_segments = []
+            while seg_idx < len(segments):
+                candidate = batch_segments + [segments[seg_idx]]
+                size = self._messages_size(candidate, [], **kwargs)
+                if size <= self.max_bytes:
+                    batch_segments = candidate
+                    seg_idx += 1
+                    continue
+                if not batch_segments:
+                    head, tail = self._split_segment_to_fit(segments[seg_idx], **kwargs)
+                    segments[seg_idx] = head
+                    segments.insert(seg_idx + 1, tail)
+                    continue
+                break
+
+            if not batch_segments:
+                raise ValueError("unable to fit any content into chunk")
+
+            chunks.append(ChunkPayload(segments=batch_segments, image_urls=[]))
+
+        if not image_urls:
+            return chunks
+
+        if not chunks:
+            chunks = [ChunkPayload(segments=[], image_urls=[])]
+
+        if not segments:
+            for image in image_urls:
+                appended = False
+                for chunk in chunks[-1:]:
+                    candidate_images = chunk.image_urls + [image]
+                    if self._messages_size(chunk.segments, candidate_images, **kwargs) <= self.max_bytes:
+                        chunk.image_urls = candidate_images
+                        appended = True
+                        break
+
+                if appended:
+                    continue
+
+                if self._messages_size([], [image], **kwargs) > self.max_bytes:
+                    raise ValueError("single image payload exceeds max_bytes")
+                chunks.append(ChunkPayload(segments=[], image_urls=[image]))
+            return chunks
+
+        chunk_count = len(chunks)
+        total_images = len(image_urls)
+        for idx, image in enumerate(image_urls):
+            preferred_idx = min(chunk_count - 1, (idx * chunk_count) // total_images)
+            placed = False
+
+            for chunk_idx in range(preferred_idx, len(chunks)):
+                chunk = chunks[chunk_idx]
+                candidate_images = chunk.image_urls + [image]
+                if self._messages_size(chunk.segments, candidate_images, **kwargs) <= self.max_bytes:
+                    chunk.image_urls = candidate_images
+                    placed = True
+                    break
+
+            if placed:
+                continue
+
+            if self._messages_size([], [image], **kwargs) > self.max_bytes:
+                raise ValueError("single image payload exceeds max_bytes")
+            chunks.append(ChunkPayload(segments=[], image_urls=[image]))
+
+        return chunks
+
+    def group_texts_by_budget(self, texts: List[str], build_messages: Callable, **kwargs) -> List[List[str]]:
+        groups: List[List[str]] = []
+        idx = 0
+        while idx < len(texts):
+            group: List[str] = []
+            while idx < len(texts):
+                candidate = group + [texts[idx]]
+                try:
+                    messages = build_messages(candidate, [], **kwargs)
+                except TypeError:
+                    messages = build_messages(candidate, **kwargs)
+                size = self.estimate(messages)
+                if size <= self.max_bytes:
+                    group = candidate
+                    idx += 1
+                    continue
+                if not group:
+                    raise ValueError("single text block exceeds max_bytes")
+                break
+            groups.append(group)
+        return groups
--- a/backend/app/gpt/universal_gpt.py
+++ b/backend/app/gpt/universal_gpt.py
@@ -1,8 +1,16 @@
 from app.gpt.base import GPT
 from app.gpt.prompt_builder import generate_base_prompt
 from app.models.gpt_model import GPTSource
-from app.gpt.prompt import BASE_PROMPT, AI_SUM, SCREENSHOT, LINK
+import os
+import hashlib
+import json
+import time
+from datetime import datetime, timezone
+from pathlib import Path
+
+from app.gpt.prompt import BASE_PROMPT, AI_SUM, SCREENSHOT, LINK, MERGE_PROMPT
 from app.gpt.utils import fix_markdown
+from app.gpt.request_chunker import RequestChunker
 from app.models.transcriber_model import TranscriptSegment
 from datetime import timedelta
 from typing import List
@@ -15,6 +23,12 @@ class UniversalGPT(GPT):
        self.temperature = temperature
        self.screenshot = False
        self.link = False
+        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:]
@@ -40,7 +54,7 @@ class UniversalGPT(GPT):
        )

        # ⛳ 组装 content 数组，支持 text + image_url 混合
-        content = [{"type": "text", "text": content_text}]
+        content: List[dict] = [{"type": "text", "text": content_text}]
        video_img_urls = kwargs.get('video_img_urls', [])

        for url in video_img_urls:
@@ -63,23 +77,231 @@ class UniversalGPT(GPT):
    def list_models(self):
        return self.client.models.list()

+    def _estimate_messages_bytes(self, messages: list) -> int:
+        import json
+        return len(json.dumps(messages, ensure_ascii=False).encode("utf-8"))
+
+    def _build_merge_messages(self, partials: list) -> list:
+        merge_text = MERGE_PROMPT + "\n\n" + "\n\n---\n\n".join(partials)
+        return [{
+            "role": "user",
+            "content": [{"type": "text", "text": merge_text}]
+        }]
+
+    def _checkpoint_path(self, checkpoint_key: str) -> Path:
+        safe_key = "".join(ch if ch.isalnum() or ch in ("-", "_") else "_" for ch in checkpoint_key)
+        return self.checkpoint_dir / f"{safe_key}.gpt.checkpoint.json"
+
+    def _build_source_signature(self, source: GPTSource) -> str:
+        payload = {
+            "model": self.model,
+            "temperature": self.temperature,
+            "max_request_bytes": self.max_request_bytes,
+            "title": source.title,
+            "tags": source.tags,
+            "format": source._format,
+            "style": source.style,
+            "extras": source.extras,
+            "video_img_urls": source.video_img_urls or [],
+            "segments": [
+                {
+                    "start": getattr(seg, "start", None),
+                    "end": getattr(seg, "end", None),
+                    "text": getattr(seg, "text", "")
+                }
+                for seg in source.segment
+            ],
+        }
+        raw = json.dumps(payload, ensure_ascii=False, sort_keys=True)
+        return hashlib.sha256(raw.encode("utf-8")).hexdigest()
+
+    def _load_checkpoint(self, checkpoint_key: str, source_signature: str) -> dict | None:
+        path = self._checkpoint_path(checkpoint_key)
+        if not path.exists():
+            return None
+        try:
+            data = json.loads(path.read_text(encoding="utf-8"))
+            if data.get("source_signature") != source_signature:
+                path.unlink(missing_ok=True)
+                return None
+            return data
+        except Exception:
+            path.unlink(missing_ok=True)
+            return None
+
+    def _save_checkpoint(self, checkpoint_key: str, source_signature: str, partials: list, phase: str) -> None:
+        path = self._checkpoint_path(checkpoint_key)
+        data = {
+            "version": 1,
+            "source_signature": source_signature,
+            "phase": phase,
+            "partials": partials,
+            "updated_at": datetime.now(timezone.utc).isoformat(),
+        }
+        tmp_path = path.with_suffix(".tmp")
+        tmp_path.write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")
+        tmp_path.replace(path)
+
+    def _clear_checkpoint(self, checkpoint_key: str) -> None:
+        self._checkpoint_path(checkpoint_key).unlink(missing_ok=True)
+
+    @staticmethod
+    def _is_insufficient_quota_error(exc: Exception) -> bool:
+        raw = str(exc)
+        return (
+            "insufficient_user_quota" in raw
+            or "预扣费额度失败" in raw
+            or "insufficient quota" in raw.lower()
+        )
+
+    @staticmethod
+    def _is_retryable_error(exc: Exception) -> bool:
+        raw = str(exc).lower()
+        retryable_tokens = (
+            "error code: 524",
+            "bad_response_status_code",
+            "timed out",
+            "timeout",
+            "rate limit",
+            "error code: 429",
+            "error code: 500",
+            "error code: 502",
+            "error code: 503",
+            "error code: 504",
+            "apiconnectionerror",
+            "connection error",
+            "service unavailable",
+        )
+        if any(token in raw for token in retryable_tokens):
+            return True
+
+        status = getattr(exc, "status_code", None) or getattr(exc, "status", None)
+        return status in {408, 409, 429, 500, 502, 503, 504, 524}
+
+    def _chat_completion_create(self, messages: list):
+        last_exc = None
+        for attempt in range(self._max_retry_attempts):
+            try:
+                return self.client.chat.completions.create(
+                    model=self.model,
+                    messages=messages,
+                    temperature=self.temperature
+                )
+            except Exception as exc:
+                last_exc = exc
+                if attempt == self._max_retry_attempts - 1 or not self._is_retryable_error(exc):
+                    raise
+                sleep_seconds = self._retry_base_backoff * (2 ** attempt)
+                time.sleep(sleep_seconds)
+
+        if last_exc is not None:
+            raise last_exc
+        raise RuntimeError("chat completion failed without exception")
+
+    def _merge_partials(self, partials: list, checkpoint_key: str | None, source_signature: str | None) -> str:
+        def build_messages(texts, *_args, **_kwargs):
+            return self._build_merge_messages(texts)
+
+        merge_chunker = RequestChunker(
+            lambda *_args, **_kwargs: [],
+            self.max_request_bytes,
+            self._estimate_messages_bytes
+        )
+
+        current_partials = list(partials)
+        while len(current_partials) > 1:
+            groups = merge_chunker.group_texts_by_budget(current_partials, build_messages)
+            new_partials = []
+            for group_idx, group in enumerate(groups):
+                messages = build_messages(group)
+                try:
+                    response = self._chat_completion_create(messages)
+                except Exception as exc:
+                    if checkpoint_key and source_signature:
+                        self._save_checkpoint(checkpoint_key, source_signature, current_partials, "merge")
+                    raise
+
+                new_partials.append(response.choices[0].message.content.strip())
+
+                if checkpoint_key and source_signature:
+                    remaining_partials = []
+                    for remaining_group in groups[group_idx + 1:]:
+                        remaining_partials.extend(remaining_group)
+                    resumable_partials = new_partials + remaining_partials
+                    self._save_checkpoint(checkpoint_key, source_signature, resumable_partials, "merge")
+
+            current_partials = new_partials
+
+        return current_partials[0]
+
    def summarize(self, source: GPTSource) -> str:
        self.screenshot = source.screenshot
        self.link = source.link
        source.segment = self.ensure_segments_type(source.segment)
+        checkpoint_key = source.checkpoint_key
+        source_signature = self._build_source_signature(source) if checkpoint_key else None

