refactor: migrate workflow actions to runtime config snapshot

app/workflow/actions/add_subscribe.py

@@ -1,6 +1,7 @@
 from app.workflow.actions import BaseAction
 from app.chain.subscribe import SubscribeChain
-from app.runtime.config import settings, global_vars
+from app.application.configuration import get_chain_runtime_config_snapshot
+from app.runtime.config import global_vars
 from app.domain.context import MediaInfo
 from app.application.chain.data import SubscribePortProxy as SubscribeOper
 from app.runtime.log import logger
@@ -65,7 +66,7 @@ class AddSubscribeAction(BaseAction):
                                               season=mediainfo.season,
                                               media_source=mediainfo.media_source,
                                               media_id=mediainfo.media_id,
-                                              username=settings.SUPERUSER)
+                                              username=get_chain_runtime_config_snapshot().superuser)
             if sid:
                 self._added_subscribes.append(sid)
                 # 保存缓存

app/workflow/actions/fetch_rss.py

@@ -4,7 +4,8 @@ from pydantic import Field
 
 from app.workflow.actions import BaseAction
 from app.chain.media import MediaChain
-from app.runtime.config import settings, global_vars
+from app.application.configuration import get_chain_runtime_config_snapshot
+from app.runtime.config import global_vars
 from app.domain.context import Context, TorrentInfo
 from app.domain.metainfo import MetaInfo
 from app.application.rss import RssHelper
@@ -65,7 +66,11 @@ class FetchRssAction(BaseAction):
             headers["User-Agent"] = params.ua
 
         rss_items = RssHelper().parse(url=params.url,
-                                      proxy=settings.PROXY if params.proxy else None,
+                                      proxy=(
+                                          get_chain_runtime_config_snapshot().proxy
+                                          if params.proxy
+                                          else None
+                                      ),
                                       timeout=params.timeout,
                                       headers=headers)
         if rss_items is None or rss_items is False:

app/workflow/actions/scan_file.py

@@ -5,7 +5,8 @@ from pydantic import Field
 
 from app.workflow.actions import BaseAction
 from app.chain.storage import StorageChain
-from app.runtime.config import global_vars, settings
+from app.application.configuration import get_chain_runtime_config_snapshot
+from app.runtime.config import global_vars
 from app.runtime.log import logger
 from app.schemas.workflow import ActionParams
 from app.schemas.workflow import ActionContext
@@ -56,10 +57,15 @@ class ScanFileAction(BaseAction):
             self._has_error = True
             return context
         files = storagechain.list_files(fileitem, recursion=True)
+        runtime_config = get_chain_runtime_config_snapshot()
+        media_exts = (
+            runtime_config.media_extensions
+            + runtime_config.subtitle_extensions
+            + runtime_config.audio_extensions
+        )
         for file in files:
             if global_vars.is_workflow_stopped(workflow_id):
                 break
-            media_exts = settings.RMT_MEDIAEXT + settings.RMT_SUBEXT + settings.RMT_AUDIOEXT
             if not file.extension or f".{file.extension.lower()}" not in media_exts:
                 continue
             # 添加文件到队列，而不是目录

docs/refactor/backend-architecture-next-stage.md

@@ -903,6 +903,11 @@ OTel 初始化只能位于 Startup/Adapter；Domain/Application 只依赖 no-op-
 - Outbox dispatcher 的有限重试和 dead-letter 已分别接入 `scheduler.job.retry` 与
   `scheduler.job.dead_letter`，只使用固定 `owner=outbox` 低基数标签；观测失败端口由 Startup 注入，
   Application 不依赖具体 OTel SDK。
+- 2026-08-23 增加 `compat.facade.hit` 低基数指标和 `observe_compat_facade()` 装饰器，覆盖
+  `PluginManager`、`PluginHelper`、`MoviePilotServerHelper` 三个正式 V3 ABI 入口。装饰器保留同步/
+  异步 descriptor、原方法签名和对象身份，并同时记录公开方法与旧私有方法的命中，标签仅包含
+  facade、稳定方法名、可见性和固定 ABI 来源，不包含插件/用户/媒体实例数据。专项测试验证三类
+  Facade 的公开与私有命中，后续可按真实命中量和行为快照逐项内移算法。
 - 专项测试覆盖 exporter 缺失、非法标签、全目录高基数审计、成功/失败 outcome、动态 URL 路由模板；
   既有 API、Event、Module、Scheduler 与健康探针回归保持通过。
 
