feat(bilibili): B站合集分P + HashRouter + 公式渲染 + prompt LaTeX 说明 + ctranslate2 兼容

BillNote_frontend/src/App.tsx

@@ -1,6 +1,6 @@
 import './App.css'
 import { lazy, Suspense, useEffect } from 'react'
-import { BrowserRouter, Navigate, Routes, Route } from 'react-router-dom'
+import { HashRouter, Navigate, Routes, Route } from 'react-router-dom'
 import { useTaskPolling } from '@/hooks/useTaskPolling.ts'
 import { useCheckBackend } from '@/hooks/useCheckBackend.ts'
 import { systemCheck } from '@/services/system.ts'
@@ -62,7 +62,7 @@ function App() {
     <>
       <StartupBanner />
       <BackendHealthIndicator />
-      <BrowserRouter>
+      <HashRouter>
         <Suspense fallback={<div className="flex h-screen items-center justify-center">加载中…</div>}>
           <Routes>
             <Route path="/onboarding" element={<Onboarding />} />
@@ -86,7 +86,7 @@ function App() {
             </Route>
           </Routes>
         </Suspense>
-      </BrowserRouter>
+      </HashRouter>
     </>
   )
 }

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

@@ -46,8 +46,8 @@ const steps = [
   { label: '保存完成', key: 'SUCCESS' },
 ]
 
-const remarkPlugins = [gfm, remarkMath]
-const rehypePlugins = [rehypeKatex]
+const remarkPlugins = [gfm, [remarkMath, { singleDollarTextMath: true }]]
+const rehypePlugins = [[rehypeKatex, { throwOnError: false, strict: false }]]
 
 /**
  * 构建 ReactMarkdown components 对象，baseURL 用于修正图片路径。

backend/app/downloaders/bilibili_downloader.py

@@ -2,6 +2,7 @@
 import json
 import logging
 import tempfile
+import re
 from abc import ABC
 from typing import Union, Optional, List
 
@@ -12,7 +13,7 @@
 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.utils.url_parser import extract_video_id, extract_bilibili_task_id
 from app.services.cookie_manager import CookieConfigManager
 
 logger = logging.getLogger(__name__)
@@ -54,7 +55,11 @@ def download(
             output_dir=self.cache_data
         os.makedirs(output_dir, exist_ok=True)
 
-        output_path = os.path.join(output_dir, "%(id)s.%(ext)s")
+        # 合集分 P 处理：识别 URL 中的 p 参数，避免总是下载第一集
+        p_match = re.search(r'[?&]p=(\d+)', video_url)
+        p_suffix = f"_p{p_match.group(1)}" if p_match else ""
+
+        output_path = os.path.join(output_dir, f"%(id)s{p_suffix}.%(ext)s")
 
         ydl_opts = {
             'format': 'bestaudio[ext=m4a]/bestaudio/best',
@@ -67,9 +72,14 @@ def download(
                     'preferredquality': '64',
                 }
             ],
-            'noplaylist': True,
             'quiet': False,
         }
+
+        if p_match:
+            ydl_opts['playlist_items'] = p_match.group(1)
+        else:
+            ydl_opts['noplaylist'] = True
+
         if self._cookiefile:
             ydl_opts['cookiefile'] = self._cookiefile
 
@@ -79,7 +89,7 @@ def download(
             title = info.get("title")
             duration = info.get("duration", 0)
             cover_url = info.get("thumbnail")
-            audio_path = os.path.join(output_dir, f"{video_id}.mp3")
+            audio_path = os.path.join(output_dir, f"{video_id}{p_suffix}.mp3")
 
         return AudioDownloadResult(
             file_path=audio_path,
@@ -106,30 +116,37 @@ def download_video(
         os.makedirs(output_dir, exist_ok=True)
         print("video_url",video_url)
         video_id=extract_video_id(video_url, "bilibili")
-        video_path = os.path.join(output_dir, f"{video_id}.mp4")
+        p_match = re.search(r'[?&]p=(\d+)', video_url)
+        p_suffix = f"_p{p_match.group(1)}" if p_match else ""
+        video_path = os.path.join(output_dir, f"{video_id}{p_suffix}.mp4")
         if os.path.exists(video_path):
             return video_path
 
         # 检查是否已经存在
 
 
-        output_path = os.path.join(output_dir, "%(id)s.%(ext)s")
+        output_path = os.path.join(output_dir, f"%(id)s{p_suffix}.%(ext)s")
 
         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 p_match:
+            ydl_opts['playlist_items'] = p_match.group(1)
+        else:
+            ydl_opts['noplaylist'] = True
+
         if self._cookiefile:
             ydl_opts['cookiefile'] = self._cookiefile
 
         with yt_dlp.YoutubeDL(ydl_opts) as ydl:
             info = ydl.extract_info(video_url, download=True)
             video_id = info.get("id")
-            video_path = os.path.join(output_dir, f"{video_id}.mp4")
+            video_path = os.path.join(output_dir, f"{video_id}{p_suffix}.mp4")
 
         if not os.path.exists(video_path):
             raise FileNotFoundError(f"视频文件未找到: {video_path}")
@@ -175,17 +192,24 @@ def download_subtitles(self, video_url: str, output_dir: str = None,
             langs = ['zh-Hans', 'zh', 'zh-CN', 'ai-zh', 'en', 'en-US']
 
         video_id = extract_video_id(video_url, "bilibili")
+        task_id = extract_bilibili_task_id(video_url)
 
         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'),
+            'outtmpl': os.path.join(output_dir, f'{task_id}.%(ext)s'),
             'quiet': True,
         }
 