-        messages = self.create_messages(
-            source.segment,
-            title=source.title,
-            tags=source.tags,
-            video_img_urls=source.video_img_urls,
-            _format=source._format,
-            style=source.style,
-            extras=source.extras
-        )
-        response = self.client.chat.completions.create(
-            model=self.model,
-            messages=messages,
-            temperature=0.7
-        )
-        return response.choices[0].message.content.strip()
+        def message_builder(segments, image_urls, **kwargs):
+            return self.create_messages(segments, video_img_urls=image_urls, **kwargs)
+
+        chunker = RequestChunker(message_builder, self.max_request_bytes, self._estimate_messages_bytes)
+
+        try:
+            chunks = chunker.chunk(
+                source.segment,
+                source.video_img_urls or [],
+                title=source.title,
+                tags=source.tags,
+                _format=source._format,
+                style=source.style,
+                extras=source.extras
+            )
+        except ValueError:
+            chunks = chunker.chunk(
+                source.segment,
+                [],
+                title=source.title,
+                tags=source.tags,
+                _format=source._format,
+                style=source.style,
+                extras=source.extras
+            )
+
+        partials = []
+        if checkpoint_key and source_signature:
+            checkpoint = self._load_checkpoint(checkpoint_key, source_signature)
+            if checkpoint and isinstance(checkpoint.get("partials"), list):
+                partials = checkpoint["partials"]
+
+        if len(partials) > len(chunks):
+            partials = []
+
+        for chunk in chunks[len(partials):]:
+            messages = self.create_messages(
+                chunk.segments,
+                title=source.title,
+                tags=source.tags,
+                video_img_urls=chunk.image_urls,
+                _format=source._format,
+                style=source.style,
+                extras=source.extras
+            )
+            try:
+                response = self._chat_completion_create(messages)
+            except Exception as exc:
+                if checkpoint_key and source_signature:
+                    self._save_checkpoint(checkpoint_key, source_signature, partials, "summarize")
+                raise
+
+            partials.append(response.choices[0].message.content.strip())
+            if checkpoint_key and source_signature:
+                self._save_checkpoint(checkpoint_key, source_signature, partials, "summarize")
+
+        if len(partials) == 1:
+            if checkpoint_key:
+                self._clear_checkpoint(checkpoint_key)
+            return partials[0]
+        merged = self._merge_partials(partials, checkpoint_key, source_signature)
+        if checkpoint_key:
+            self._clear_checkpoint(checkpoint_key)
+        return merged
--- a/backend/app/models/gpt_model.py
+++ b/backend/app/models/gpt_model.py
@@ -15,4 +15,5 @@ class GPTSource:
    extras: Optional[str] = None
    _format: Optional[list] = None
    video_img_urls:  Optional[list] = None
+    checkpoint_key: Optional[str] = None

--- a/backend/app/routers/chat.py
+++ b/backend/app/routers/chat.py
@@ -0,0 +1,101 @@
+from fastapi import APIRouter, BackgroundTasks
+from pydantic import BaseModel
+
+from app.services.chat_service import chat as chat_service
+from app.services.vector_store import VectorStoreManager
+from app.utils.logger import get_logger
+from app.utils.response import ResponseWrapper as R
+
+logger = get_logger(__name__)
+
+router = APIRouter()
+
+# 索引状态追踪: task_id -> "indexing" | "indexed" | "failed"
+_index_status: dict[str, str] = {}
+
+
+class IndexRequest(BaseModel):
+    task_id: str
+
+
+class ChatMessage(BaseModel):
+    role: str
+    content: str
+
+
+class AskRequest(BaseModel):
+    task_id: str
+    question: str
+    history: list[ChatMessage] = []
+    provider_id: str
+    model_name: str
+
+
+def _do_index(task_id: str):
+    """后台执行索引任务。"""
+    try:
+        _index_status[task_id] = "indexing"
+        store = VectorStoreManager()
+        store.index_task(task_id)
+        _index_status[task_id] = "indexed"
+        logger.info(f"索引完成: {task_id}")
+    except Exception as e:
+        _index_status[task_id] = "failed"
+        logger.error(f"索引失败: {task_id}, {e}")
+
+
+@router.post("/chat/index")
+def index_task(data: IndexRequest, background_tasks: BackgroundTasks):
+    """触发后台索引，立即返回。"""
+    if _index_status.get(data.task_id) == "indexing":
+        return R.success(msg="正在索引中")
+
+    # 如果已经索引过，直接返回
+    store = VectorStoreManager()
+    if store.is_indexed(data.task_id):
+        _index_status[data.task_id] = "indexed"
+        return R.success(msg="已完成索引")
+
+    _index_status[data.task_id] = "indexing"
+    background_tasks.add_task(_do_index, data.task_id)
+    return R.success(msg="开始索引")
+
+
+@router.get("/chat/status")
+def chat_status(task_id: str):
+    """返回索引状态：idle / indexing / indexed / failed。"""
+    try:
+        # 优先检查内存状态
+        status = _index_status.get(task_id)
+        if status:
+            return R.success(data={"status": status, "indexed": status == "indexed"})
+
+        # 内存没有记录，检查持久化
+        store = VectorStoreManager()
+        indexed = store.is_indexed(task_id)
+        if indexed:
+            _index_status[task_id] = "indexed"
+        return R.success(data={"status": "indexed" if indexed else "idle", "indexed": indexed})
+    except Exception as e:
+        logger.error(f"查询索引状态失败: {e}")
+        return R.success(data={"status": "idle", "indexed": False})
+
+
+@router.post("/chat/ask")
+def ask_question(data: AskRequest):
+    """基于笔记内容的 RAG 问答。"""
+    try:
+        history = [{"role": m.role, "content": m.content} for m in data.history]
+        result = chat_service(
+            task_id=data.task_id,
+            question=data.question,
+            history=history,
+            provider_id=data.provider_id,
+            model_name=data.model_name,
+        )
+        return R.success(data=result)
+    except ValueError as e:
+        return R.error(msg=str(e))
+    except Exception as e:
+        logger.error(f"Chat 问答失败: {e}", exc_info=True)
+        return R.error(msg=f"问答失败: {str(e)}")
--- a/backend/app/routers/config.py
+++ b/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:
@@ -42,4 +211,38 @@ async def sys_health():

@router.get("/sys_check")
 async def sys_check():
-    return R.success()
+    return R.success()
+
+
+@router.get("/deploy_status")
+async def deploy_status():
+    """返回部署监控所需的所有状态信息"""
+    import torch
+    import os
+    
+    # CUDA 状态
+    cuda_available = torch.cuda.is_available()
+    cuda_info = {
+        "available": cuda_available,
+        "version": torch.version.cuda if cuda_available else None,
+        "gpu_name": torch.cuda.get_device_name(0) if cuda_available else None,
+    }
+    
+    # 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:
+        ensure_ffmpeg_or_raise()
+        ffmpeg_ok = True
+    except:
+        ffmpeg_ok = False
+    
+    return R.success(data={
+        "backend": {"status": "running", "port": int(os.getenv("BACKEND_PORT", 8483))},
+        "cuda": cuda_info,
+        "whisper": {"model_size": model_size, "transcriber_type": transcriber_type},
+        "ffmpeg": {"available": ffmpeg_ok},
+    })
--- a/backend/app/routers/note.py
+++ b/backend/app/routers/note.py
@@ -15,6 +15,7 @@ from app.enmus.exception import NoteErrorEnum
 from app.enmus.note_enums import DownloadQuality
 from app.exceptions.note import NoteError
 from app.services.note import NoteGenerator, logger
+from app.services.task_serial_executor import task_serial_executor
 from app.utils.response import ResponseWrapper as R
 from app.utils.url_parser import extract_video_id
 from app.validators.video_url_validator import is_supported_video_url
@@ -82,28 +83,38 @@ def run_note_task(task_id: str, video_url: str, platform: str, quality: Download
    if not model_name or not provider_id:
        raise HTTPException(status_code=400, detail="请选择模型和提供者")

-    note = NoteGenerator().generate(
-        video_url=video_url,
-        platform=platform,
-        quality=quality,
-        task_id=task_id,
-        model_name=model_name,
-        provider_id=provider_id,
-        link=link,
-        _format=_format,
-        style=style,
-        extras=extras,
-        screenshot=screenshot
-        , video_understanding=video_understanding,
-        video_interval=video_interval,
-        grid_size=grid_size
-    )
+    def _execute_note_task():
+        return NoteGenerator().generate(
+            video_url=video_url,
+            platform=platform,
+            quality=quality,
+            task_id=task_id,
+            model_name=model_name,
+            provider_id=provider_id,
+            link=link,
+            _format=_format,
+            style=style,
+            extras=extras,
+            screenshot=screenshot,
+            video_understanding=video_understanding,
+            video_interval=video_interval,
+            grid_size=grid_size,
+        )
+
+    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:
        logger.warning(f"任务 {task_id} 执行失败，跳过保存")
        return
    save_note_to_file(task_id, note)

+    # 自动建立向量索引（用于 AI 问答），失败不影响笔记生成
+    try:
+        from app.services.vector_store import VectorStoreManager
+        VectorStoreManager().index_task(task_id)
+    except Exception as e:
+        logger.warning(f"向量索引失败（不影响笔记）: {e}")


@router.post('/delete_task')
@@ -144,13 +155,14 @@ 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
-            # 更新之前的状态
-            NoteGenerator()._update_status(task_id, TaskStatus.PENDING)
            logger.info(f"重试模式，复用已有 task_id={task_id}")
        else:
            # 正常新建任务
            task_id = str(uuid.uuid4())

+        # 统一先写入 PENDING，表示已进入队列等待串行执行
+        NoteGenerator()._update_status(task_id, TaskStatus.PENDING)
+
        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)
--- a/backend/app/routers/provider.py
+++ b/backend/app/routers/provider.py
@@ -77,11 +77,14 @@ def update_provider(data: ProviderUpdateRequest):
        ):
            return R.error(msg='请至少填写一个参数')