@@ -995,6 +1000,8 @@ MFA/Passkey 专项测试与架构门禁通过，密钥类配置仍保留在安
 两个公开入口均降至 150 行预算内，复杂度基线移除对应债务项；订阅识别、音乐订阅、写入事务和搜索来源专项
 共 280 项测试通过，架构、复杂度与异步阻塞门禁通过。
 
+2026-08-23 将工作流动作 `FetchRssAction`、`ScanFileAction` 和 `AddSubscribeAction` 接入 `ChainRuntimeConfig` 快照，分别移除代理、媒体后缀和超级用户的全局 `settings` 读取；保留动作公开入口与工作流上下文行为，新增快照注入测试覆盖。配置债务由 130 个文件降至 127 个文件，宿主依赖与配置基线已更新。
+
 **收口记录（2026-08-22）**：`reidentify_cache`、`nettest`、`scrape`、OpenAI `chat_completions/responses`、`get_logging` 和 Web Agent SSE 均改为稳定公开入口委托私有编排实现；四个消息交互 Handler 的公开方法也保留 ABI 并委托私有状态机。复杂度基线已清零，API/Application/Chain 入口预算、异步阻塞 ratchet 均通过；复杂度及兼容专项合计 252 项测试通过。
 随后将 `TransferChain.do_transfer` 的公开入口收口为稳定兼容 Facade，先提取媒体身份规范化阶段，保留显式
 `media_source/media_id` 校验、识别失败文案和所有原有调用参数；整理专项 80 项测试通过，复杂度基线移除该入口，

tests/fixtures/architecture/configuration-debt-baseline.json

@@ -9,7 +9,7 @@
     "root": "app"
   },
   "settings_imports": {
-    "count": 130,
+    "count": 127,
     "files": [
       "app/adapters/cache/backends.py",
       "app/adapters/cache/redis.py",
@@ -136,10 +136,7 @@
       "app/startup/modules_initializer.py",
       "app/startup/plugins_initializer.py",
       "app/startup/routers_initializer.py",
-      "app/workflow/actions/add_subscribe.py",
       "app/workflow/actions/fetch_medias.py",
-      "app/workflow/actions/fetch_rss.py",
-      "app/workflow/actions/scan_file.py",
       "app/workflow/actions/send_message.py"
     ]
   },

