fix(subscribe): preserve source identity when adding subscriptions

app/api/endpoints/media.py

@@ -366,6 +366,9 @@ async def seasons(
             )
             if mediainfo:
                 return _build_media_seasons(mediainfo, season)
+        # 明确来源的查询不能按标题切换到默认识别源，避免辅助 TMDB 信息替换主身份。
+        if media_source and source_media_id:
+            return []
     if title:
         meta = MetaInfo(title)
         if year:

app/chain/subscribe.py

@@ -904,6 +904,16 @@ class SubscribeChain(ChainBase):
             metainfo.begin_season = season
         if not media_source and not media_id and mediaid:
             media_source, media_id = parse_media_key(mediaid)
+        resolved_source, resolved_media_id = resolve_media_identity(
+            source=media_source,
+            media_id=media_id,
+            tmdbid=tmdbid,
+            doubanid=doubanid,
+            bangumiid=bangumiid,
+            anilistid=anilistid,
+        )
+        if resolved_source and resolved_media_id:
+            media_source, media_id = resolved_source, resolved_media_id
         if any((media_id, tmdbid, doubanid, bangumiid, anilistid)):
             mediainfo = self.recognize_media(
                 meta=metainfo,
@@ -926,10 +936,11 @@ class SubscribeChain(ChainBase):
             if season is None:
                 season = meta.begin_season
 
-        # 使用名称识别兜底
+        # 明确来源时只允许在同一来源内按名称兜底，不能切换主识别源。
         if not mediainfo:
             mediainfo = MediaChain().recognize_by_meta(
                 metainfo,
+                source=media_source,
                 episode_group=episode_group,
                 obtain_images=False,
             )
@@ -1099,6 +1110,16 @@ class SubscribeChain(ChainBase):
             metainfo.begin_season = season
         if not media_source and not media_id and mediaid:
             media_source, media_id = parse_media_key(mediaid)
+        resolved_source, resolved_media_id = resolve_media_identity(
+            source=media_source,
+            media_id=media_id,
+            tmdbid=tmdbid,
+            doubanid=doubanid,
+            bangumiid=bangumiid,
+            anilistid=anilistid,
+        )
+        if resolved_source and resolved_media_id:
+            media_source, media_id = resolved_source, resolved_media_id
         if any((media_id, tmdbid, doubanid, bangumiid, anilistid)):
             mediainfo = await self.async_recognize_media(
                 meta=metainfo,
@@ -1121,10 +1142,11 @@ class SubscribeChain(ChainBase):
             if season is None:
                 season = meta.begin_season
 
-        # 使用名称识别兜底
+        # 明确来源时只允许在同一来源内按名称兜底，不能切换主识别源。
         if not mediainfo:
             mediainfo = await MediaChain().async_recognize_by_meta(
                 metainfo,
+                source=media_source,
                 episode_group=episode_group,
                 obtain_images=False,
             )

tests/test_search_media_sources.py

@@ -1,7 +1,12 @@
 import asyncio
+from unittest.mock import AsyncMock, Mock, patch
+
+import pytest
 
 from app.api.endpoints import media as media_endpoint
 from app.api.endpoints import search as search_endpoint
+from app.chain import subscribe as subscribe_module
+from app.chain.subscribe import SubscribeChain
 from app.core.context import MediaInfo
 from app.schemas.types import MediaType
 