-        provider_id =ProviderService.update_provider(
+        updated_provider =ProviderService.update_provider(
            id=data.id,
            data=dict(data)
        )
-        return R.success(msg='更新模型供应商成功',data={'id': provider_id})
+        if updated_provider:
+            return R.success(msg='更新模型供应商成功', data=updated_provider)
+        else:
+            return R.error(msg='更新模型供应商失败')
    except Exception as e:
        print(e)
        return R.error(msg=str(e))
--- a/backend/app/services/chat_service.py
+++ b/backend/app/services/chat_service.py
@@ -0,0 +1,158 @@
+import json
+from typing import Optional
+
+from app.gpt.gpt_factory import GPTFactory
+from app.models.model_config import ModelConfig
+from app.services.provider import ProviderService
+from app.services.vector_store import VectorStoreManager
+from app.services.chat_tools import TOOLS, execute_tool
+from app.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+SYSTEM_PROMPT = """你是一个视频笔记问答助手。你拥有以下能力：
+
+1. 系统已自动检索了一些相关内容作为初始参考（见下方）
+2. 你可以调用工具主动查询更多信息：
+   - lookup_transcript: 查询视频原始转录文本（支持按时间、关键词、位置筛选）
+   - get_video_info: 获取视频元信息（标题、作者、简介、标签等）
+   - get_note_content: 获取完整笔记内容
+
+--- 初始检索内容 ---
+{context}
+---
+
+回答要求：
+- 如果初始检索内容不足以回答问题，请主动调用工具获取更多信息
+- 回答关于视频具体原话、细节时，用 lookup_transcript 查询原文
+- 回答关于作者、标题等基本信息时，用 get_video_info 查询
+- 请用中文回答，保持简洁准确"""
+
+
+def _build_context(chunks: list[dict]) -> str:
+    """将检索到的片段拼接为上下文文本。"""
+    parts = []
+    for chunk in chunks:
+        meta = chunk.get("metadata", {})
+        source_type = meta.get("source_type", "unknown")
+        if source_type == "meta":
+            label = "[视频信息]"
+        elif source_type == "markdown":
+            label = f"[笔记 - {meta.get('section_title', '')}]"
+        else:
+            start = meta.get("start_time", 0)
+            end = meta.get("end_time", 0)
+            label = f"[转录 - {start:.0f}s~{end:.0f}s]"
+        parts.append(f"{label}\n{chunk['text']}")
+    return "\n\n".join(parts)
+
+
+def _build_sources(chunks: list[dict]) -> list[dict]:
+    """从检索片段中提取来源信息。"""
+    sources = []
+    for chunk in chunks:
+        meta = chunk.get("metadata", {})
+        source = {
+            "text": chunk["text"][:200],
+            "source_type": meta.get("source_type", "unknown"),
+        }
+        if meta.get("section_title"):
+            source["section_title"] = meta["section_title"]
+        if meta.get("start_time") is not None:
+            source["start_time"] = meta["start_time"]
+        if meta.get("end_time") is not None:
+            source["end_time"] = meta["end_time"]
+        sources.append(source)
+    return sources
+
+
+def chat(
+    task_id: str,
+    question: str,
+    history: list[dict],
+    provider_id: str,
+    model_name: str,
+) -> dict:
+    """
+    RAG + Tool Calling 问答。
+    1. 向量检索初始上下文
+    2. 调用 LLM（带 tools）
+    3. 如果 LLM 调用了工具，执行工具并将结果返回给 LLM
+    4. 循环直到 LLM 给出最终回答
+    """
+    vector_store = VectorStoreManager()
+
+    # 1. 检索初始上下文
+    chunks = vector_store.query(task_id, question, n_results=6)
+    context = _build_context(chunks) if chunks else "（未检索到相关内容，请使用工具查询）"
+    sources = _build_sources(chunks) if chunks else []
+
+    # 2. 构建消息
+    system_msg = SYSTEM_PROMPT.format(context=context)
+    messages = [{"role": "system", "content": system_msg}]
+
+    for msg in history[-20:]:
+        messages.append({"role": msg["role"], "content": msg["content"]})
+
+    messages.append({"role": "user", "content": question})
+
+    # 3. 获取 LLM client
+    provider = ProviderService.get_provider_by_id(provider_id)
+    if not provider:
+        raise ValueError(f"未找到模型供应商: {provider_id}")
+
+    config = ModelConfig(
+        api_key=provider["api_key"],
+        base_url=provider["base_url"],
+        model_name=model_name,
+        provider=provider["type"],
+        name=provider["name"],
+    )
+    gpt = GPTFactory.from_config(config)
+
+    logger.info(f"Chat: task_id={task_id}, model={model_name}")
+
+    # 4. Tool calling 循环（最多 3 轮）
+    max_rounds = 3
+    for round_i in range(max_rounds):
+        response = gpt.client.chat.completions.create(
+            model=gpt.model,
+            messages=messages,
+            tools=TOOLS,
+            temperature=0.7,
+        )
+
+        msg = response.choices[0].message
+
+        # 没有工具调用，直接返回
+        if not msg.tool_calls:
+            return {"answer": msg.content or "", "sources": sources}
+
+        # 处理工具调用
+        messages.append(msg)
+
+        for tool_call in msg.tool_calls:
+            fn_name = tool_call.function.name
+            try:
+                fn_args = json.loads(tool_call.function.arguments)
+            except json.JSONDecodeError:
+                fn_args = {}
+
+            logger.info(f"Tool call [{round_i+1}/{max_rounds}]: {fn_name}({fn_args})")
+
+            result = execute_tool(task_id, fn_name, fn_args)
+
+            messages.append({
+                "role": "tool",
+                "tool_call_id": tool_call.id,
+                "content": result,
+            })
+
+    # 超过最大轮次，做最后一次不带 tools 的调用
+    response = gpt.client.chat.completions.create(
+        model=gpt.model,
+        messages=messages,
+        temperature=0.7,
+    )
+
+    return {"answer": response.choices[0].message.content or "", "sources": sources}
--- a/backend/app/services/chat_tools.py
+++ b/backend/app/services/chat_tools.py
@@ -0,0 +1,184 @@
+"""
+Chat function calling 工具定义与执行。
+提供给 LLM 调用，用于主动查询视频原文、笔记、元信息。
+"""
+
+import json
+import os
+from typing import Optional
+
+from app.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+NOTE_OUTPUT_DIR = os.getenv("NOTE_OUTPUT_DIR", "note_results")
+
+
+def _load_note_data(task_id: str) -> Optional[dict]:
+    path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}.json")
+    if not os.path.exists(path):
+        return None
+    with open(path, "r", encoding="utf-8") as f:
+        return json.load(f)
+
+
+# ── 工具定义（OpenAI function calling 格式）──────────────────────
+
+TOOLS = [
+    {
+        "type": "function",
+        "function": {
+            "name": "lookup_transcript",
+            "description": "查询视频原始转录文本。可按时间范围筛选、按关键词搜索、或获取指定位置的内容。",
+            "parameters": {
+                "type": "object",
+                "properties": {
+                    "start_time": {
+                        "type": "number",
+                        "description": "起始时间（秒），例如 0 表示视频开头，60 表示第1分钟",
+                    },
+                    "end_time": {
+                        "type": "number",
+                        "description": "结束时间（秒），不传则到末尾",
+                    },
+                    "keyword": {
+                        "type": "string",
+                        "description": "搜索关键词，返回包含该关键词的转录片段",
+                    },
+                    "position": {
+                        "type": "string",
+                        "enum": ["start", "end"],
+                        "description": "快捷位置：start=视频开头前30句，end=视频结尾后30句",
+                    },
+                },
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_video_info",
+            "description": "获取视频的完整元信息，包括标题、作者、简介、标签、时长、播放量等。",
+            "parameters": {
+                "type": "object",
+                "properties": {},
+                "required": [],
+            },
+        },
+    },
+    {
+        "type": "function",
+        "function": {
+            "name": "get_note_content",
+            "description": "获取 AI 生成的完整笔记内容（Markdown 格式）。",
+            "parameters": {
+                "type": "object",
+                "properties": {},
+                "required": [],
+            },
+        },
+    },
+]
+
+
+# ── 工具执行 ──────────────────────────────────────────────────
+
+def execute_tool(task_id: str, tool_name: str, arguments: dict) -> str:
+    """执行工具调用，返回结果字符串。"""
+    data = _load_note_data(task_id)
+    if not data:
+        return json.dumps({"error": "笔记数据不存在"}, ensure_ascii=False)
+
+    if tool_name == "lookup_transcript":
+        return _lookup_transcript(data, arguments)
+    elif tool_name == "get_video_info":
+        return _get_video_info(data)
+    elif tool_name == "get_note_content":
+        return _get_note_content(data)
+    else:
+        return json.dumps({"error": f"未知工具: {tool_name}"}, ensure_ascii=False)
+
+
+def _lookup_transcript(data: dict, args: dict) -> str:
+    segments = data.get("transcript", {}).get("segments", [])
+    if not segments:
+        return json.dumps({"error": "没有转录数据"}, ensure_ascii=False)
+
+    position = args.get("position")
+    start_time = args.get("start_time")
+    end_time = args.get("end_time")
+    keyword = args.get("keyword", "").strip()
+
+    # 快捷位置
+    if position == "start":
+        filtered = segments[:30]
+    elif position == "end":
+        filtered = segments[-30:]
+    else:
+        filtered = segments
+
+    # 时间筛选
+    if start_time is not None:
+        filtered = [s for s in filtered if s.get("end", 0) >= start_time]
+    if end_time is not None:
+        filtered = [s for s in filtered if s.get("start", 0) <= end_time]
+
+    # 关键词筛选
+    if keyword:
+        filtered = [s for s in filtered if keyword.lower() in s.get("text", "").lower()]
+
+    # 限制返回量，避免 token 爆炸
+    if len(filtered) > 50:
+        filtered = filtered[:50]
+        truncated = True
+    else:
+        truncated = False
+
+    result = {
+        "total_segments": len(data.get("transcript", {}).get("segments", [])),
+        "returned": len(filtered),
+        "truncated": truncated,
+        "segments": [
+            {
+                "start": round(s.get("start", 0), 1),
+                "end": round(s.get("end", 0), 1),
+                "text": s.get("text", ""),
+            }
+            for s in filtered
+        ],
+    }
+    return json.dumps(result, ensure_ascii=False)
+
+
+def _get_video_info(data: dict) -> str:
+    am = data.get("audio_meta", {})
+    raw = am.get("raw_info", {}) or {}
+
+    info = {
+        "title": am.get("title") or raw.get("title", ""),
+        "uploader": raw.get("uploader", ""),
+        "description": raw.get("description", "")[:1000],
+        "tags": raw.get("tags", [])[:20] if isinstance(raw.get("tags"), list) else [],
+        "duration_seconds": am.get("duration", 0),
+        "platform": am.get("platform", ""),
+        "video_id": am.get("video_id", ""),
+        "url": raw.get("webpage_url", ""),
+        "view_count": raw.get("view_count"),
+        "like_count": raw.get("like_count"),
+        "comment_count": raw.get("comment_count"),
+    }
+    # 去除 None 值
+    info = {k: v for k, v in info.items() if v is not None and v != ""}
+    return json.dumps(info, ensure_ascii=False)
+
+
+def _get_note_content(data: dict) -> str:
+    md = data.get("markdown", "")
+    if isinstance(md, list):
+        # 多版本，取最新
+        md = md[-1].get("content", "") if md else ""
+    # 限制长度
+    if len(md) > 5000:
+        md = md[:5000] + "\n\n... (内容过长已截断)"
+    return json.dumps({"markdown": md}, ensure_ascii=False)
--- a/backend/app/services/note.py
+++ b/backend/app/services/note.py
@@ -1,7 +1,6 @@
 import json
 import logging
 import os
-import re
 from dataclasses import asdict
 from pathlib import Path
 from typing import List, Optional, Tuple, Union, Any
@@ -32,7 +31,8 @@ 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.utils.note_helper import replace_content_markers
+from app.utils.note_helper import replace_content_markers, prepend_source_link
+from app.utils.screenshot_marker import extract_screenshot_timestamps
 from app.utils.status_code import StatusCode
 from app.utils.video_helper import generate_screenshot
 from app.utils.video_reader import VideoReader
@@ -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=[]
@@ -131,8 +133,46 @@ class NoteGenerator:
            audio_cache_file = NOTE_OUTPUT_DIR / f"{task_id}_audio.json"
            transcript_cache_file = NOTE_OUTPUT_DIR / f"{task_id}_transcript.json"
            markdown_cache_file = NOTE_OUTPUT_DIR / f"{task_id}_markdown.md"
-            print(audio_cache_file)
-            # 1. 下载音频/视频
+            # 1. 获取字幕/转写：优先缓存 → 平台字幕 → 音频转写
+            transcript = None
+
+            # 尝试读取缓存
+            if transcript_cache_file.exists():
+                logger.info(f"检测到转写缓存 ({transcript_cache_file})，尝试读取")
+                try:
+                    data = json.loads(transcript_cache_file.read_text(encoding="utf-8"))
+                    segments = [TranscriptSegment(**seg) for seg in data.get("segments", [])]
+                    transcript = TranscriptResult(
+                        language=data.get("language"),
+                        full_text=data["full_text"],
+                        segments=segments,
+                    )
+                    logger.info(f"已从缓存加载转写结果，共 {len(segments)} 段")
+                except Exception as e:
+                    logger.warning(f"加载转写缓存失败: {e}")
+
+            # 缓存没有，尝试获取平台字幕
+            if transcript is None:
+                logger.info("尝试获取平台字幕（优先于音频下载）...")
+                try:
+                    transcript = downloader.download_subtitles(video_url)
+                    if transcript and transcript.segments:
+                        logger.info(f"成功获取平台字幕，共 {len(transcript.segments)} 段")
+                        transcript_cache_file.write_text(
+                            json.dumps(asdict(transcript), ensure_ascii=False, indent=2),
+                            encoding="utf-8",
+                        )
+                    else:
+                        transcript = None
+                        logger.info("平台无可用字幕，将下载音频后转写")
+                except Exception as e:
+                    logger.warning(f"获取平台字幕失败: {e}，将下载音频后转写")
+                    transcript = None
+
+            # 2. 下载音频/视频
+            # 有字幕时只提取元信息，不下载音视频文件（除非需要截图/视频理解）
+            has_transcript = transcript is not None
+            need_full_download = not has_transcript or screenshot or video_understanding
            audio_meta = self._download_media(
                downloader=downloader,
                video_url=video_url,
@@ -145,14 +185,19 @@ class NoteGenerator:
                video_understanding=video_understanding,
                video_interval=video_interval,
                grid_size=grid_size,
+                skip_download=not need_full_download,
            )