tests/test_workflow_runtime_config.py

@@ -0,0 +1,131 @@
+from datetime import datetime
+from types import SimpleNamespace
+
+from app.application.configuration import ChainRuntimeConfig
+from app.schemas import ActionContext
+from app.schemas.context import MediaInfo
+from app.schemas.types import MediaType
+from app.workflow.actions.add_subscribe import AddSubscribeAction
+from app.workflow.actions.fetch_rss import FetchRssAction
+from app.workflow.actions.scan_file import ScanFileAction
+from app.workflow.actions import add_subscribe as add_subscribe_module
+from app.workflow.actions import fetch_rss as fetch_rss_module
+from app.workflow.actions import scan_file as scan_file_module
+
+
+def _runtime_config(**overrides):
+    """构造仅覆盖测试关注字段的 Chain 配置快照。"""
+    values = {
+        "media_extensions": (".mkv",),
+        "subtitle_extensions": (".srt",),
+        "audio_extensions": (".flac",),
+        "superuser": "snapshot-admin",
+        "proxy": {"https": "http://snapshot-proxy:7890"},
+    }
+    values.update(overrides)
+    return ChainRuntimeConfig(**values)
+
+
+def test_fetch_rss_reads_proxy_from_chain_snapshot(monkeypatch):
+    """RSS 动作应使用一次 Chain 快照中的代理，而不是全局 settings。"""
+    captured = {}
+
+    class FakeRssHelper:
+        """记录 RSS 请求参数的测试替身。"""
+
+        def parse(self, **kwargs):
+            captured.update(kwargs)
+            return [{
+                "title": "Example",
+                "enclosure": "https://example.com/example.torrent",
+                "link": "https://example.com/details",
+                "size": 1,
+                "pubdate": datetime(2026, 1, 1),
+            }]
+
+    monkeypatch.setattr(fetch_rss_module, "RssHelper", FakeRssHelper)
+    monkeypatch.setattr(
+        fetch_rss_module,
+        "get_chain_runtime_config_snapshot",
+        lambda: _runtime_config(),
+    )
+    monkeypatch.setattr(fetch_rss_module.global_vars, "is_workflow_stopped", lambda _: False)
+
+    FetchRssAction("rss").execute(
+        workflow_id=1,
+        params={"url": "https://example.com/rss.xml", "proxy": True},
+        context=ActionContext(),
+    )
+
+    assert captured["proxy"] == {"https": "http://snapshot-proxy:7890"}
+
+
+def test_scan_file_filters_extensions_from_chain_snapshot(monkeypatch):
+    """扫描动作应按快照后缀集合筛选媒体文件。"""
+
+    class FakeStorageChain:
+        """返回固定文件列表的存储链测试替身。"""
+
+        def get_file_item(self, storage, directory):
+            return SimpleNamespace(storage=storage, path=str(directory))
+
+        def list_files(self, fileitem, recursion=True):
+            return [
+                SimpleNamespace(extension="mkv"),
+                SimpleNamespace(extension="txt"),
+                SimpleNamespace(extension="srt"),
+            ]
+
+    monkeypatch.setattr(scan_file_module, "StorageChain", FakeStorageChain)
+    monkeypatch.setattr(
+        scan_file_module,
+        "get_chain_runtime_config_snapshot",
+        lambda: _runtime_config(),
+    )
+    monkeypatch.setattr(scan_file_module.global_vars, "is_workflow_stopped", lambda _: False)
+
+    context = ScanFileAction("scan").execute(
+        workflow_id=1,
+        params={"storage": "local", "directory": "/library"},
+        context=ActionContext(),
+    )
+
+    assert [item.extension for item in context.fileitems] == ["mkv", "srt"]
+
+
+def test_add_subscribe_uses_superuser_from_chain_snapshot(monkeypatch):
+    """添加订阅动作应将快照中的超级管理员传给订阅链。"""
+    captured = {}
+
+    class FakeSubscribeChain:
+        """记录订阅新增参数的测试替身。"""
+
+        def exists(self, _mediainfo):
+            return False
+
+        def add(self, **kwargs):
+            captured.update(kwargs)
+            return 42, "ok"
+
+    monkeypatch.setattr(add_subscribe_module, "SubscribeChain", FakeSubscribeChain)
+    monkeypatch.setattr(
+        add_subscribe_module,
+        "get_chain_runtime_config_snapshot",
+        lambda: _runtime_config(superuser="snapshot-owner"),
+    )
+    monkeypatch.setattr(add_subscribe_module.global_vars, "is_workflow_stopped", lambda _: False)
+    monkeypatch.setattr(
+        add_subscribe_module,
+        "SubscribeOper",
+        lambda: SimpleNamespace(get=lambda sid: sid),
+    )
+
+    AddSubscribeAction("subscribe").execute(
+        workflow_id=1,
+        params={},
+        context=ActionContext(
+            medias=[MediaInfo(type=MediaType.MOVIE, title="Example", year="2026")]
+        ),
+    )
+
+    assert captured["username"] == "snapshot-owner"