fix(cntfilter): allow legacy sensitive-word lists over 64 patterns (#2467)

* fix(cntfilter): allow legacy sensitive-word lists over 64 patterns

Legacy sensitive-words.json files shipped ~70 rules. After v4.10.7,
BanWordFilter treated the 64-pattern safe_regex cap as a hard failure
and blocked every message. Raise the cap only on the sensitive-word
path, keep the 50ms CPU budget, and truncate oversized lists with a
one-time warning.

Fixes #2443

* fix(cntfilter): reject oversized sensitive-word lists

---------

Co-authored-by: dadachann <185672915+dadachann@users.noreply.github.com>

src/langbot/libs/qq_official_api/api.py

@@ -422,6 +422,69 @@ class QQOfficialClient:
                 await self.logger.error(f'Failed to send private message: {response_data}')
                 raise ValueError(response)
 
+    async def _send_markdown_msg(
+        self,
+        target_type: str,
+        target_id: str,
+        content: str,
+        msg_id: Optional[str] = None,
+        event_id: Optional[str] = None,
+        msg_seq: int = 1,
+    ) -> None:
+        """Send a Markdown message to a C2C user or QQ group."""
+        if not await self.check_access_token():
+            await self.get_access_token()
+
+        if target_type == 'c2c':
+            url = f'{self.base_url}/v2/users/{target_id}/messages'
+        elif target_type == 'group':
+            url = f'{self.base_url}/v2/groups/{target_id}/messages'
+        else:
+            raise ValueError(f'Unsupported Markdown target type: {target_type}')
+
+        data: dict[str, Any] = {
+            'msg_type': 2,
+            'markdown': {'content': content},
+            'msg_seq': msg_seq,
+        }
+        if msg_id:
+            data['msg_id'] = msg_id
+        if event_id:
+            data['event_id'] = event_id
+
+        async with self._http_client_context() as client:
+            headers = {
+                'Authorization': f'QQBot {self.access_token}',
+                'Content-Type': 'application/json',
+            }
+            response = await client.post(url, headers=headers, json=data)
+            if response.status_code != 200:
+                response_data = await httpclient.parse_json_response(response)
+                await self.logger.error(f'Failed to send Markdown message: {response_data}')
+                raise ValueError(response)
+
+    async def send_private_markdown_msg(
+        self,
+        user_openid: str,
+        content: str,
+        msg_id: Optional[str] = None,
+        event_id: Optional[str] = None,
+        msg_seq: int = 1,
+    ) -> None:
+        """Send a Markdown C2C message."""
+        await self._send_markdown_msg('c2c', user_openid, content, msg_id, event_id, msg_seq)
+
+    async def send_group_markdown_msg(
+        self,
+        group_openid: str,
+        content: str,
+        msg_id: Optional[str] = None,
+        event_id: Optional[str] = None,
+        msg_seq: int = 1,
+    ) -> None:
+        """Send a Markdown QQ group message."""
+        await self._send_markdown_msg('group', group_openid, content, msg_id, event_id, msg_seq)
+
     async def send_group_text_msg(
         self,
         group_openid: str,

src/langbot/pkg/pipeline/cntfilter/filters/banwords.py

@@ -5,6 +5,11 @@ from .. import entities
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
 from ....utils.safe_regex import SafeRegexError, mask_patterns
 
+# Legacy sensitive-words.json files shipped ~70 rules, which exceeds the
+# default safe_regex per-call cap of 64 and used to fail-close every message.
+# Keep one 50ms CPU budget for the whole list; only raise the pattern cap.
+_MAX_SENSITIVE_WORD_PATTERNS = 256
+
 
 @filter_model.filter_class('ban-word-filter')
 class BanWordFilter(filter_model.ContentFilter):
@@ -14,12 +19,17 @@ class BanWordFilter(filter_model.ContentFilter):
         pass
 