-            # 2. 转写文字
-            transcript = self._transcribe_audio(
-                audio_file=audio_meta.file_path,
-                transcript_cache_file=transcript_cache_file,
-                status_phase=TaskStatus.TRANSCRIBING,
-            )
+            # 3. 如果前面没拿到字幕，走转写流程
+            if transcript is None:
+                transcript = self._get_transcript(
+                    downloader=downloader,
+                    video_url=video_url,
+                    audio_file=audio_meta.file_path,
+                    transcript_cache_file=transcript_cache_file,
+                    status_phase=TaskStatus.TRANSCRIBING,
+                    task_id=task_id,
+                )

            # 3. GPT 总结
            markdown = self._summarize_text(
@@ -178,6 +223,8 @@ class NoteGenerator:
                    platform=platform,
                )

+            markdown = prepend_source_link(markdown, str(video_url))
+
            # 5. 保存记录到数据库
            self._update_status(task_id, TaskStatus.SAVING)
            self._save_metadata(video_id=audio_meta.video_id, platform=platform, task_id=task_id)
@@ -323,6 +370,7 @@ class NoteGenerator:
        video_understanding: bool,
        video_interval: int,
        grid_size: List[int],
+        skip_download: bool = False,
    ) -> AudioDownloadResult | None:
        """
        1. 检查音频缓存；若不存在，则根据需要下载音频或视频（若需截图/可视化）。
@@ -345,34 +393,6 @@ class NoteGenerator:
        task_id = audio_cache_file.stem.split("_")[0]
        self._update_status(task_id, status_phase)

-
-
-        # 判断是否需要下载视频
-        need_video = screenshot or video_understanding
-        if need_video:
-            try:
-                logger.info("开始下载视频")
-                video_path_str = downloader.download_video(video_url)
-                self.video_path = Path(video_path_str)
-                logger.info(f"视频下载完成：{self.video_path}")
-
-                # 若指定了 grid_size，则生成缩略图
-                if grid_size:
-                    self.video_img_urls=VideoReader(
-                        video_path=str(self.video_path),
-                        grid_size=tuple(grid_size),
-                        frame_interval=video_interval,
-                        unit_width=1280,
-                        unit_height=720,
-                        save_quality=90,
-                    ).run()
-                else:
-                    logger.info("未指定 grid_size，跳过缩略图生成")
-            except Exception as exc:
-                logger.error(f"视频下载失败：{exc}")
-
-                self._handle_exception(task_id, exc)
-                raise
        # 已有缓存，尝试加载
        if audio_cache_file.exists():
            logger.info(f"检测到音频缓存 ({audio_cache_file})，直接读取")
@@ -381,6 +401,56 @@ class NoteGenerator:
                return AudioDownloadResult(**data)
            except Exception as e:
                logger.warning(f"读取音频缓存失败，将重新下载：{e}")
+
+        # 有字幕且不需要截图/视频理解时，只提取元信息不下载文件
+        if skip_download:
+            logger.info("已有字幕，仅提取视频元信息（不下载音视频）")
+            try:
+                audio = downloader.download(
+                    video_url=video_url,
+                    quality=quality,
+                    output_dir=output_path,
+                    need_video=False,
+                    skip_download=True,
+                )
+                audio_cache_file.write_text(
+                    json.dumps(asdict(audio), ensure_ascii=False, indent=2),
+                    encoding="utf-8",
+                )
+                logger.info(f"元信息提取完成 ({audio_cache_file})")
+                return audio
+            except Exception as exc:
+                logger.warning(f"元信息提取失败，将尝试完整下载: {exc}")
+
+        # 判断是否需要下载视频
+        need_video = screenshot or video_understanding
+        if screenshot and not grid_size:
+            grid_size = [2, 2]
+
+        frame_interval = video_interval if video_interval and video_interval > 0 else 6
+        if need_video:
+            try:
+                logger.info("开始下载视频")
+                video_path_str = downloader.download_video(video_url)
+                self.video_path = Path(video_path_str)
+                logger.info(f"视频下载完成：{self.video_path}")
+
+                if grid_size:
+                    self.video_img_urls = VideoReader(
+                        video_path=str(self.video_path),
+                        grid_size=tuple(grid_size),
+                        frame_interval=frame_interval,
+                        unit_width=960,
+                        unit_height=540,
+                        save_quality=80,
+                    ).run()
+                else:
+                    logger.info("未指定 grid_size，跳过缩略图生成")
+            except Exception as exc:
+                logger.error(f"视频下载失败：{exc}")
+                self._handle_exception(task_id, exc)
+                raise
+
        # 下载音频
        try:
            logger.info("开始下载音频")
@@ -390,7 +460,6 @@ class NoteGenerator:
                output_dir=output_path,
                need_video=need_video,
            )
-            # 缓存 audio 元信息到本地 JSON
            audio_cache_file.write_text(json.dumps(asdict(audio), ensure_ascii=False, indent=2), encoding="utf-8")
            logger.info(f"音频下载并缓存成功 ({audio_cache_file})")
            return audio
@@ -400,6 +469,62 @@ class NoteGenerator:
            raise


+    def _get_transcript(
+        self,
+        downloader: Downloader,
+        video_url: str,
+        audio_file: str,
+        transcript_cache_file: Path,
+        status_phase: TaskStatus,
+        task_id: Optional[str] = None,
+    ) -> TranscriptResult | None:
+        """
+        优先获取平台字幕，没有则 fallback 到音频转写
+
+        :param downloader: 下载器实例
+        :param video_url: 视频链接
+        :param audio_file: 音频文件路径（用于 fallback 转写）
+        :param transcript_cache_file: 缓存文件路径
+        :param status_phase: 状态枚举
+        :param task_id: 任务 ID
+        :return: TranscriptResult 对象
+        """
+        self._update_status(task_id, status_phase)
+
+        # 已有缓存，直接返回
+        if transcript_cache_file.exists():
+            logger.info(f"检测到转写缓存 ({transcript_cache_file})，尝试读取")
+            try:
+                data = json.loads(transcript_cache_file.read_text(encoding="utf-8"))
+                segments = [TranscriptSegment(**seg) for seg in data.get("segments", [])]
+                return TranscriptResult(language=data.get("language"), full_text=data["full_text"], segments=segments)
+            except Exception as e:
+                logger.warning(f"加载转写缓存失败，将重新获取：{e}")
+
+        # 1. 先尝试获取平台字幕
+        logger.info("尝试获取平台字幕...")
+        try:
+            transcript = downloader.download_subtitles(video_url)
+            if transcript and transcript.segments:
+                logger.info(f"成功获取平台字幕，共 {len(transcript.segments)} 段")
+                # 缓存结果
+                transcript_cache_file.write_text(
+                    json.dumps(asdict(transcript), ensure_ascii=False, indent=2),
+                    encoding="utf-8"
+                )
+                return transcript
+            else:
+                logger.info("平台无可用字幕，将使用音频转写")
+        except Exception as e:
+            logger.warning(f"获取平台字幕失败: {e}，将使用音频转写")
+
+        # 2. Fallback 到音频转写
+        return self._transcribe_audio(
+            audio_file=audio_file,
+            transcript_cache_file=transcript_cache_file,
+            status_phase=status_phase,
+        )
+
    def _transcribe_audio(
        self,
        audio_file: str,
@@ -480,6 +605,7 @@ class NoteGenerator:
            _format=formats,
            style=style,
            extras=extras,
+            checkpoint_key=task_id,
        )

