在应用程序中添加会话级别的临时文件管理功能，优化临时文件清理逻辑，并引入系统资源检查以提高稳定性和用户体验。同时更新依赖项以支持新功能。

requirements.txt

@@ -9,6 +9,7 @@ feedparser>=6.0.11
 requests>=2.32.3
 gradio>=5.30.0
 spaces>=0.36.0
+psutil>=5.9.0
 
 # 可选依赖 - whisper.cpp 绑定
 pywhispercpp>=1.3.0

src/podcast_transcribe/webui/app.py

@@ -8,6 +8,9 @@ import os
 import uuid
 import atexit
 import shutil
+import threading
+import time
+from typing import Dict, Set
 
 # 尝试相对导入，这在通过 `python -m src.podcast_transcribe.webui.app` 运行时有效
 try:
@@ -33,28 +36,129 @@ except ImportError:
     from podcast_transcribe.schemas import PodcastChannel, PodcastEpisode, CombinedTranscriptionResult, EnhancedSegment
     from podcast_transcribe.transcriber import transcribe_podcast_audio
 
-# 用于存储应用程序使用的所有临时文件路径
-temp_files = []
+# 会话级别的临时文件管理
+class SessionFileManager:
+    def __init__(self):
+        self.session_files: Dict[str, Set[str]] = {}
+        self.lock = threading.Lock()
+        self.cleanup_thread = None
+        self.start_cleanup_thread()
+    
+    def start_cleanup_thread(self):
+        """启动后台清理线程"""
+        if self.cleanup_thread is None or not self.cleanup_thread.is_alive():
+            self.cleanup_thread = threading.Thread(target=self._periodic_cleanup, daemon=True)
+            self.cleanup_thread.start()
+    
+    def _periodic_cleanup(self):
+        """定期清理过期的临时文件"""
+        while True:
+            try:
+                time.sleep(300)  # 每5分钟清理一次
+                self._cleanup_old_files()
+            except Exception as e:
+                print(f"清理线程错误: {e}")
+    
+    def _cleanup_old_files(self):
+        """清理超过30分钟的临时文件"""
+        current_time = time.time()
+        with self.lock:
+            for session_id, files in list(self.session_files.items()):
+                files_to_remove = []
+                for filepath in list(files):
+                    try:
+                        if os.path.exists(filepath):
+                            # 检查文件创建时间
+                            file_age = current_time - os.path.getctime(filepath)
+                            if file_age > 1800:  # 30分钟
+                                os.remove(filepath)
+                                files_to_remove.append(filepath)
+                                print(f"自动清理过期临时文件: {filepath}")
+                        else:
+                            files_to_remove.append(filepath)
+                    except Exception as e:
+                        print(f"清理文件 {filepath} 时出错: {e}")
+                        files_to_remove.append(filepath)
+                
+                # 从集合中移除已清理的文件
+                for filepath in files_to_remove:
+                    files.discard(filepath)
+                
+                # 如果会话没有文件了，移除会话记录
+                if not files:
+                    del self.session_files[session_id]
+    
+    def add_file(self, session_id: str, filepath: str):
+        """添加文件到会话管理"""
+        with self.lock:
+            if session_id not in self.session_files:
+                self.session_files[session_id] = set()
+            self.session_files[session_id].add(filepath)
+    
+    def cleanup_session(self, session_id: str):
+        """清理特定会话的所有文件"""
+        with self.lock:
+            if session_id in self.session_files:
+                files = self.session_files[session_id]
+                for filepath in list(files):
+                    try:
+                        if os.path.exists(filepath):
+                            os.remove(filepath)
+                            print(f"清理会话文件: {filepath}")
+                    except Exception as e:
+                        print(f"无法删除文件 {filepath}: {e}")
+                del self.session_files[session_id]
+    
+    def cleanup_all(self):
+        """清理所有临时文件"""
+        with self.lock:
+            total_files = 0
+            for session_id in list(self.session_files.keys()):
+                total_files += len(self.session_files[session_id])
+                self.cleanup_session(session_id)
+            print(f"应用程序退出，清理了 {total_files} 个临时文件")
+
+# 全局文件管理器
+file_manager = SessionFileManager()
 
 def cleanup_temp_files():
     """清理应用程序使用的临时文件"""
-    global temp_files
-    print(f"应用程序退出，清理 {len(temp_files)} 个临时文件...")
-    
-    for filepath in temp_files:
-        try:
-            if os.path.exists(filepath):
-                os.remove(filepath)
-                print(f"已删除临时文件: {filepath}")
-        except Exception as e:
-            print(f"无法删除临时文件 {filepath}: {e}")
-    
-    # 清空列表
-    temp_files = []
+    file_manager.cleanup_all()
 
