refactor(chain): 处理链功能域 mixin 化，清理未使用导入并根治兼容层循环导入

- ChainBase 拆分为 RecognitionMixin/MessageProcessingMixin/NotificationMixin
- TransferChain 拆分为 7 个功能 mixin（_mixins.py），SubscribeChain 音乐订阅域拆出 _music.py
- 斜杠命令交互四件套收敛为 InteractionChainMixin 委托，会话管理器移至 application 层，chain 层不再 re-export
- 模块基础类收敛到 app/modules/_base（notification/mediaserver 语义重命名）
- 清理 app/chain/__init__.py 24 个未使用导入，修正 49 处测试 patch 目标到实际命名空间
- 兼容层 legacy 符号不再并入 __all__，根治 schemas 初始化反向拉起 application.transfer 的循环导入
- 修复 bangumi 集数为字符串时 set_bangumi_info 抛 TypeError
- 新增重复代码等架构门禁测试；capability 清单校验排除下划线内部目录

app/chain/message.py

@@ -11,8 +11,12 @@ from pathlib import Path
 from typing import Any, Optional, Dict, Union, List, Tuple
 from urllib.parse import unquote, urlparse
 
-from app.agent.orchestrator import agent_manager
-from app.agent.llm import AgentCapabilityManager, LLMHelper
+from app.application.agent import (
+    get_agent_manager,
+    is_audio_input_available,
+    supports_image_input,
+    transcribe_audio,
+)
 from app.chain import ChainBase
 from app.chain.download import DownloadChain
 from app.chain.media import MediaChain
@@ -68,7 +72,7 @@ class MessageChain(ChainBase):
             return
         clear_task = None
         try:
-            clear_task = agent_manager.clear_session(session_id=session_id, user_id=str(userid))
+            clear_task = get_agent_manager().clear_session(session_id=session_id, user_id=str(userid))
             asyncio.run_coroutine_threadsafe(
                 clear_task,
                 global_vars.loop,
@@ -346,7 +350,7 @@ class MessageChain(ChainBase):
         if not session_info:
             return False
         session_id, _ = session_info
-        if not agent_manager.matches_secret_confirmation(
+        if not get_agent_manager().matches_secret_confirmation(
             session_id,
             str(userid),
             channel=channel.value,
@@ -966,7 +970,7 @@ class MessageChain(ChainBase):
         if session_id:
             clear_task = None
             try:
-                clear_task = agent_manager.clear_session(
+                clear_task = get_agent_manager().clear_session(
                     session_id=session_id, user_id=str(userid)
                 )
                 asyncio.run_coroutine_threadsafe(
@@ -1015,7 +1019,7 @@ class MessageChain(ChainBase):
             session_id, _ = session_info
             try:
                 future = asyncio.run_coroutine_threadsafe(
-                    agent_manager.stop_current_task(session_id=session_id),
+                    get_agent_manager().stop_current_task(session_id=session_id),
                     global_vars.loop,
                 )
                 stopped = future.result(timeout=10)
@@ -1180,7 +1184,7 @@ class MessageChain(ChainBase):
             return
 
         session_id, _ = session_info
-        status = agent_manager.get_session_status(session_id=session_id)
+        status = get_agent_manager().get_session_status(session_id=session_id)
         self.post_message(
             Notification(
                 channel=channel,
@@ -1254,7 +1258,7 @@ class MessageChain(ChainBase):
             # 将可直接输入给 LLM 的附件统一转换为 data URL
             original_images = images
             all_files = list(files or [])
-            if images and LLMHelper.supports_image_input(
+            if images and supports_image_input(
                     provider=settings.LLM_PROVIDER,
                     model=settings.LLM_MODEL,
             ):
@@ -1333,7 +1337,7 @@ class MessageChain(ChainBase):
                 process_kwargs["has_audio_input"] = True
             # 在事件循环中处理
             asyncio.run_coroutine_threadsafe(
-                agent_manager.process_message(**process_kwargs),
+                get_agent_manager().process_message(**process_kwargs),
                 global_vars.loop,
             )
             return True
@@ -1353,7 +1357,7 @@ class MessageChain(ChainBase):
         """
         if not audio_refs:
             return None
-        if not AgentCapabilityManager.is_audio_input_available():
+        if not is_audio_input_available():
             logger.warning("音频输入能力未配置或未启用，跳过语音识别")
             return None
 
@@ -1460,7 +1464,7 @@ class MessageChain(ChainBase):
                     )
                     continue
 
-                transcript = AgentCapabilityManager.transcribe_audio(
+                transcript = transcribe_audio(
                     content=content, filename=filename
                 )
                 if transcript: