fix(ingest): raise ref cap to 50 and clarify invalid bvid error

DEVELOPMENT.md

@@ -489,7 +489,7 @@ P3 实现说明：
 - `callback_data` 使用 `add:<aid>` / `ign:<aid>` / `blk:<aid>`（`fav:` 兼容旧消息）；
 - 入库只写本地库；`video` 的 `removed` / `dropped` / `unfaved` 状态会被重新激活；
 - 引用解析是纯函数 `video_ref.extract_video_refs`，`b23.tv` 短链由 `bilibili.client.resolve_short_link` 跟随跳转；
-- 手动入库不做「标题含咻咻满」等规则校验；单条消息的引用数量有上限；
+- 手动入库不做「标题含咻咻满」等规则校验；单条消息最多处理 50 个引用，超出会在回执中提示分批发送；无效 BV/av 号（`-400`）逐条报错、不中断其余引用；
 - Telegram 只接受 `chat_id` 与 `from.id` 都等于配置私聊用户的回调与消息。
 
 ---

reaction_board/notify/telegram.py

@@ -80,6 +80,9 @@ def format_manual_reply(manual_result: Any) -> str:
         return "未识别到 B 站视频引用，请发送视频链接或 BV 号，例如 BV1x8x166EzY。"
     results = list(getattr(manual_result, "results", []) or [])
     lines = [_result_line(item) for item in results]
+    truncated = int(getattr(manual_result, "truncated", 0) or 0)
+    if truncated > 0:
+        lines.append(f"⚠️ 本消息还有 {truncated} 个引用未处理，请分批再发一次。")
     header = f"共识别 {refs} 个引用，入库 {int(getattr(manual_result, 'added', 0) or 0)} 条："
     return "\n".join([header, *lines])
 

reaction_board/services/library.py

@@ -11,7 +11,11 @@ import time
 from dataclasses import dataclass, field
 from typing import Any
 
-from ..bilibili.client import BilibiliError, BilibiliProtocolError
+from ..bilibili.client import (
+    BilibiliAPIError,
+    BilibiliError,
+    BilibiliProtocolError,
+)
 from ..config import ConfigStore
 from ..repository import Repository
 from ..timeutil import shanghai_date
@@ -42,6 +46,7 @@ class IngestResult:
 class ManualIngestResult:
     refs: int = 0
     added: int = 0
+    truncated: int = 0
     results: list[IngestResult] = field(default_factory=list)
 
     @property
@@ -52,6 +57,7 @@ class ManualIngestResult:
         return {
             "refs": self.refs,
             "added": self.added,
+            "truncated": self.truncated,
             "ok": self.ok,
             "results": [result.as_dict() for result in self.results],
         }
@@ -180,16 +186,30 @@ class LibraryService:
     # ------------------------------------------------------------ manual
 
     def ingest_text(self, text: object) -> ManualIngestResult:
-        refs = extract_video_refs(text, limit=self.max_refs)
-        result = ManualIngestResult(refs=len(refs))
-        if not refs:
+        all_refs = extract_video_refs(text, limit=None)
+        result = ManualIngestResult(refs=len(all_refs))
+        if not all_refs:
             return result
+        if len(all_refs) > self.max_refs:
+            result.truncated = len(all_refs) - self.max_refs
+            all_refs = all_refs[: self.max_refs]
+            result.refs = len(all_refs)
 
         client = self.read_provider.get()
         ups: list[tuple[int | None, str]] = []
-        for ref in refs:
+        for ref in all_refs:
             try:
                 detail = self._resolve_ref(client, ref)
+            except BilibiliAPIError as exc:
+                message = (
+                    "无法解析该视频引用：BV 号/av 号无效或视频不存在（code=-400）"
+                    if exc.code == -400 else str(exc)
+                )
+                result.results.append(IngestResult(
+                    ok=False, aid=None, state="error", message=message,
+                ))
+                self._log("warning", "ingest.manual_failed", error=str(exc))
+                continue
             except BilibiliError as exc:
                 result.results.append(IngestResult(
                     ok=False, aid=None, state="error", message=str(exc),

reaction_board/video_ref.py

@@ -25,7 +25,8 @@ _AID_RE = re.compile(r"(?<![0-9A-Za-z])av(\d{1,20})(?![0-9A-Za-z])", re.IGNORECA
 _SHORT_RE = re.compile(r"b23\.tv/([0-9A-Za-z]+)")
 
 # 单条消息最多处理的引用数，避免误粘贴一段文本导致突发请求。
-DEFAULT_MAX_REFS = 10
+# 15 个 BV 号批量测试手动入库是常见用法，上限放宽到 50，超出时回执会提示分批发送。
+DEFAULT_MAX_REFS = 50
 