        try:
@@ -532,7 +658,7 @@ class NoteGenerator:
        :param video_path: 本地视频文件路径
        :return: 替换后的 Markdown 字符串
        """
-        matches: List[Tuple[str, int]] = self._extract_screenshot_timestamps(markdown)
+        matches: List[Tuple[str, int]] = extract_screenshot_timestamps(markdown)
        for idx, (marker, ts) in enumerate(matches):
            try:
                img_path = generate_screenshot(str(video_path), str(IMAGE_OUTPUT_DIR), ts, idx)
@@ -555,14 +681,7 @@ class NoteGenerator:
        :param markdown: 原始 Markdown 文本
        :return: 标记与对应时间戳秒数的列表
        """
-        pattern = r"(?:\*Screenshot-(\d{2}):(\d{2})|Screenshot-\[(\d{2}):(\d{2})\])"
-        results: List[Tuple[str, int]] = []
-        for match in re.finditer(pattern, markdown):
-            mm = match.group(1) or match.group(3)
-            ss = match.group(2) or match.group(4)
-            total_seconds = int(mm) * 60 + int(ss)
-            results.append((match.group(0), total_seconds))
-        return results
+        return extract_screenshot_timestamps(markdown)

    def _save_metadata(self, video_id: str, platform: str, task_id: str) -> None:
        """
@@ -576,4 +695,4 @@ class NoteGenerator:
            insert_video_task(video_id=video_id, platform=platform, task_id=task_id)
            logger.info(f"已保存任务记录到数据库 (video_id={video_id}, platform={platform}, task_id={task_id})")
        except Exception as e:
-            logger.error(f"保存任务记录失败：{e}")
+            logger.error(f"保存任务记录失败：{e}")
--- a/backend/app/services/provider.py
+++ b/backend/app/services/provider.py
@@ -123,7 +123,12 @@ class ProviderService:
            filtered_data = {k: v for k, v in data.items() if v is not None and k != 'id'}
            print('更新模型供应商',filtered_data)
            update_provider(id, **filtered_data)
-            return id
+            # 获取更新后的供应商信息
+            updated_provider = get_provider_by_id(id)
+            return {
+                'id': id,
+                'enabled': updated_provider.enabled,
+            }

        except Exception as e:
            print('更新模型供应商失败：',e)
--- a/backend/app/services/task_serial_executor.py
+++ b/backend/app/services/task_serial_executor.py
@@ -0,0 +1,23 @@
+import os
+from concurrent.futures import ThreadPoolExecutor, Future
+from typing import Any, Callable
+
+
+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:
+        future: Future = self._pool.submit(fn, *args, **kwargs)
+        return future.result()
+
+    def shutdown(self, wait: bool = True):
+        self._pool.shutdown(wait=wait)
+
+
+# 保持向后兼容的导出名
+SerialTaskExecutor = ConcurrentTaskExecutor
+task_serial_executor = ConcurrentTaskExecutor()
--- a/backend/app/services/transcriber_config_manager.py
+++ b/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"]
--- a/backend/app/services/vector_store.py
+++ b/backend/app/services/vector_store.py
@@ -0,0 +1,226 @@
+import json
+import os
+import re
+from typing import Optional
+
+import chromadb
+from chromadb.config import Settings
+
+from app.utils.logger import get_logger
+
+logger = get_logger(__name__)
+
+NOTE_OUTPUT_DIR = os.getenv("NOTE_OUTPUT_DIR", "note_results")
+VECTOR_DB_DIR = os.getenv("VECTOR_DB_DIR", "vector_db")
+
+
+def _chunk_markdown(markdown: str) -> list[dict]:
+    """按 H2/H3 标题拆分 markdown 为语义块。"""
+    sections = re.split(r'(?=^#{2,3}\s)', markdown, flags=re.MULTILINE)
+    chunks = []
+    for section in sections:
+        section = section.strip()
+        if not section or len(section) < 30:
+            continue
+        heading_match = re.match(r'^(#{2,3})\s+(.+)', section)
+        title = heading_match.group(2).strip() if heading_match else "intro"
+        chunks.append({
+            "text": section,
+            "metadata": {"source_type": "markdown", "section_title": title},
+        })
+    return chunks
+
+
+def _chunk_transcript(segments: list[dict], window_size: int = 15, overlap: int = 3) -> list[dict]:
+    """将转录 segments 按滑动窗口分组。"""
+    if not segments:
+        return []
+    chunks = []
+    step = max(window_size - overlap, 1)
+    for i in range(0, len(segments), step):
+        window = segments[i:i + window_size]
+        if not window:
+            break
+        text = "\n".join(
+            f"[{seg.get('start', 0):.0f}s] {seg.get('text', '')}" for seg in window
+        )
+        chunks.append({
+            "text": text,
+            "metadata": {
+                "source_type": "transcript",
+                "start_time": window[0].get("start", 0),
+                "end_time": window[-1].get("end", 0),
+            },
+        })
+    return chunks
+
+
+def _build_meta_chunk(audio_meta: dict) -> list[dict]:
+    """将视频元信息（标题、作者、描述、标签等）构建为可检索的 chunk。"""
+    if not audio_meta:
+        return []
+
+    raw = audio_meta.get("raw_info", {}) or {}
+    parts = []
+
+    title = audio_meta.get("title") or raw.get("title", "")
+    if title:
+        parts.append(f"视频标题：{title}")
+
+    uploader = raw.get("uploader", "")
+    if uploader:
+        parts.append(f"视频作者/UP主：{uploader}")
+
+    desc = raw.get("description", "")
+    if desc:
+        parts.append(f"视频简介：{desc[:500]}")
+
+    tags = raw.get("tags", [])
+    if tags and isinstance(tags, list):
+        parts.append(f"标签：{', '.join(str(t) for t in tags[:20])}")
+
+    duration = audio_meta.get("duration", 0)
+    if duration:
+        m, s = divmod(int(duration), 60)
+        parts.append(f"视频时长：{m}分{s}秒")
+
+    platform = audio_meta.get("platform", "")
+    if platform:
+        parts.append(f"平台：{platform}")
+
+    url = raw.get("webpage_url", "")
+    if url:
+        parts.append(f"链接：{url}")
+
+    if not parts:
+        return []
+
+    return [{
+        "text": "\n".join(parts),
+        "metadata": {"source_type": "meta"},
+    }]
+
+
+class VectorStoreManager:
+    """基于 ChromaDB 的笔记向量存储管理器。"""
+
+    def __init__(self):
+        os.makedirs(VECTOR_DB_DIR, exist_ok=True)
+        self._client = chromadb.PersistentClient(
+            path=VECTOR_DB_DIR,
+            settings=Settings(anonymized_telemetry=False),
+        )
+
+    def _collection_name(self, task_id: str) -> str:
+        """ChromaDB collection 名称：直接使用 task_id（UUID 格式合法）。"""
+        return task_id
+
+    def index_task(self, task_id: str) -> None:
+        """读取笔记结果并建立向量索引。"""
+        result_path = os.path.join(NOTE_OUTPUT_DIR, f"{task_id}.json")
+        if not os.path.exists(result_path):
+            logger.warning(f"笔记文件不存在，跳过索引: {result_path}")
+            return
+
+        with open(result_path, "r", encoding="utf-8") as f:
+            note_data = json.load(f)
+
+        markdown = note_data.get("markdown", "")
+        transcript = note_data.get("transcript", {})
+        segments = transcript.get("segments", [])
+
+        audio_meta = note_data.get("audio_meta", {})
+
+        meta_chunks = _build_meta_chunk(audio_meta)
+        md_chunks = _chunk_markdown(markdown)
+        tr_chunks = _chunk_transcript(segments)
+        all_chunks = meta_chunks + md_chunks + tr_chunks
+
+        if not all_chunks:
+            logger.warning(f"笔记内容为空，跳过索引: {task_id}")
+            return
+
+        col_name = self._collection_name(task_id)
+
+        # 删除旧 collection（幂等）
+        try:
+            self._client.delete_collection(col_name)
+        except Exception:
+            pass
+
+        collection = self._client.create_collection(
+            name=col_name,
+            metadata={"hnsw:space": "cosine"},
+        )
+
+        documents = [c["text"] for c in all_chunks]
+        metadatas = [c["metadata"] for c in all_chunks]
+        ids = [f"{task_id}_{i}" for i in range(len(all_chunks))]
+
+        collection.add(documents=documents, metadatas=metadatas, ids=ids)
+        logger.info(f"向量索引完成: task_id={task_id}, chunks={len(all_chunks)}")
+
+    def _parse_results(self, results: dict) -> list[dict]:
+        """将 ChromaDB query 结果转换为 chunk 列表。"""
+        chunks = []
+        if not results or not results.get("documents") or not results["documents"][0]:
+            return chunks
+        for i in range(len(results["documents"][0])):
+            chunks.append({
+                "text": results["documents"][0][i],
+                "metadata": results["metadatas"][0][i] if results["metadatas"] else {},
+                "distance": results["distances"][0][i] if results["distances"] else None,
+            })
+        return chunks
+
+    def query(self, task_id: str, query_text: str, n_results: int = 6) -> list[dict]:
+        """
+        按固定配额从各来源检索：meta 1 条、markdown 2 条、transcript 3 条，
+        确保三种来源都被召回。
+        """
+        col_name = self._collection_name(task_id)
+        try:
+            collection = self._client.get_collection(col_name)
+        except Exception:
+            logger.warning(f"Collection 不存在: {col_name}")
+            return []
+
+        all_chunks = []
+
+        # 每种来源的配额
+        quotas = {"meta": 1, "markdown": 2, "transcript": 3}
+
+        for source_type, quota in quotas.items():
+            try:
+                results = collection.query(
+                    query_texts=[query_text],
+                    n_results=quota,
+                    where={"source_type": source_type},
+                )
+                all_chunks.extend(self._parse_results(results))
+            except Exception:
+                pass
+
+        return all_chunks
+
+    def delete_index(self, task_id: str) -> None:
+        """删除指定任务的向量索引。"""
+        col_name = self._collection_name(task_id)
+        try:
+            self._client.delete_collection(col_name)
+            logger.info(f"已删除向量索引: {task_id}")
+        except Exception:
+            pass
+
+    def is_indexed(self, task_id: str) -> bool:
+        """检查指定任务是否已建立完整索引（含 meta 信息）。"""
+        col_name = self._collection_name(task_id)
+        try:
+            col = self._client.get_collection(col_name)
+            if col.count() == 0:
+                return False
+            # 检查是否包含 meta chunk，旧索引可能缺失
+            meta = col.get(where={"source_type": "meta"}, limit=1)
+            return len(meta["ids"]) > 0
+        except Exception:
+            return False
--- a/backend/app/transcriber/transcriber_provider.py
+++ b/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:
--- a/backend/app/transcriber/whisper.py
+++ b/backend/app/transcriber/whisper.py
@@ -64,7 +64,6 @@ class WhisperTranscriber(Transcriber):
            model_size_or_path=model_path,
            device=self.device,
            compute_type=self.compute_type,
-            cpu_threads=cpu_threads,
            download_root=model_dir
        )
    @staticmethod
--- a/backend/app/utils/note_helper.py
+++ b/backend/app/utils/note_helper.py
@@ -1,9 +1,35 @@
 import re