     async def process(self, query: pipeline_query.Query, message: str) -> entities.FilterResult:
+        words = self.ap.sensitive_meta.data.get('words') or []
+        mask = self.ap.sensitive_meta.data['mask']
+        mask_word = self.ap.sensitive_meta.data['mask_word']
+
         try:
-            found, message = await mask_patterns(
+            found, current = await mask_patterns(
-                self.ap.sensitive_meta.data['words'],
+                words,
                 message,
-                mask=self.ap.sensitive_meta.data['mask'],
+                mask=mask,
-                mask_word=self.ap.sensitive_meta.data['mask_word'],
+                mask_word=mask_word,
+                max_pattern_count=_MAX_SENSITIVE_WORD_PATTERNS,
             )
         except SafeRegexError as exc:
             return entities.FilterResult(
@@ -31,7 +41,7 @@ class BanWordFilter(filter_model.ContentFilter):
 
         return entities.FilterResult(
             level=entities.ResultLevel.MASKED if found else entities.ResultLevel.PASS,
-            replacement=message,
+            replacement=current,
             user_notice='消息中存在不合适的内容, 请修改' if found else '',
             console_notice='',
         )

src/langbot/pkg/platform/sources/line.py

@@ -101,7 +101,7 @@ class LINEEventConverter(abstract_platform_adapter.AbstractEventConverter):
         if event.source.type == 'user':
             return platform_events.FriendMessage(
                 sender=platform_entities.Friend(
-                    id=event.message.id,
+                    id=event.source.user_id,
                     nickname=event.source.user_id,
                     remark='',
                 ),
@@ -110,13 +110,19 @@ class LINEEventConverter(abstract_platform_adapter.AbstractEventConverter):
                 source_platform_object=event,
             )
         else:
+            # 'group' and 'room' sources carry the stable chat id under different
+            # field names; user_id may be absent for some members, so fall back
+            # to the group/room id rather than the per-message id.
+            group_id = event.source.group_id if event.source.type == 'group' else event.source.room_id
+            member_id = event.source.user_id or group_id
+
             return platform_events.GroupMessage(
                 sender=platform_entities.GroupMember(
-                    id=event.event.sender.sender_id.open_id,
+                    id=member_id,
-                    member_name=event.event.sender.sender_id.union_id,
+                    member_name=member_id,
                     permission=platform_entities.Permission.Member,
                     group=platform_entities.Group(
-                        id=event.message.id,
+                        id=group_id,
                         name='',
                         permission=platform_entities.Permission.Member,
                     ),

src/langbot/pkg/platform/sources/qqofficial.py

@@ -329,17 +329,12 @@ class QQOfficialAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter
             content_type = content.get('type', 'text')
 
             if content_type == 'text':
-                if target_type == 'c2c':
+                if target_type in {'c2c', 'group'}:
-                    await self.bot.send_private_text_msg(
+                    await self._send_c2c_or_group_text_reply(
+                        target_type,
                         target_id,
                         content['content'],
-                        qq_official_event.d_id,
+                        msg_id=qq_official_event.d_id,
-                    )
-                elif target_type == 'group':
-                    await self.bot.send_group_text_msg(
-                        target_id,
-                        content['content'],
-                        qq_official_event.d_id,
                     )
 
             elif content_type == 'image':
@@ -383,6 +378,39 @@ class QQOfficialAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter
     async def send_message(self, target_type: str, target_id: str, message: platform_message.MessageChain):
         pass
 
+    async def _send_c2c_or_group_text_reply(
+        self,
+        target_type: str,
+        target_id: str,
+        content: str,
+        *,
+        msg_id: typing.Optional[str] = None,
+        event_id: typing.Optional[str] = None,
+        msg_seq: int = 1,
+    ) -> None:
+        """Send a text reply using the configured C2C/group render mode."""
+        use_markdown = self.config.get('enable-markdown-rendering', False)
+        if target_type == 'c2c':
+            send = self.bot.send_private_markdown_msg if use_markdown else self.bot.send_private_text_msg
+            await send(
+                user_openid=target_id,
+                content=content,
+                msg_id=msg_id,
+                event_id=event_id,
+                msg_seq=msg_seq,
+            )
+        elif target_type == 'group':
+            send = self.bot.send_group_markdown_msg if use_markdown else self.bot.send_group_text_msg
+            await send(
+                group_openid=target_id,
+                content=content,
+                msg_id=msg_id,
+                event_id=event_id,
+                msg_seq=msg_seq,
+            )
+        else:
+            raise ValueError(f'Unsupported QQ Official text reply target: {target_type}')
+
     def register_listener(
         self,
         event_type: typing.Type[platform_events.Event],
@@ -650,13 +678,13 @@ class QQOfficialAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter
             # 用第一个 chunk 的文本建立会话（不发 "..." 避免污染前缀）
             ctx['session_started'] = True
 