 
 @dataclass(frozen=True)

tests/test_library.py

@@ -13,7 +13,10 @@ sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
 sys.stdout.reconfigure(encoding="utf-8")
 
 from fixtures import view_for  # noqa: E402
-from reaction_board.bilibili.client import BilibiliError  # noqa: E402
+from reaction_board.bilibili.client import (  # noqa: E402
+    BilibiliAPIError,
+    BilibiliError,
+)
 from reaction_board.bilibili.models import VideoDetail  # noqa: E402
 from reaction_board.config import ConfigStore  # noqa: E402
 from reaction_board.database import Database  # noqa: E402
@@ -51,14 +54,18 @@ def detail(
 
 
 class FakeReadClient:
-    def __init__(self, by_bvid=None, by_aid=None, short_map=None, fail_bvids=()):
+    def __init__(self, by_bvid=None, by_aid=None, short_map=None, fail_bvids=(),
+                 api_errors=None):
         self.by_bvid = by_bvid or {}
         self.by_aid = by_aid or {}
         self.short_map = short_map or {}
         self.fail_bvids = set(fail_bvids)
+        self.api_errors = dict(api_errors or {})
 
     def view_detail(self, *, aid=None, bvid=None):
         if bvid is not None:
+            if bvid in self.api_errors:
+                raise BilibiliAPIError(self.api_errors[bvid], "请求错误")
             if bvid in self.fail_bvids:
                 raise BilibiliError(f"view 失败: {bvid}")
             if bvid in self.by_bvid:
@@ -263,6 +270,44 @@ def test_snapshot_not_overwritten():
             db.close()
 
 
+def test_manual_ingest_truncation():
+    with tempfile.TemporaryDirectory() as tmp:
+        bvs = {
+            "BV1x8x166E00": detail(1, bvid="BV1x8x166E00"),
+            "BV1x8x166E01": detail(2, bvid="BV1x8x166E01"),
+            "BV1x8x166E02": detail(3, bvid="BV1x8x166E02"),
+        }
+        client = FakeReadClient(by_bvid=bvs)
+        store = ConfigStore(write_config(tmp), env={})
+        db = Database(Path(tmp) / "app.db", backup_dir=Path(tmp) / "backup")
+        db.initialize()
+        try:
+            repo = Repository(db, clock=lambda: NOW)
+            service = LibraryService(
+                store, repo, FakeProvider(client), clock=lambda: NOW, max_refs=2
+            )
+            result = service.ingest_text(" ".join(bvs))
+            assert result.refs == 2
+            assert result.truncated == 1
+            assert result.added == 2
+            assert len(result.results) == 2
+        finally:
+            db.close()
+
+
+def test_manual_ingest_invalid_bvid_message():
+    with tempfile.TemporaryDirectory() as tmp:
+        client = FakeReadClient(api_errors={"BV1x8x166E00": -400})
+        db, repo, service = setup(tmp, client)
+        try:
+            result = service.ingest_text("BV1x8x166E00")
+            assert result.refs == 1
+            assert result.results[0].state == "error"
+            assert "无效或视频不存在" in result.results[0].message
+        finally:
+            db.close()
+
+
 def test_ignore_and_blacklist():
     with tempfile.TemporaryDirectory() as tmp:
         client = FakeReadClient()
@@ -289,6 +334,8 @@ def main():
     test_manual_ingest_forms_and_multiple()
     test_manual_ingest_error_isolation()
     test_snapshot_not_overwritten()
+    test_manual_ingest_truncation()
+    test_manual_ingest_invalid_bvid_message()
     test_ignore_and_blacklist()
     print("OK: 入库、手动入库、忽略、拉黑测试全部通过")
 

tests/test_telegram.py

@@ -15,6 +15,7 @@ from reaction_board.notify.telegram import (  # noqa: E402
     TelegramClient,
     TelegramError,
     format_duration,
+    format_manual_reply,
 )
 from reaction_board.services.library import (  # noqa: E402
     ActionResult,
@@ -51,6 +52,19 @@ def test_format_duration():
     assert format_duration(3725) == "1:02:05"
 
 
+def test_manual_reply_truncation_notice():
+    result = ManualIngestResult(
+        refs=50, added=1, truncated=2,
+        results=[IngestResult(True, 1, "added", title="t", message="已入库")],
+    )
+    text = format_manual_reply(result)
+    assert "共识别 50 个引用" in text
+    assert "还有 2 个引用未处理" in text
+    assert format_manual_reply(ManualIngestResult()) == (
+        "未识别到 B 站视频引用，请发送视频链接或 BV 号，例如 BV1x8x166EzY。"
+    )
+
+
 def test_send_candidates_and_chunking():
     transport = FakeTransport()
     bot = TelegramClient("token", 42, transport=transport)
@@ -300,6 +314,7 @@ def test_poller_backoff_and_recovery():
 
 def main():
     test_format_duration()
+    test_manual_reply_truncation_notice()
     test_send_candidates_and_chunking()
     test_send_candidates_dedupes_by_aid()
     test_get_updates_and_answer()

tests/test_video_ref.py

@@ -63,6 +63,17 @@ def test_limit():
     assert len(extract_video_refs(text, limit=None)) == 3
 
 
+def test_default_cap_allows_bulk_paste():
+    from reaction_board.video_ref import DEFAULT_MAX_REFS
+
+    assert DEFAULT_MAX_REFS >= 15
+    text = " ".join(f"BV1x8x166E{i:02d}" for i in range(15))
+    assert len(extract_video_refs(text)) == 15
+    assert len(extract_video_refs(" ".join(
+        f"BV1x8x166E{i:02d}" for i in range(DEFAULT_MAX_REFS + 5)
+    ))) == DEFAULT_MAX_REFS
+
+
 def test_invalid_and_boundaries():
     assert extract_video_refs("") == []
     assert extract_video_refs(None) == []
@@ -85,6 +96,7 @@ def main():
     test_short_link()
     test_multiple_and_order_and_dedup()
     test_limit()
+    test_default_cap_allows_bulk_paste()
     test_invalid_and_boundaries()
     print("OK: 视频引用（BV/av/URL/短链）解析测试全部通过")