-import re
+def prepend_source_link(markdown: str | None, source_url: str) -> str | None:
+    """
+    在笔记开头添加来源链接；若首个非空行已包含来源链接，则更新该行并避免重复。
+    """
+    if markdown is None:
+        return None
+
+    source = (source_url or "").strip()
+    if not source:
+        return markdown
+
+    header = f"> 来源链接：{source}"
+    lines = markdown.splitlines()
+    first_non_empty_idx = None
+    for idx, line in enumerate(lines):
+        if line.strip():
+            first_non_empty_idx = idx
+            break
+
+    if first_non_empty_idx is not None:
+        first_line = lines[first_non_empty_idx].strip()
+        if first_line.startswith("> 来源链接：") or first_line.startswith("来源链接："):
+            lines[first_non_empty_idx] = header
+            return "\n".join(lines)
+
+    if markdown.strip():
+        return f"{header}\n\n{markdown}"
+    return header

-import re

 def replace_content_markers(markdown: str, video_id: str, platform: str = 'bilibili') -> str:
    """
@@ -12,17 +38,24 @@ def replace_content_markers(markdown: str, video_id: str, platform: str = 'bilib
    # 匹配三种形式：*Content-04:16*、Content-04:16、Content-[04:16]
    pattern = r"(?:\*?)Content-(?:\[(\d{2}):(\d{2})\]|(\d{2}):(\d{2}))"

+    safe_video_id = video_id
+
    def replacer(match):
        mm = match.group(1) or match.group(3)
        ss = match.group(2) or match.group(4)
        total_seconds = int(mm) * 60 + int(ss)

        if platform == 'bilibili':
-            url = f"https://www.bilibili.com/video/{video_id}?t={total_seconds}"
+            video_id = video_id.replace("_p", "?p=")
+            url = f"https://www.bilibili.com/video/{video_id}&t={total_seconds}"
+            parsed_video_id = safe_video_id.replace("_p", "?p=")
+            url = f"https://www.bilibili.com/video/{parsed_video_id}&t={total_seconds}"
        elif platform == 'youtube':
            url = f"https://www.youtube.com/watch?v={video_id}&t={total_seconds}s"
+            url = f"https://www.youtube.com/watch?v={safe_video_id}&t={total_seconds}s"
        elif platform == 'douyin':
            url = f"https://www.douyin.com/video/{video_id}"
+            url = f"https://www.douyin.com/video/{safe_video_id}"
            return f"[原片 @ {mm}:{ss}]({url})"
        else:
            return f"({mm}:{ss})"
@@ -30,3 +63,4 @@ def replace_content_markers(markdown: str, video_id: str, platform: str = 'bilib
        return f"[原片 @ {mm}:{ss}]({url})"

    return re.sub(pattern, replacer, markdown)
+
--- a/backend/app/utils/screenshot_marker.py
+++ b/backend/app/utils/screenshot_marker.py
@@ -0,0 +1,13 @@
+import re
+from typing import List, Tuple
+
+
+def extract_screenshot_timestamps(markdown: str) -> List[Tuple[str, int]]:
+    pattern = r"(\*?Screenshot-(?:\[(\d{2}):(\d{2})\]|(\d{2}):(\d{2})))"
+    results: List[Tuple[str, int]] = []
+    for match in re.finditer(pattern, markdown):
+        mm = match.group(2) or match.group(4)
+        ss = match.group(3) or match.group(5)
+        total_seconds = int(mm) * 60 + int(ss)
+        results.append((match.group(1), total_seconds))
+    return results
--- a/backend/app/utils/url_parser.py
+++ b/backend/app/utils/url_parser.py
@@ -1,5 +1,6 @@
 import re
 from typing import Optional
+import requests


 def extract_video_id(url: str, platform: str) -> Optional[str]:
@@ -11,6 +12,12 @@ def extract_video_id(url: str, platform: str) -> Optional[str]:
    :return: 提取到的视频 ID 或 None
    """
    if platform == "bilibili":
+        # 如果是短链接，则解析真实链接
+        if "b23.tv" in url:
+            resolved_url = resolve_bilibili_short_url(url)
+            if resolved_url:
+                url = resolved_url
+
        # 匹配 BV号（如 BV1vc411b7Wa）
        match = re.search(r"BV([0-9A-Za-z]+)", url)
        return f"BV{match.group(1)}" if match else None
@@ -26,3 +33,18 @@ def extract_video_id(url: str, platform: str) -> Optional[str]:
        return match.group(1) if match else None

    return None
+
+
+def resolve_bilibili_short_url(short_url: str) -> Optional[str]:
+    """
+    解析哔哩哔哩短链接以获取真实视频链接
+
+    :param short_url: Bilibili短链接（如"https://b23.tv/xxxxxx"）
+    :return: 真实的视频链接或None
+    """
+    try:
+        response = requests.head(short_url, allow_redirects=True)
+        return response.url
+    except requests.RequestException as e:
+        print(f"Error resolving short URL: {e}")
+        return None
--- a/backend/app/utils/video_reader.py
+++ b/backend/app/utils/video_reader.py
@@ -1,7 +1,9 @@
 import base64
+import hashlib
 import os
 import re
 import subprocess
+from concurrent.futures import ThreadPoolExecutor, as_completed
 import ffmpeg
 from PIL import Image, ImageDraw, ImageFont

@@ -14,6 +16,7 @@ class VideoReader:
                 video_path: str,
                 grid_size=(3, 3),
                 frame_interval=2,
+                 dedupe_enabled=True,
                 unit_width=960,
                 unit_height=540,
                 save_quality=90,
@@ -23,6 +26,7 @@ class VideoReader:
        self.video_path = video_path
        self.grid_size = grid_size
        self.frame_interval = frame_interval
+        self.dedupe_enabled = dedupe_enabled
        self.unit_width = unit_width
        self.unit_height = unit_height
        self.save_quality = save_quality
@@ -31,6 +35,14 @@ class VideoReader:
        print(f"视频路径：{video_path}",self.frame_dir,self.grid_dir)
        self.font_path = font_path

+    @staticmethod
+    def _calculate_file_md5(file_path: str) -> str:
+        hasher = hashlib.md5()
+        with open(file_path, "rb") as f:
+            for chunk in iter(lambda: f.read(8192), b""):
+                hasher.update(chunk)
+        return hasher.hexdigest()
+
    def format_time(self, seconds: float) -> str:
        mm = int(seconds // 60)
        ss = int(seconds % 60)
@@ -43,6 +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:
@@ -50,13 +74,30 @@ 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)
+                    if frame_hash == last_hash:
+                        os.remove(output_path)
+                        continue
+                    last_hash = frame_hash
+
                image_paths.append(output_path)
            return image_paths
        except Exception as e:
--- a/backend/app/validators/video_url_validator.py
+++ b/backend/app/validators/video_url_validator.py
@@ -1,5 +1,6 @@
 from pydantic import AnyUrl, validator, BaseModel, field_validator
 import re
+from urllib.parse import urlparse

 SUPPORTED_PLATFORMS = {
    "bilibili": r"(https?://)?(www\.)?bilibili\.com/video/[a-zA-Z0-9]+",
@@ -10,6 +11,12 @@ SUPPORTED_PLATFORMS = {