-        # 发送内容 = 全量累积文本
+        # `replace` mode requires every update to contain the previously
-        # QQ API 的 replace 模式不允许修改已下发前缀，所以：
+        # delivered content as its prefix. `sent_length` only tells us whether
-        # - 首次：发送全部文本，建立会话
+        # a non-final snapshot has new content; it must not truncate the
-        # - 后续：只能发送新增部分（append 行为）
+        # content sent to QQ.
-        content_to_send = ctx['accumulated_text'][ctx['sent_length'] :]
+        if len(ctx['accumulated_text']) <= ctx['sent_length'] and not is_final:
-        if not content_to_send and not is_final:
             return
+        content_to_send = ctx['accumulated_text']
 
         input_state = 10 if is_final else 1
 
@@ -778,20 +806,13 @@ class QQOfficialAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter
             return
 
         try:
-            if target_type == 'c2c':
+            await self._send_c2c_or_group_text_reply(
-                await self.bot.send_private_text_msg(
+                target_type,
-                    user_openid=target_id,
+                target_id,
-                    content=text,
+                text,
-                    event_id=event_id,
+                event_id=event_id,
-                    msg_seq=msg_seq,
+                msg_seq=msg_seq,
-                )
+            )
-            elif target_type == 'group':
-                await self.bot.send_group_text_msg(
-                    group_openid=target_id,
-                    content=text,
-                    event_id=event_id,
-                    msg_seq=msg_seq,
-                )
         except Exception:
             await self.logger.error(f'QQ Official: synthetic reply delivery failed: {traceback.format_exc()}')
 

src/langbot/pkg/platform/sources/qqofficial.yaml

@@ -95,6 +95,18 @@ spec:
       type: boolean
       required: true
       default: false
+    - name: enable-markdown-rendering
+      label:
+        en_US: Enable Markdown Rendering
+        zh_Hans: 启用 Markdown 渲染
+        zh_Hant: 啟用 Markdown 渲染
+      description:
+        en_US: Render non-stream C2C and QQ group text replies as Markdown. Channel messages always use plain text and are not affected by this setting.
+        zh_Hans: 将非流式 C2C 私聊和 QQ 群聊文本回复渲染为 Markdown。频道消息始终以纯文本发送，不受此设置影响。
+        zh_Hant: 將非串流 C2C 私聊與 QQ 群聊文字回覆渲染為 Markdown。頻道訊息一律以純文字傳送，不受此設定影響。
+      type: boolean
+      required: true
+      default: false
     - name: webhook_url
       label:
         en_US: Webhook Callback URL

src/langbot/pkg/utils/safe_regex.py

@@ -27,10 +27,16 @@ class SafeRegexTimeoutError(SafeRegexError):
     """Raised when the regex engine exhausts the operation CPU budget."""
 
 
-def _validate_patterns(patterns: Sequence[str]) -> tuple[str, ...]:
+def _validate_patterns(
+    patterns: Sequence[str],
+    *,
+    max_pattern_count: int = MAX_PATTERN_COUNT,
+) -> tuple[str, ...]:
+    if max_pattern_count < 1:
+        raise ValueError('max_pattern_count must be positive')
+    if len(patterns) > max_pattern_count:
+        raise SafeRegexLimitError(f'At most {max_pattern_count} regex patterns are allowed')
     normalized = tuple(patterns)
-    if len(normalized) > MAX_PATTERN_COUNT:
-        raise SafeRegexLimitError(f'At most {MAX_PATTERN_COUNT} regex patterns are allowed')
     for pattern in normalized:
         if not isinstance(pattern, str):
             raise SafeRegexError('Regex patterns must be strings')
@@ -115,8 +121,9 @@ def _mask_patterns_sync(
     mask: str,
     mask_word: str,
     timeout_seconds: float,
+    max_pattern_count: int,
 ) -> tuple[bool, str]:
-    normalized_patterns = _validate_patterns(patterns)
+    normalized_patterns = _validate_patterns(patterns, max_pattern_count=max_pattern_count)
     _validate_input(value)
     if len(mask) > MAX_REPLACEMENT_CHARS or len(mask_word) > MAX_REPLACEMENT_CHARS:
         raise SafeRegexLimitError(f'Regex replacements may contain at most {MAX_REPLACEMENT_CHARS} characters')
@@ -162,6 +169,7 @@ async def mask_patterns(
     mask: str,
     mask_word: str,
     timeout_seconds: float = DEFAULT_OPERATION_TIMEOUT_SECONDS,
+    max_pattern_count: int = MAX_PATTERN_COUNT,
 ) -> tuple[bool, str]:
     """Apply untrusted masking patterns with bounded CPU and output growth."""
 
@@ -174,4 +182,5 @@ async def mask_patterns(
         mask=mask,
         mask_word=mask_word,
         timeout_seconds=timeout_seconds,
+        max_pattern_count=max_pattern_count,
     )

src/langbot/pkg/workspace/invitation_delivery.py

@@ -240,7 +240,7 @@ class InvitationDeliveryService:
     @staticmethod
     def _plain_text(workspace_name: str, invitation_link: str) -> str:
         return (
-            'You have been invited to LangBot Cloud\n\n'
+            'You have been invited to join a Workspace in LangBot\n\n'
             f'Join the Workspace “{workspace_name}” to collaborate with your team.\n\n'
             f'Accept invitation: {invitation_link}\n\n'
             'This secure invitation expires in 7 days and can only be accepted by the email address '
@@ -258,30 +258,77 @@ class InvitationDeliveryService:
 <head>
   <meta charset="utf-8">
   <meta name="viewport" content="width=device-width,initial-scale=1">
-  <title>Join {escaped_workspace} on LangBot Cloud</title>
+  <meta http-equiv="X-UA-Compatible" content="IE=edge">
+  <title>Join {escaped_workspace} in LangBot</title>
 </head>
-<body style="margin:0;background:#f4f7fb;color:#152033;font-family:Inter,-apple-system,BlinkMacSystemFont,'Segoe UI',sans-serif;">
+<body style="margin:0;padding:0;background:#f4f7fb;color:#111827;font-family:Arial,'Helvetica Neue',sans-serif;">
-  <div style="display:none;max-height:0;overflow:hidden;opacity:0;">You have been invited to join {escaped_workspace} on LangBot Cloud.</div>
+  <div style="display:none;max-height:0;overflow:hidden;opacity:0;">You have been invited to join {escaped_workspace} in LangBot.</div>
-  <table role="presentation" width="100%" cellspacing="0" cellpadding="0" style="background:#f4f7fb;padding:40px 16px;">
+  <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0" style="width:100%;background:#f4f7fb;">
-    <tr><td align="center">
+    <tr>
-      <table role="presentation" width="100%" cellspacing="0" cellpadding="0" style="max-width:600px;background:#ffffff;border:1px solid #e5eaf2;border-radius:16px;overflow:hidden;box-shadow:0 12px 32px rgba(20,49,93,.08);">
+      <td align="center" style="padding:48px 16px;">
-        <tr><td style="padding:28px 36px;background:linear-gradient(135deg,#0f172a,#1d4ed8);color:#ffffff;">
+        <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0" style="width:100%;max-width:600px;">
-          <div style="font-size:14px;font-weight:700;letter-spacing:.08em;text-transform:uppercase;opacity:.78;">LangBot Cloud</div>
+          <tr>
-          <div style="font-size:26px;font-weight:700;margin-top:8px;line-height:1.25;">You’re invited</div>
+            <td style="padding:0 4px 20px;">
-        </td></tr>
+              <img src="https://docs.langbot.app/langbot-logo.png" alt="LangBot" width="34" height="34" style="display:inline-block;width:34px;height:34px;border:0;vertical-align:middle;">
-        <tr><td style="padding:36px;">
+              <span style="display:inline-block;margin-left:10px;vertical-align:middle;font-size:18px;font-weight:700;letter-spacing:-.01em;">LangBot</span>
-          <p style="margin:0 0 18px;font-size:16px;line-height:1.65;color:#475569;">You have been invited to collaborate in this Workspace:</p>
+            </td>
-          <div style="margin:0 0 26px;padding:18px 20px;background:#f8fafc;border:1px solid #e2e8f0;border-radius:12px;font-size:18px;font-weight:700;color:#0f172a;">{escaped_workspace}</div>
+          </tr>
-          <table role="presentation" cellspacing="0" cellpadding="0"><tr><td style="border-radius:9px;background:#2563eb;">
+          <tr>
-            <a href="{escaped_link}" style="display:inline-block;padding:13px 22px;color:#ffffff;text-decoration:none;font-size:15px;font-weight:700;">Accept invitation</a>
+            <td style="background:#ffffff;border-radius:10px;overflow:hidden;">
-          </td></tr></table>
+              <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0">