+        p_match = re.search(r'[?&]p=(\d+)', video_url)
+        if p_match:
+            ydl_opts['playlist_items'] = p_match.group(1)
+        else:
+            ydl_opts['noplaylist'] = True
+
         # 通过 CookieConfigManager 注入 B站 Cookie（Netscape cookiefile）
         if self._cookiefile:
             ydl_opts['cookiefile'] = self._cookiefile
@@ -229,7 +253,7 @@ def download_subtitles(self, video_url: str, output_dir: str = None,
 
                 # 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}")
+                subtitle_file = os.path.join(output_dir, f"{task_id}.{detected_lang}.{ext}")
 
                 if not os.path.exists(subtitle_file):
                     logger.info(f"字幕文件不存在: {subtitle_file}")

backend/app/gpt/prompt.py

@@ -38,7 +38,20 @@
 2. **去除无关内容**：省略广告、填充词、问候语和不相关的言论。
 3. **保留关键细节**：保留重要事实、示例、结论和建议。(如果额外重要的任务有格式需求可以不遵守)
 4. **可读布局**：必要时使用项目符号，并保持段落简短，增强可读性。(如果额外重要的任务有格式需求可以不遵守)
-5. 视频中提及的数学公式必须保留，并以 LaTeX 语法形式呈现，适合 Markdown 渲染。
+5. 视频中提及的数学公式必须保留，并以 LaTeX 语法形式呈现：
+   - 行内公式用 `$...$`，$ 与公式内容之间**不能有空格**
+   - 块级公式用 `$$...$$`，**必须独占一行**，前后各空一行，$$ 与公式内容之间**不能有空格**
+   - 例如：
+
+     原问题为：
+
+     $$\begin{{aligned}}
+     \min \quad & f = 3x_1 - 5x_2
+     \end{{aligned}}$$
+
+     转换为：$z = -f$
+   - **禁止**用代码块（如 ` ```math ` 或 ` ``` `）包裹公式。
+6. **禁止元叙述**：笔记中不要出现"视频讲了……"、"视频中提到……"、"在本视频中……"等指向视频本身的表述。直接总结内容即可，把信息当作客观事实陈述。
 
 
 请始终遵循此规则。

backend/app/routers/config.py

@@ -172,18 +172,23 @@ def get_transcriber_models_status():
     mlx_available = platform.system() == "Darwin"
     mlx_statuses = []
     if mlx_available:
-        from app.transcriber.mlx_whisper_transcriber import MLX_MODEL_MAP
-        for size in WHISPER_MODEL_SIZES:
-            mlx_key = f"mlx-{size}"
-            repo_id = MLX_MODEL_MAP.get(size)
-            # 用 config.json 判定，和 _check_mlx_whisper_model_exists / 加载逻辑保持一致
-            downloaded = _check_mlx_whisper_model_exists(size)
-            mlx_statuses.append({
-                "model_size": size,
-                "downloaded": downloaded,
-                "downloading": _downloading.get(mlx_key) == "downloading",
-                "available": repo_id is not None,
-            })
+        try:
+            from app.transcriber.mlx_whisper_transcriber import MLX_MODEL_MAP
+            for size in WHISPER_MODEL_SIZES:
+                mlx_key = f"mlx-{size}"
+                repo_id = MLX_MODEL_MAP.get(size)
+                # 用 config.json 判定，和 _check_mlx_whisper_model_exists / 加载逻辑保持一致
+                downloaded = _check_mlx_whisper_model_exists(size)
+                mlx_statuses.append({
+                    "model_size": size,
+                    "downloaded": downloaded,
+                    "downloading": _downloading.get(mlx_key) == "downloading",
+                    "available": repo_id is not None,
+                })
+        except ImportError:
+            # mlx-whisper 未安装或导入失败，跳过该分支，并反映实际不可用状态
+            mlx_available = False
+            logger.debug("MLX Whisper 未安装，跳过模型状态检查")
 
     return R.success(data={
         "whisper": statuses,

backend/app/utils/url_parser.py

@@ -20,7 +20,9 @@ def extract_video_id(url: str, platform: str) -> Optional[str]:
 
         # 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
+        if not match:
+            return None
+        return f"BV{match.group(1)}"
 
     elif platform == "youtube":
         # 匹配 v=xxxxx 或 youtu.be/xxxxx，ID 长度通常为 11
@@ -35,6 +37,22 @@ def extract_video_id(url: str, platform: str) -> Optional[str]:
     return None
 
 
+def extract_bilibili_task_id(url: str) -> str:
+    """
+    从 B 站链接中提取带分 P 信息的任务签名，用于缓存文件名区分不同 P。
+
+    :param url: Bilibili 视频链接
+    :return: 纯 BV 号，若存在 p 参数则追加为 BVxxxx_p2
+    """
+    bvid = extract_video_id(url, "bilibili")
+    if not bvid:
+        return ""
+    p_match = re.search(r'[?&]p=(\d+)', url)
+    if p_match:
+        return f"{bvid}_p{p_match.group(1)}"
+    return bvid
+
+
 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

backend/main.py

@@ -20,7 +20,9 @@
 from ffmpeg_helper import ensure_ffmpeg_or_raise
 
 logger = get_logger(__name__)
-load_dotenv()
+# 支持在 backend/ 子目录中运行时也能加载项目根目录的 .env
+dotenv_path = os.path.join(os.path.dirname(os.path.dirname(__file__)), '.env')
+load_dotenv(dotenv_path)
 
 # 读取 .env 中的路径
 static_path = os.getenv('STATIC', '/static')

backend/requirements.txt

@@ -24,7 +24,7 @@ click-repl==0.3.0
 colorama==0.4.6
 coloredlogs==15.0.1
 cssselect2==0.8.0
-ctranslate2==4.6.0
+ctranslate2>=4.6.0,<5
 distro==1.9.0
 dnspython==2.7.0
 email_validator==2.2.0