 # 注册应用程序退出时的清理函数
 atexit.register(cleanup_temp_files)
 
+def get_session_id(request: gr.Request = None) -> str:
+    """获取会话ID，用于文件管理"""
+    if request and hasattr(request, 'session_hash'):
+        return request.session_hash
+    else:
+        # 如果无法获取会话ID，使用UUID
+        return str(uuid.uuid4())
+
+# 添加资源限制检查
+def check_system_resources():
+    """检查系统资源是否足够"""
+    try:
+        import psutil
+        # 检查可用内存
+        memory = psutil.virtual_memory()
+        if memory.available < 500 * 1024 * 1024:  # 少于500MB
+            return False, "系统内存不足，请稍后再试"
+        
+        # 检查磁盘空间
+        disk = psutil.disk_usage(tempfile.gettempdir())
+        if disk.free < 1024 * 1024 * 1024:  # 少于1GB
+            return False, "磁盘空间不足，请稍后再试"
+        
+        return True, "资源充足"
+    except ImportError:
+        # 如果没有psutil，跳过检查
+        return True, "无法检查资源状态"
+    except Exception as e:
+        print(f"资源检查错误: {e}")
+        return True, "资源检查失败，继续执行"
+
 def parse_rss_feed(rss_url: str):
     """回调函数：解析 RSS Feed"""
     print(f"开始解析RSS: {rss_url}")
@@ -162,11 +266,27 @@ def parse_rss_feed(rss_url: str):
             selected_episode_index_state: None
         }
 
-def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel):
+def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel, request: gr.Request = None):
     """回调函数：当用户从下拉菜单选择一个剧集时加载音频"""
-    global temp_files
     print(f"开始加载剧集音频，选择的索引: {selected_episode_index}")
     
+    # 获取会话ID
+    session_id = get_session_id(request)
+    
+    # 检查系统资源
+    resource_ok, resource_msg = check_system_resources()
+    if not resource_ok:
+        return {
+            audio_player: gr.update(value=None),
+            current_audio_url_state: None,
+            status_message_area: gr.update(value=f"⚠️ {resource_msg}"),
+            episode_shownotes: gr.update(value="", visible=False),
+            transcription_output_df: gr.update(value=None),
+            local_audio_file_path: None,
+            transcribe_button: gr.update(interactive=False),
+            selected_episode_index_state: None
+        }
+    
     if selected_episode_index is None or podcast_data is None or not podcast_data.episodes:
         print("未选择剧集或无播客数据")
         return {
@@ -235,12 +355,29 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
                 
                 # 创建临时文件
                 temp_dir = tempfile.gettempdir()
-                # 使用UUID生成唯一文件名，避免冲突
-                unique_filename = f"podcast_audio_{uuid.uuid4().hex}"
+                # 使用会话ID和UUID生成唯一文件名，避免冲突
+                unique_filename = f"podcast_audio_{session_id[:8]}_{uuid.uuid4().hex[:8]}"
                 
-                # 先发送一个HEAD请求获取内容类型
+                # 先发送一个HEAD请求获取内容类型和文件大小
                 head_response = requests.head(audio_url, timeout=30, headers=headers)
                 
+                # 检查文件大小限制（例如限制为200MB）
+                content_length = head_response.headers.get('Content-Length')
+                if content_length:
+                    file_size = int(content_length)
+                    max_size = 200 * 1024 * 1024  # 200MB
+                    if file_size > max_size:
+                        return {
+                            audio_player: gr.update(value=None),
+                            current_audio_url_state: None,
+                            status_message_area: gr.update(value=f"⚠️ 音频文件过大 ({file_size/1024/1024:.1f}MB)，超过限制 ({max_size/1024/1024}MB)"),
+                            episode_shownotes: gr.update(value=episode_shownotes_content, visible=True),
+                            transcription_output_df: gr.update(value=None),
+                            local_audio_file_path: None,
+                            transcribe_button: gr.update(interactive=False),
+                            selected_episode_index_state: None
+                        }
+                
                 # 根据内容类型确定文件扩展名
                 content_type = head_response.headers.get('Content-Type', '').lower()
                 if 'mp3' in content_type:
@@ -259,8 +396,8 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
                 
                 temp_filepath = os.path.join(temp_dir, unique_filename + file_ext)
                 