-          <p style="margin:26px 0 8px;font-size:14px;line-height:1.6;color:#64748b;">This invitation expires in 7 days and is bound to the email address that received it.</p>
+                <tr>
-          <p style="margin:0 0 8px;font-size:13px;line-height:1.6;color:#94a3b8;">If the button does not work, copy and paste this URL into your browser:</p>
+                  <td style="padding:42px 42px 38px;">
-          <p style="margin:0;padding:12px;background:#f8fafc;border-radius:8px;word-break:break-all;font-size:12px;line-height:1.55;color:#475569;">{escaped_link}</p>
+                    <div style="margin:0 0 12px;font-size:13px;line-height:1.4;font-weight:600;color:#5f6f84;">Workspace invitation</div>
-        </td></tr>
+                    <h1 style="margin:0 0 16px;font-size:28px;line-height:1.25;font-weight:700;letter-spacing:-.025em;color:#111827;">You’re invited to collaborate</h1>
-        <tr><td style="padding:20px 36px;border-top:1px solid #eef2f7;font-size:12px;line-height:1.6;color:#94a3b8;">If you were not expecting this invitation, you can safely ignore this email.</td></tr>
+                    <p style="margin:0 0 28px;font-size:15px;line-height:1.7;color:#526173;">Join your team in LangBot and start building together in this Workspace.</p>
-      </table>
+
-    </td></tr>
+                    <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0" style="background:#f6f8fb;border-radius:8px;">
+                      <tr>
+                        <td style="padding:16px 18px;">
+                          <div style="margin:0 0 4px;font-size:11px;line-height:1.4;font-weight:700;letter-spacing:.08em;text-transform:uppercase;color:#5f6f84;">Workspace</div>
+                          <div style="font-size:18px;line-height:1.4;font-weight:700;color:#111827;">{escaped_workspace}</div>
+                        </td>
+                      </tr>
+                    </table>
+
+                    <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0">
+                      <tr><td height="28" style="height:28px;font-size:0;line-height:0;">&nbsp;</td></tr>
+                    </table>
+
+                    <table role="presentation" cellspacing="0" cellpadding="0" border="0">
+                      <tr>
+                        <td style="background:#2563eb;border-radius:8px;">
+                          <a href="{escaped_link}" target="_blank" style="display:inline-block;padding:13px 22px;font-size:15px;line-height:1.2;font-weight:700;color:#ffffff;text-decoration:none;border-radius:8px;">Accept invitation</a>
+                        </td>
+                      </tr>
+                    </table>
+
+                    <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0">
+                      <tr><td height="32" style="height:32px;font-size:0;line-height:0;">&nbsp;</td></tr>
+                    </table>
+
+                    <table role="presentation" width="100%" cellspacing="0" cellpadding="0" border="0" style="border-top:1px solid #e8edf4;">
+                      <tr>
+                        <td style="padding-top:22px;">
+                          <p style="margin:0 0 10px;font-size:13px;line-height:1.6;color:#5f6f84;">For your security, this invitation expires in 7 days and only works for the email address that received it.</p>
+                          <a href="{escaped_link}" target="_blank" style="font-size:13px;line-height:1.6;font-weight:600;color:#2563eb;text-decoration:none;">Open invitation link&nbsp;&rarr;</a>
+                        </td>
+                      </tr>
+                    </table>
+                  </td>
+                </tr>
+              </table>
+            </td>
+          </tr>
+          <tr>
+            <td align="center" style="padding:20px 24px 0;font-size:12px;line-height:1.6;color:#5f6f84;">
+              Sent by LangBot<br>
+              If you were not expecting this invitation, you can safely ignore this email.
+            </td>
+          </tr>
+        </table>
+      </td>
+    </tr>
   </table>
 </body>
 </html>'''