@@ -150,3 +155,146 @@ def test_media_seasons_builds_anilist_season_response(monkeypatch) -> None:
     assert result[0].poster_path == media.poster_path
     assert captured["source"] == "anilist"
     assert captured["mediaid"] == "154587"
+
+
+@pytest.mark.parametrize(
+    ("mediaid", "media_kwargs", "episode_count"),
+    [
+        (
+            "douban:db-7301",
+            {
+                "douban_info": {
+                    "episodes_count": 12,
+                    "id": "db-7301",
+                    "subtype": "tv",
+                    "title": "豆瓣剧集",
+                }
+            },
+            12,
+        ),
+        (
+            "bangumi:7302",
+            {
+                "bangumi_info": {
+                    "id": 7302,
+                    "name_cn": "Bangumi 剧集",
+                    "platform": "TV",
+                    "total_episodes": 13,
+                }
+            },
+            13,
+        ),
+        (
+            "anilist:7303",
+            {
+                "anilist_info": {
+                    "episodes": 14,
+                    "format": "TV",
+                    "id": 7303,
+                    "title": {"native": "AniList 剧集"},
+                }
+            },
+            14,
+        ),
+    ],
+)
+def test_media_seasons_uses_source_episode_count_and_defaults_to_first_season(
+        monkeypatch, mediaid: str, media_kwargs: dict, episode_count: int,
+) -> None:
+    """非 TMDB 来源应使用自身总集数构造第 1 季，不依赖 TMDB。"""
+    media = MediaInfo(**media_kwargs)
+    media_chain = Mock()
+    media_chain.async_recognize_media = AsyncMock(return_value=media)
+    monkeypatch.setattr(media_endpoint, "MediaChain", Mock(return_value=media_chain))
+
+    result = asyncio.run(
+        media_endpoint.seasons(mediaid=mediaid, season=None, _=None)
+    )
+
+    assert media.season is None
+    assert media.seasons[1] == list(range(1, episode_count + 1))
+    assert len(result) == 1
+    assert result[0].season_number == 1
+    assert result[0].episode_count == episode_count
+
+
+@pytest.mark.parametrize(
+    "mediaid",
+    ["douban:db-7401", "bangumi:7402", "anilist:7403"],
+)
+def test_media_seasons_does_not_fallback_to_default_source_for_explicit_identity(
+        monkeypatch, mediaid: str,
+) -> None:
+    """明确来源查询失败时应直接返回空列表，不能按标题切换到默认源。"""
+    media_chain = Mock()
+    media_chain.async_recognize_media = AsyncMock(return_value=None)
+    media_chain.async_recognize_by_meta = AsyncMock(
+        side_effect=AssertionError("不应按标题切换识别源")
+    )
+    monkeypatch.setattr(media_endpoint, "MediaChain", Mock(return_value=media_chain))
+
+    result = asyncio.run(
+        media_endpoint.seasons(
+            mediaid=mediaid,
+            title="来源查询失败剧集",
+            year="2026",
+            _=None,
+        )
+    )
+
+    assert result == []
+    media_chain.async_recognize_media.assert_awaited_once()
+    media_chain.async_recognize_by_meta.assert_not_awaited()
+
+
+def test_subscribe_add_keeps_inferred_anilist_source_during_title_fallback() -> None:
+    """同步新增订阅按兼容 ID 推导来源后，标题兜底仍应限定 AniList。"""
+    media_chain = Mock()
+    media_chain.recognize_by_meta.return_value = None
+    chain = object.__new__(SubscribeChain)
+
+    with patch.object(SubscribeChain, "recognize_media", return_value=None) as recognize, \
+            patch.object(subscribe_module, "MediaChain", return_value=media_chain):
+        sid, message = chain.add(
+            title="AniList 同步订阅",
+            year="2026",
+            mtype=MediaType.TV,
+            anilistid=154587,
+            tmdbid=209867,
+            media_source="anilist",
+        )
+
+    assert sid is None
+    assert message == "未识别到媒体信息"
+    assert recognize.call_args.kwargs["source"] == "anilist"
+    assert recognize.call_args.kwargs["mediaid"] == "154587"
+    assert media_chain.recognize_by_meta.call_args.kwargs["source"] == "anilist"
+
+
+def test_subscribe_async_add_keeps_inferred_anilist_source_during_title_fallback() -> None:
+    """异步新增订阅按兼容 ID 推导来源后，标题兜底仍应限定 AniList。"""
+    media_chain = Mock()
+    media_chain.async_recognize_by_meta = AsyncMock(return_value=None)
+    chain = object.__new__(SubscribeChain)
+
+    with patch.object(
+        SubscribeChain, "async_recognize_media", new=AsyncMock(return_value=None)
+    ) as recognize, patch.object(
+        subscribe_module, "MediaChain", return_value=media_chain
+    ):
+        sid, message = asyncio.run(
+            chain.async_add(
+                title="AniList 异步订阅",
+                year="2026",
+                mtype=MediaType.TV,
+                anilistid=154587,
+                tmdbid=209867,
+                media_source="anilist",
+            )
+        )
+
+    assert sid is None
+    assert message == "未识别到媒体信息"
+    assert recognize.await_args.kwargs["source"] == "anilist"
+    assert recognize.await_args.kwargs["mediaid"] == "154587"
+    assert media_chain.async_recognize_by_meta.await_args.kwargs["source"] == "anilist"