 def is_supported_video_url(url: str) -> bool:
+    parsed = urlparse(url)
+
+    # 检查是否为Bilibili的短链接
+    if parsed.netloc == "b23.tv":
+        return True
+
    for name, pattern in SUPPORTED_PLATFORMS.items():
        if pattern in ["douyin", "kuaishou"]:
            if pattern in url:
--- a/backend/build.sh
+++ b/backend/build.sh
@@ -25,6 +25,7 @@ cp .env.example backend/.env
 # 步骤 2: PyInstaller 打包，直接添加已存在的 .env 文件
 echo "开始 PyInstaller 打包..."
 pyinstaller \
+  -y \
  --name BiliNoteBackend \
  --paths backend \
  --distpath ./BillNote_frontend/src-tauri/bin \
--- a/backend/ffmpeg_helper.py
+++ b/backend/ffmpeg_helper.py
@@ -14,7 +14,17 @@ def check_ffmpeg_exists() -> bool:
    logger.info(f"FFMPEG_BIN_PATH: {ffmpeg_bin_path}")
    if ffmpeg_bin_path and os.path.isdir(ffmpeg_bin_path):
        os.environ["PATH"] = ffmpeg_bin_path + os.pathsep + os.environ.get("PATH", "")
-        logger.info(f"ffmpeg 未配置路径，尝试使用系统路径PATH: {os.environ.get('PATH')}")
+        logger.info(f"使用FFMPEG_BIN_PATH: {ffmpeg_bin_path}")
+    else:
+        # 遍历系统PATH寻找ffmpeg.exe
+        system_path = os.environ.get("PATH", "")
+        path_dirs = system_path.split(os.pathsep)
+        for path_dir in path_dirs:
+            ffmpeg_exe_path = os.path.join(path_dir, "ffmpeg.exe")
+            if os.path.isfile(ffmpeg_exe_path):
+                os.environ["PATH"] = path_dir + os.pathsep + system_path
+                logger.info(f"在系统PATH中找到ffmpeg: {path_dir}")
+                break
    try:
        subprocess.run(["ffmpeg", "-version"], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=True)
        logger.info("ffmpeg 已安装")
@@ -36,4 +46,4 @@ def ensure_ffmpeg_or_raise():
            "🪟 Windows 推荐：https://www.gyan.dev/ffmpeg/builds/\n"
            "💡 如果你已安装，请将其路径写入 `.env` 文件，例如：\n"
            "FFMPEG_BIN_PATH=/your/custom/ffmpeg/bin"
-        )
+        )
--- a/backend/main.py
+++ b/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")
--- a/backend/requirements.txt
+++ b/backend/requirements.txt
@@ -16,6 +16,7 @@ celery==5.5.1
 certifi==2025.1.31
 cffi==1.17.1
 charset-normalizer==3.4.1
+chromadb>=0.5.0
 click==8.1.8
 click-didyoumean==0.3.1
 click-plugins==1.1.1
@@ -123,5 +124,6 @@ weasyprint==65.1
 webencodings==0.5.1
 websockets==15.0.1
 yarl==1.19.0
+youtube-transcript-api>=1.0.0
 yt-dlp==2025.3.31
 zopfli==0.2.3.post1
--- a/backend/run.bat
+++ b/backend/run.bat
@@ -0,0 +1 @@
+python main.py
--- a/backend/tests/init.py
+++ b/backend/tests/init.py
--- a/backend/tests/test_note_helper.py
+++ b/backend/tests/test_note_helper.py
@@ -0,0 +1,35 @@
+import importlib.util
+import pathlib
+import unittest
+
+
+ROOT = pathlib.Path(__file__).resolve().parents[1]
+MODULE_PATH = ROOT / "app" / "utils" / "note_helper.py"
+spec = importlib.util.spec_from_file_location("note_helper", MODULE_PATH)
+if spec is None or spec.loader is None:
+    raise ImportError("note_helper module spec not found")
+note_helper = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(note_helper)
+
+
+class TestNoteHelper(unittest.TestCase):
+    def test_prepend_source_link_adds_header_at_top(self):
+        source_url = "https://www.bilibili.com/video/BV1xx411c7mD"
+        markdown = "## 标题\n\n内容"
+
+        result = note_helper.prepend_source_link(markdown, source_url)
+
+        self.assertTrue(result.startswith(f"> 来源链接：{source_url}\n\n"))
+        self.assertIn("## 标题", result)
+
+    def test_prepend_source_link_does_not_duplicate_when_header_exists(self):
+        source_url = "https://www.youtube.com/watch?v=abc123"
+        markdown = f"> 来源链接：{source_url}\n\n## 标题\n\n内容"
+
+        result = note_helper.prepend_source_link(markdown, source_url)
+
+        self.assertEqual(result, markdown)
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/backend/tests/test_request_chunker.py
+++ b/backend/tests/test_request_chunker.py
@@ -0,0 +1,97 @@
+import importlib.util
+import pathlib
+import unittest
+from dataclasses import dataclass
+
+ROOT = pathlib.Path(__file__).resolve().parents[1]
+MODULE_PATH = ROOT / "app" / "gpt" / "request_chunker.py"
+spec = importlib.util.spec_from_file_location("request_chunker", MODULE_PATH)
+if spec is None or spec.loader is None:
+    raise ImportError("request_chunker module spec not found")
+request_chunker = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(request_chunker)
+RequestChunker = request_chunker.RequestChunker
+
+
+@dataclass
+class DummySeg:
+    start: float
+    end: float
+    text: str
+
+
+def build_messages(segments, image_urls, **_):
+    content = [{"type": "text", "text": "".join(s.text for s in segments)}]
+    for url in image_urls:
+        content.append({"type": "image_url", "image_url": {"url": url, "detail": "auto"}})
+    return [{"role": "user", "content": content}]
+
+
+def size_estimator(messages):
+    size = 0
+    for part in messages[0]["content"]:
+        if part["type"] == "text":
+            size += len(part["text"])
+        else:
+            size += len(part["image_url"]["url"])
+    return size
+
+
+class TestRequestChunker(unittest.TestCase):
+    def test_chunk_segments_preserves_order_and_content(self):
+        segments = [
+            DummySeg(0, 1, "aaaa"),
+            DummySeg(1, 2, "bbbb"),
+            DummySeg(2, 3, "cccc"),
+        ]
+        chunker = RequestChunker(build_messages, max_bytes=8, size_estimator=size_estimator)
+        chunks = chunker.chunk(segments, [])
+        texts = ["".join(seg.text for seg in c.segments) for c in chunks]
+        self.assertEqual("".join(texts), "aaaabbbbcccc")
+        self.assertTrue(all(texts))
+
+    def test_chunk_images_distributed_across_batches(self):
+        segments = [DummySeg(0, 1, "aa")]
+        images = ["i" * 6, "j" * 6, "k" * 6]
+        chunker = RequestChunker(build_messages, max_bytes=10, size_estimator=size_estimator)
+        chunks = chunker.chunk(segments, images)
+        all_images = [img for c in chunks for img in c.image_urls]
+        self.assertEqual(all_images, images)
+
+    def test_chunk_images_are_not_front_loaded_when_multiple_segment_chunks(self):
+        segments = [
+            DummySeg(0, 1, "aaaaaa"),
+            DummySeg(1, 2, "bbbbbb"),
+            DummySeg(2, 3, "cccccc"),
+        ]
+        images = ["11111", "22222", "33333"]
+        chunker = RequestChunker(build_messages, max_bytes=12, size_estimator=size_estimator)
+        chunks = chunker.chunk(segments, images)
+
+        self.assertGreaterEqual(len(chunks), 3)
+        image_counts = [len(c.image_urls) for c in chunks]
+        self.assertGreater(image_counts[1], 0)
+        self.assertGreater(image_counts[2], 0)
+        all_images = [img for c in chunks for img in c.image_urls]
+        self.assertEqual(all_images, images)
+
+    def test_split_oversized_segment(self):
+        segments = [DummySeg(0, 1, "x" * 25)]
+        chunker = RequestChunker(build_messages, max_bytes=10, size_estimator=size_estimator)
+        chunks = chunker.chunk(segments, [])
+        combined = "".join(seg.text for c in chunks for seg in c.segments)
+        self.assertEqual(combined, "x" * 25)
+
+    def test_group_texts_by_budget(self):
+        chunker = RequestChunker(build_messages, max_bytes=10, size_estimator=size_estimator)
+
+        def build_text_messages(texts, *_args, **_kwargs):
+            content = [{"type": "text", "text": "".join(texts)}]
+            return [{"role": "user", "content": content}]
+
+        groups = chunker.group_texts_by_budget(["aaaaa", "bbbbb", "ccccc"], build_text_messages)
+        self.assertEqual(groups, [["aaaaa", "bbbbb"], ["ccccc"]])
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/backend/tests/test_screenshot_marker.py
+++ b/backend/tests/test_screenshot_marker.py
@@ -0,0 +1,35 @@
+import importlib.util
+import pathlib
+import unittest
+
+
+ROOT = pathlib.Path(__file__).resolve().parents[1]
+MODULE_PATH = ROOT / "app" / "utils" / "screenshot_marker.py"
+spec = importlib.util.spec_from_file_location("screenshot_marker", MODULE_PATH)
+if spec is None or spec.loader is None:
+    raise ImportError("screenshot_marker module spec not found")
+screenshot_marker = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(screenshot_marker)
+extract_screenshot_timestamps = screenshot_marker.extract_screenshot_timestamps
+
+
+class TestScreenshotMarker(unittest.TestCase):
+    def test_extract_accepts_star_bracket_format(self):
+        markdown = "A\n*Screenshot-[01:02]\nB"
+        matches = extract_screenshot_timestamps(markdown)
+        self.assertEqual(matches, [("*Screenshot-[01:02]", 62)])
+
+    def test_extract_accepts_legacy_formats(self):
+        markdown = "*Screenshot-03:04 and Screenshot-[05:06]"
+        matches = extract_screenshot_timestamps(markdown)
+        self.assertEqual(
+            matches,
+            [
+                ("*Screenshot-03:04", 184),
+                ("Screenshot-[05:06]", 306),
+            ],
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/backend/tests/test_task_serial_executor.py
+++ b/backend/tests/test_task_serial_executor.py
@@ -0,0 +1,42 @@
+import importlib.util
+import pathlib
+import threading
+import time
+import unittest
+
+
+ROOT = pathlib.Path(__file__).resolve().parents[1]
+MODULE_PATH = ROOT / "app" / "services" / "task_serial_executor.py"
+spec = importlib.util.spec_from_file_location("task_serial_executor", MODULE_PATH)
+if spec is None or spec.loader is None:
+    raise ImportError("task_serial_executor module spec not found")
+task_serial_executor = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(task_serial_executor)
+SerialTaskExecutor = task_serial_executor.SerialTaskExecutor
+
+
+class TestTaskSerialExecutor(unittest.TestCase):
+    def test_executor_runs_tasks_one_by_one(self):
+        executor = SerialTaskExecutor()
+        state_lock = threading.Lock()
+        state = {"active": 0, "peak_active": 0}
+
+        def critical_work():
+            with state_lock:
+                state["active"] += 1
+                state["peak_active"] = max(state["peak_active"], state["active"])
+            time.sleep(0.05)
+            with state_lock:
+                state["active"] -= 1
+
+        threads = [threading.Thread(target=lambda: executor.run(critical_work)) for _ in range(2)]
+        for t in threads:
+            t.start()
+        for t in threads:
+            t.join()
+
+        self.assertEqual(state["peak_active"], 1)
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/backend/tests/test_universal_gpt_checkpoint.py
+++ b/backend/tests/test_universal_gpt_checkpoint.py
@@ -0,0 +1,147 @@
+import importlib.util
+import json
+import os
+import pathlib
+import sys
+import tempfile
+import types
+import unittest
+from pathlib import Path
+
+
+def _install_stubs():
+    app_mod = types.ModuleType("app")
+    gpt_pkg = types.ModuleType("app.gpt")
+    models_pkg = types.ModuleType("app.models")
+
+    base_mod = types.ModuleType("app.gpt.base")
+
+    class _GPT:
+        pass
+
+    base_mod.GPT = _GPT
+
+    prompt_builder_mod = types.ModuleType("app.gpt.prompt_builder")
+