tests/unit_tests/pipeline/test_banwords.py

@@ -0,0 +1,113 @@
+"""BanWordFilter regression tests for legacy sensitive-word lists.
+
+v4.10.7 introduced a 64-pattern cap in safe_regex. Older installs still carry
+the previous default list (~70 patterns). The filter must keep applying those
+rules instead of blocking every message.
+"""
+
+from __future__ import annotations
+
+from importlib import import_module
+from unittest.mock import Mock
+
+import pytest
+
+from tests.factories import FakeApp
+
+
+def _load_banwords():
+    import_module('langbot.pkg.pipeline.pipelinemgr')
+    banwords = import_module('langbot.pkg.pipeline.cntfilter.filters.banwords')
+    entities = import_module('langbot.pkg.pipeline.cntfilter.entities')
+    safe_regex = import_module('langbot.pkg.utils.safe_regex')
+    return banwords, entities, safe_regex
+
+
+def _filter_with_words(words: list[str], *, mask: str = '*', mask_word: str = ''):
+    banwords, entities, _ = _load_banwords()
+    app = FakeApp()
+    app.sensitive_meta = Mock()
+    app.sensitive_meta.data = {
+        'words': words,
+        'mask': mask,
+        'mask_word': mask_word,
+    }
+    return banwords.BanWordFilter(app), entities, app
+
+
+@pytest.mark.asyncio
+async def test_legacy_word_list_over_pattern_cap_does_not_block_clean_message():
+    """A pre-v4.10.7 word list must not fail closed on every message."""
+    _, _, safe_regex = _load_banwords()
+    words = [f'word{i}' for i in range(safe_regex.MAX_PATTERN_COUNT + 6)]
+    filt, entities, _ = _filter_with_words(words)
+
+    result = await filt.process(Mock(), 'hello there, nothing banned')
+
+    assert result.level == entities.ResultLevel.PASS
+    assert result.replacement == 'hello there, nothing banned'
+    assert result.user_notice == ''
+
+
+@pytest.mark.asyncio
+async def test_legacy_word_list_still_masks_match_beyond_first_batch():
+    """Words past the first 64-pattern batch must still be applied."""
+    _, _, safe_regex = _load_banwords()
+    words = [f'word{i}' for i in range(safe_regex.MAX_PATTERN_COUNT)] + ['secret-token']
+    filt, entities, _ = _filter_with_words(words, mask_word='[hidden]')
+
+    result = await filt.process(Mock(), 'please hide secret-token now')
+
+    assert result.level == entities.ResultLevel.MASKED
+    assert 'secret-token' not in result.replacement
+    assert '[hidden]' in result.replacement
+
+
+@pytest.mark.asyncio
+async def test_legacy_word_list_masks_match_in_first_batch():
+    _, _, safe_regex = _load_banwords()
+    words = ['alpha-secret'] + [f'word{i}' for i in range(safe_regex.MAX_PATTERN_COUNT)]
+    filt, entities, _ = _filter_with_words(words, mask_word='[hidden]')
+
+    result = await filt.process(Mock(), 'alpha-secret is here')
+
+    assert result.level == entities.ResultLevel.MASKED
+    assert result.replacement == '[hidden] is here'
+
+
+@pytest.mark.asyncio
+async def test_invalid_sensitive_word_regex_still_blocks():
+    filt, entities, _ = _filter_with_words(['(unclosed'])
+
+    result = await filt.process(Mock(), 'any message')
+
+    assert result.level == entities.ResultLevel.BLOCK
+    assert result.user_notice == '内容检查规则执行失败，请联系管理员'
+    assert 'rejected' in result.console_notice.lower() or 'invalid' in result.console_notice.lower()
+
+
+@pytest.mark.asyncio
+async def test_oversized_word_list_is_blocked():
+    """Configured rules must never be silently skipped when the list is oversized."""
+    banwords, _, _ = _load_banwords()
+    words = [f'word{i}' for i in range(banwords._MAX_SENSITIVE_WORD_PATTERNS + 10)]
+    filt, entities, _ = _filter_with_words(words)
+
+    result = await filt.process(Mock(), 'hello there, nothing banned')
+
+    assert result.level == entities.ResultLevel.BLOCK
+    assert result.replacement == ''
+    assert result.user_notice == '内容检查规则执行失败，请联系管理员'
+    assert 'at most 256 regex patterns are allowed' in result.console_notice.lower()
+
+
+@pytest.mark.asyncio
+async def test_match_beyond_total_cap_cannot_bypass_filter():
+    banwords, _, _ = _load_banwords()
+    words = [f'word{i}' for i in range(banwords._MAX_SENSITIVE_WORD_PATTERNS)] + ['late-secret']
+    filt, entities, _ = _filter_with_words(words, mask_word='[hidden]')
+
+    result = await filt.process(Mock(), 'please hide late-secret now')
+
+    assert result.level == entities.ResultLevel.BLOCK
+    assert result.replacement == ''