-                # 将文件路径添加到全局临时文件列表
-                temp_files.append(temp_filepath)
+                # 将文件路径添加到会话文件管理器
+                file_manager.add_file(session_id, temp_filepath)
                 
                 # 保存到临时文件
                 # 使用流式下载，避免一次性加载整个文件到内存
@@ -289,7 +426,7 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
                 return {
                     audio_player: gr.update(value=temp_filepath, label=f"Now Playing: {episode.title or 'Untitled'}"),
                     current_audio_url_state: audio_url,
-                    status_message_area: gr.update(value=f"Episode loaded: {episode.title or 'Untitled'}."),
+                    status_message_area: gr.update(value=f"✅ Episode loaded: {episode.title or 'Untitled'}."),
                     episode_shownotes: gr.update(value=episode_shownotes_content, visible=True),
                     transcription_output_df: gr.update(value=None),
                     local_audio_file_path: temp_filepath,
@@ -302,7 +439,7 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
                 return {
                     audio_player: gr.update(value=None),
                     current_audio_url_state: None,
-                    status_message_area: gr.update(value=f"Error: Failed to download audio: {e}"),
+                    status_message_area: gr.update(value=f"❌ Error: Failed to download audio: {e}"),
                     episode_shownotes: gr.update(value=episode_shownotes_content, visible=True),
                     transcription_output_df: gr.update(value=None),
                     local_audio_file_path: None,
@@ -314,7 +451,7 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
             return {
                 audio_player: gr.update(value=None),
                 current_audio_url_state: None,
-                status_message_area: gr.update(value=f"Error: Selected episode '{episode.title}' does not provide a valid audio URL."),
+                status_message_area: gr.update(value=f"❌ Error: Selected episode '{episode.title}' does not provide a valid audio URL."),
                 episode_shownotes: gr.update(value=episode_shownotes_content, visible=True),
                 transcription_output_df: gr.update(value=None),
                 local_audio_file_path: None,
@@ -326,7 +463,7 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
         return {
             audio_player: gr.update(value=None),
             current_audio_url_state: None,
-            status_message_area: gr.update(value="Error: Invalid episode index selected."),
+            status_message_area: gr.update(value="❌ Error: Invalid episode index selected."),
             episode_shownotes: gr.update(value="", visible=False),
             transcription_output_df: gr.update(value=None),
             local_audio_file_path: None,
@@ -339,7 +476,7 @@ def load_episode_audio(selected_episode_index: int, podcast_data: PodcastChannel
         return {
             audio_player: gr.update(value=None),
             current_audio_url_state: None,
-            status_message_area: gr.update(value=f"Serious error occurred while loading audio: {e}"),
+            status_message_area: gr.update(value=f"❌ Serious error occurred while loading audio: {e}"),
             episode_shownotes: gr.update(value="", visible=False),
             transcription_output_df: gr.update(value=None),
             local_audio_file_path: None,
@@ -361,11 +498,22 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
     """回调函数：开始转录当前加载的音频"""
     print(f"开始转录本地音频文件: {local_audio_file_path}, 选中剧集索引: {selected_episode_index}")
     
+    # 检查系统资源
+    resource_ok, resource_msg = check_system_resources()
+    if not resource_ok:
+        return {
+            transcription_output_df: gr.update(value=None),
+            status_message_area: gr.update(value=f"⚠️ {resource_msg}"),
+            parse_button: gr.update(interactive=True),
+            episode_dropdown: gr.update(interactive=True),
+            transcribe_button: gr.update(interactive=True)
+        }
+    
     if not local_audio_file_path or not os.path.exists(local_audio_file_path):
         print("没有可用的本地音频文件")
         return {
             transcription_output_df: gr.update(value=None),
-            status_message_area: gr.update(value="Error: No valid audio file for transcription. Please select an episode first."),
+            status_message_area: gr.update(value="❌ Error: No valid audio file for transcription. Please select an episode first."),
             parse_button: gr.update(interactive=True),
             episode_dropdown: gr.update(interactive=True),
             transcribe_button: gr.update(interactive=True)
@@ -380,7 +528,19 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
         