+    def _generate_base_prompt(**_kwargs):
+        return "prompt"
+
+    prompt_builder_mod.generate_base_prompt = _generate_base_prompt
+
+    prompt_mod = types.ModuleType("app.gpt.prompt")
+    prompt_mod.BASE_PROMPT = ""
+    prompt_mod.AI_SUM = ""
+    prompt_mod.SCREENSHOT = ""
+    prompt_mod.LINK = ""
+    prompt_mod.MERGE_PROMPT = "merge"
+
+    utils_mod = types.ModuleType("app.gpt.utils")
+
+    def _fix_markdown(text):
+        return text
+
+    utils_mod.fix_markdown = _fix_markdown
+
+    request_chunker_mod = types.ModuleType("app.gpt.request_chunker")
+
+    class _RequestChunker:
+        def __init__(self, *_args, **_kwargs):
+            pass
+
+        def group_texts_by_budget(self, texts, _builder, **_kwargs):
+            return [texts]
+
+    request_chunker_mod.RequestChunker = _RequestChunker
+
+    gpt_model_mod = types.ModuleType("app.models.gpt_model")
+
+    class _GPTSource:
+        pass
+
+    gpt_model_mod.GPTSource = _GPTSource
+
+    transcriber_model_mod = types.ModuleType("app.models.transcriber_model")
+
+    class _TranscriptSegment:
+        def __init__(self, **kwargs):
+            self.start = kwargs.get("start", 0)
+            self.end = kwargs.get("end", 0)
+            self.text = kwargs.get("text", "")
+
+    transcriber_model_mod.TranscriptSegment = _TranscriptSegment
+
+    sys.modules.setdefault("app", app_mod)
+    sys.modules.setdefault("app.gpt", gpt_pkg)
+    sys.modules.setdefault("app.models", models_pkg)
+    sys.modules["app.gpt.base"] = base_mod
+    sys.modules["app.gpt.prompt_builder"] = prompt_builder_mod
+    sys.modules["app.gpt.prompt"] = prompt_mod
+    sys.modules["app.gpt.utils"] = utils_mod
+    sys.modules["app.gpt.request_chunker"] = request_chunker_mod
+    sys.modules["app.models.gpt_model"] = gpt_model_mod
+    sys.modules["app.models.transcriber_model"] = transcriber_model_mod
+
+
+def _load_universal_gpt_class():
+    _install_stubs()
+    root = pathlib.Path(__file__).resolve().parents[1]
+    module_path = root / "app" / "gpt" / "universal_gpt.py"
+    spec = importlib.util.spec_from_file_location("universal_gpt", module_path)
+    if spec is None or spec.loader is None:
+        raise ImportError("universal_gpt module spec not found")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module.UniversalGPT
+
+
+UniversalGPT = _load_universal_gpt_class()
+
+
+class _FailingCompletions:
+    def create(self, **_kwargs):
+        raise Exception("Error code: 524 - bad_response_status_code")
+
+
+class _DummyChat:
+    def __init__(self):
+        self.completions = _FailingCompletions()
+
+
+class _DummyModels:
+    @staticmethod
+    def list():
+        return []
+
+
+class _DummyClient:
+    def __init__(self):
+        self.chat = _DummyChat()
+        self.models = _DummyModels()
+
+
+class TestUniversalGPTCheckpoint(unittest.TestCase):
+    def test_merge_524_error_persists_checkpoint(self):
+        original_attempts = os.environ.get("OPENAI_RETRY_ATTEMPTS")
+        os.environ["OPENAI_RETRY_ATTEMPTS"] = "1"
+        gpt = UniversalGPT(_DummyClient(), model="mock-model")
+        try:
+            with tempfile.TemporaryDirectory() as tmp_dir:
+                gpt.checkpoint_dir = Path(tmp_dir)
+
+                with self.assertRaises(Exception):
+                    gpt._merge_partials(["part-a", "part-b"], "task-1", "sig-1")
+
+                checkpoint_path = gpt._checkpoint_path("task-1")
+                self.assertTrue(checkpoint_path.exists())
+                payload = json.loads(checkpoint_path.read_text(encoding="utf-8"))
+                self.assertEqual(payload["phase"], "merge")
+                self.assertEqual(payload["partials"], ["part-a", "part-b"])
+        finally:
+            if original_attempts is None:
+                os.environ.pop("OPENAI_RETRY_ATTEMPTS", None)
+            else:
+                os.environ["OPENAI_RETRY_ATTEMPTS"] = original_attempts
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/backend/tests/test_video_reader_dedupe.py
+++ b/backend/tests/test_video_reader_dedupe.py
@@ -0,0 +1,142 @@
+import importlib.util
+import pathlib
+import re
+import sys
+import tempfile
+import types
+import unittest
+from unittest.mock import patch
+
+
+def _install_stubs():
+    app_mod = types.ModuleType("app")
+    utils_pkg = types.ModuleType("app.utils")
+
+    logger_mod = types.ModuleType("app.utils.logger")
+
+    class _Logger:
+        @staticmethod
+        def info(*_args, **_kwargs):
+            return None
+
+        @staticmethod
+        def warning(*_args, **_kwargs):
+            return None
+
+        @staticmethod
+        def error(*_args, **_kwargs):
+            return None
+
+    def _get_logger(_name):
+        return _Logger()
+
+    logger_mod.get_logger = _get_logger
+
+    path_helper_mod = types.ModuleType("app.utils.path_helper")
+    ffmpeg_mod = types.ModuleType("ffmpeg")
+
+    pil_mod = types.ModuleType("PIL")
+    pil_image_mod = types.ModuleType("PIL.Image")
+    pil_draw_mod = types.ModuleType("PIL.ImageDraw")
+    pil_font_mod = types.ModuleType("PIL.ImageFont")
+
+    class _FakeImage:
+        pass
+
+    class _FakeImageDraw:
+        @staticmethod
+        def Draw(*_args, **_kwargs):
+            return None
+
+    class _FakeImageFont:
+        @staticmethod
+        def truetype(*_args, **_kwargs):
+            return None
+
+        @staticmethod
+        def load_default():
+            return None
+
+    pil_image_mod.Image = _FakeImage
+    pil_draw_mod.ImageDraw = _FakeImageDraw
+    pil_font_mod.ImageFont = _FakeImageFont
+
+    def _get_app_dir(name):
+        return name
+
+    path_helper_mod.get_app_dir = _get_app_dir
+    ffmpeg_mod.probe = lambda *_args, **_kwargs: {"format": {"duration": "0"}}
+
+    sys.modules.setdefault("app", app_mod)
+    sys.modules.setdefault("app.utils", utils_pkg)
+    sys.modules["PIL"] = pil_mod
+    sys.modules["PIL.Image"] = pil_image_mod
+    sys.modules["PIL.ImageDraw"] = pil_draw_mod
+    sys.modules["PIL.ImageFont"] = pil_font_mod
+    sys.modules["ffmpeg"] = ffmpeg_mod
+    sys.modules["app.utils.logger"] = logger_mod
+    sys.modules["app.utils.path_helper"] = path_helper_mod
+
+
+def _load_video_reader_module():
+    _install_stubs()
+    root = pathlib.Path(__file__).resolve().parents[1]
+    module_path = root / "app" / "utils" / "video_reader.py"
+    spec = importlib.util.spec_from_file_location("video_reader", module_path)
+    if spec is None or spec.loader is None:
+        raise ImportError("video_reader module spec not found")
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+video_reader_module = _load_video_reader_module()
+VideoReader = video_reader_module.VideoReader
+
+
+def _make_fake_ffmpeg_runner(colors_by_second):
+    def _runner(cmd, check=True):
+        output_path = next((arg for arg in cmd if isinstance(arg, str) and arg.endswith(".jpg")), None)
+        if output_path is None:
+            raise AssertionError("Output path not found in ffmpeg cmd")
+        match = re.search(r"frame_(\d{2})_(\d{2})\.jpg$", output_path)
+        if match is None:
+            raise AssertionError("Unexpected output path")
+        sec = int(match.group(1)) * 60 + int(match.group(2))
+        payload = colors_by_second[sec]
+        with open(output_path, "wb") as f:
+            f.write(payload)
+        return 0
+
+    return _runner
+
+
+class TestVideoReaderDeduplicateFrames(unittest.TestCase):
+    def test_extract_frames_skips_adjacent_duplicates_when_enabled(self):
+        with tempfile.TemporaryDirectory() as tmp_dir:
+            frame_dir = pathlib.Path(tmp_dir) / "frames"
+            grid_dir = pathlib.Path(tmp_dir) / "grids"
+            reader = VideoReader(
+                video_path="dummy.mp4",
+                frame_interval=1,
+                frame_dir=str(frame_dir),
+                grid_dir=str(grid_dir),
+            )
+
+            fake_colors = {
+                0: b"frame-a",
+                1: b"frame-a",
+                2: b"frame-b",
+                3: b"frame-b",
+            }
+
+            with patch.object(video_reader_module.ffmpeg, "probe", return_value={"format": {"duration": "4"}}), \
+                    patch.object(video_reader_module.subprocess, "run", side_effect=_make_fake_ffmpeg_runner(fake_colors)):
+                paths = reader.extract_frames(max_frames=10)
+
+            names = [pathlib.Path(p).name for p in paths]
+            self.assertEqual(names, ["frame_00_00.jpg", "frame_00_02.jpg"])
+
+
+if __name__ == "__main__":
+    unittest.main()
--- a/docker-compose.gpu.yml
+++ b/docker-compose.gpu.yml
@@ -0,0 +1,47 @@
+
+services:
+  backend:
+    container_name: bilinote-backend
+    build:
+      context: .
+      dockerfile: backend/Dockerfile.gpu
+      args:
+        APT_MIRROR: ${APT_MIRROR:-mirrors.tuna.tsinghua.edu.cn}
+        PIP_INDEX: ${PIP_INDEX:-https://pypi.tuna.tsinghua.edu.cn/simple}
+    env_file:
+      - .env
+    environment:
+      - BACKEND_PORT=${BACKEND_PORT}
+      - BACKEND_HOST=${BACKEND_HOST}
+    volumes:
+      - ./backend:/app
+    expose:
+      - "${BACKEND_PORT}"  # 不再对外暴露，用于 nginx 内部通信
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: "nvidia"
+              count: "all"
+              capabilities: ["gpu"]
+
+  frontend:
+    container_name: bilinote-frontend
+    build:
+      context: .
+      dockerfile: BillNote_frontend/Dockerfile
+    env_file:
+      - .env
+    expose:
+      - "80"  # 不暴露给宿主机，只供 nginx 访问
+
+  nginx:
+    container_name: bilinote-nginx
+    image: nginx:1.25-alpine
+    ports:
+      - "${APP_PORT}:80"
+    volumes:
+      - ./nginx/default.conf:/etc/nginx/conf.d/default.conf
+    depends_on:
+      - backend
+      - frontend
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -5,6 +5,9 @@ services:
    build:
      context: .
      dockerfile: backend/Dockerfile
+      args:
+        APT_MIRROR: ${APT_MIRROR:-mirrors.tuna.tsinghua.edu.cn}
+        PIP_INDEX: ${PIP_INDEX:-https://pypi.tuna.tsinghua.edu.cn/simple}
    env_file:
      - .env
    environment:
@@ -14,6 +17,14 @@ services:
      - ./backend:/app
    expose:
      - "${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 +35,8 @@ 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 +46,9 @@ 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
--- a/nginx/default.conf
+++ b/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;
@@ -8,15 +16,16 @@ server {

  # 所有 /api 请求代理给 backend 容器
  location /api/ {
-    proxy_pass http://backend:8000;
+    proxy_pass http://backend:8483;
    proxy_set_header Host $host;
    proxy_set_header X-Real-IP $remote_addr;
  }
-    
-   location /static/ {
-    proxy_pass http://backend:8000/static/;
+
+  location /static/ {
+    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";
  }
 }
--- a/run.bat
+++ b/run.bat
@@ -0,0 +1,7 @@
+@echo off
+cd /d "%~dp0"
+
+start "Backend" powershell -NoExit -Command "cd backend; conda activate bili; python main.py"
+start "Frontend" powershell -NoExit -Command "cd BillNote_frontend; npm run dev"
+
+start http://localhost:3015/