         # 从文件加载音频
         audio_segment = AudioSegment.from_file(local_audio_file_path)
-        print(f"音频加载完成，时长: {len(audio_segment)/1000}秒")
+        audio_duration = len(audio_segment) / 1000  # 转换为秒
+        print(f"音频加载完成，时长: {audio_duration}秒")
+        
+        # 检查音频时长限制（例如限制为60分钟）
+        max_duration = 60 * 60  # 60分钟
+        if audio_duration > max_duration:
+            return {
+                transcription_output_df: gr.update(value=None),
+                status_message_area: gr.update(value=f"⚠️ 音频时长过长 ({audio_duration/60:.1f}分钟)，超过限制 ({max_duration/60}分钟)"),
+                parse_button: gr.update(interactive=True),
+                episode_dropdown: gr.update(interactive=True),
+                transcribe_button: gr.update(interactive=True)
+            }
             
         progress(0.4, desc="Audio loaded, starting transcription (this may take a while)...")
         
@@ -396,7 +556,7 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
         result: CombinedTranscriptionResult = transcribe_podcast_audio(audio_segment,
                                                                        podcast_info=podcast_data,
                                                                        episode_info=episode_info,
-                                                                       segmentation_batch_size=64,
+                                                                       segmentation_batch_size=32,  # 减少批次大小以节省内存
                                                                        parallel=True)
         print(f"转录完成，结果: {result is not None}, 段落数: {len(result.segments) if result and result.segments else 0}")
         progress(0.9, desc="Transcription completed, formatting results...")
@@ -410,7 +570,7 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
             progress(1.0, desc="Transcription results generated!")
             return {
                 transcription_output_df: gr.update(value=formatted_segments),
-                status_message_area: gr.update(value=f"Transcription completed! {len(result.segments)} segments generated. {result.num_speakers} speakers detected."),
+                status_message_area: gr.update(value=f"✅ Transcription completed! {len(result.segments)} segments generated. {result.num_speakers} speakers detected."),
                 parse_button: gr.update(interactive=True),
                 episode_dropdown: gr.update(interactive=True),
                 transcribe_button: gr.update(interactive=True)
@@ -419,7 +579,7 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
             progress(1.0, desc="Transcription completed, but no text segments")
             return {
                 transcription_output_df: gr.update(value=None),
-                status_message_area: gr.update(value="Transcription completed, but no text segments were generated."),
+                status_message_area: gr.update(value="⚠️ Transcription completed, but no text segments were generated."),
                 parse_button: gr.update(interactive=True),
                 episode_dropdown: gr.update(interactive=True),
                 transcribe_button: gr.update(interactive=True)
@@ -428,7 +588,7 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
             progress(1.0, desc="Transcription failed")
             return {
                 transcription_output_df: gr.update(value=None),
-                status_message_area: gr.update(value="Transcription failed, no results obtained."),
+                status_message_area: gr.update(value="❌ Transcription failed, no results obtained."),
                 parse_button: gr.update(interactive=True),
                 episode_dropdown: gr.update(interactive=True),
                 transcribe_button: gr.update(interactive=True)
@@ -439,7 +599,7 @@ def start_transcription(local_audio_file_path: str, podcast_data: PodcastChannel
         progress(1.0, desc="Transcription failed: processing error")
         return {
             transcription_output_df: gr.update(value=None),
-            status_message_area: gr.update(value=f"Serious error occurred during transcription: {e}"),
+            status_message_area: gr.update(value=f"❌ Serious error occurred during transcription: {e}"),
             parse_button: gr.update(interactive=True),
             episode_dropdown: gr.update(interactive=True),
             transcribe_button: gr.update(interactive=True)
@@ -467,6 +627,14 @@ with gr.Blocks(title="Podcast Transcriber v2", css="""
     border-radius: 8px;
     box-shadow: 0 4px 8px rgba(0,0,0,0.1);
 }
+.resource-warning {
+    background-color: #fff3cd;
+    border: 1px solid #ffeaa7;
+    border-radius: 6px;
+    padding: 10px;
+    margin: 10px 0;
+    color: #856404;
+}
 """) as demo:
     gr.Markdown("# 🎙️ Podcast Transcriber")