tests/unit_tests/platform/test_line_session_identity.py

@@ -0,0 +1,69 @@
+from __future__ import annotations
+
+import pytest
+from unittest.mock import MagicMock
+
+from linebot.v3.webhooks import TextMessageContent
+
+from langbot.pkg.platform import botmgr as _botmgr  # noqa: F401
+from langbot.pkg.platform.sources import line
+
+
+def _make_event(*, source_type: str, user_id, group_id=None, room_id=None, message_id: str, text: str = 'hi'):
+    event = MagicMock()
+    event.timestamp = 1700000000000
+    event.message = MagicMock(spec=TextMessageContent)
+    event.message.id = message_id
+    event.message.text = text
+    event.message.webhook_event_id = f'webhook-{message_id}'
+    event.message.timestamp = event.timestamp
+
+    source = MagicMock()
+    source.type = source_type
+    source.user_id = user_id
+    if group_id is not None:
+        source.group_id = group_id
+    if room_id is not None:
+        source.room_id = room_id
+    event.source = source
+
+    return event
+
+
+@pytest.mark.asyncio
+async def test_user_message_launcher_id_stable_across_messages() -> None:
+    """Two distinct messages from the same LINE user must resolve to the same
+    sender id, otherwise every message starts a brand new session (context loss).
+    """
+    event1 = _make_event(source_type='user', user_id='U-stable-user', message_id='msg-1')
+    event2 = _make_event(source_type='user', user_id='U-stable-user', message_id='msg-2')
+
+    result1 = await line.LINEEventConverter.target2yiri(event1, bot_client=None)
+    result2 = await line.LINEEventConverter.target2yiri(event2, bot_client=None)
+
+    assert result1.sender.id == 'U-stable-user'
+    assert result1.sender.id == result2.sender.id
+    assert result1.sender.id != event1.message.id
+
+
+@pytest.mark.asyncio
+async def test_group_message_uses_group_id_not_message_id() -> None:
+    event1 = _make_event(source_type='group', user_id='U-member', group_id='G-stable-group', message_id='msg-1')
+    event2 = _make_event(source_type='group', user_id='U-member', group_id='G-stable-group', message_id='msg-2')
+
+    result1 = await line.LINEEventConverter.target2yiri(event1, bot_client=None)
+    result2 = await line.LINEEventConverter.target2yiri(event2, bot_client=None)
+
+    assert result1.sender.group.id == 'G-stable-group'
+    assert result1.sender.group.id == result2.sender.group.id
+    assert result1.sender.id == 'U-member'
+
+
+@pytest.mark.asyncio
+async def test_room_message_uses_room_id_and_falls_back_when_user_id_missing() -> None:
+    event = _make_event(source_type='room', user_id=None, room_id='R-stable-room', message_id='msg-1')
+
+    result = await line.LINEEventConverter.target2yiri(event, bot_client=None)
+
+    assert result.sender.group.id == 'R-stable-room'
+    assert result.sender.id == 'R-stable-room'

tests/unit_tests/platform/test_qqofficial_api.py

@@ -1,9 +1,11 @@
-"""Tests for QQ Official keyboard payload helpers."""
+"""Tests for QQ Official message and keyboard payload helpers."""
 
 import asyncio
+import json
 import time
 from unittest.mock import AsyncMock, MagicMock, patch
 
+import httpx
 import pytest
 
 import langbot_plugin.api.entities.builtin.platform.message as platform_message
@@ -99,6 +101,12 @@ def _stream_test_adapter():
     adapter.bot = MagicMock()
     adapter.bot.send_stream_msg = AsyncMock(return_value={'id': 'stream-1'})
     adapter.bot.send_markdown_keyboard = AsyncMock(return_value={'id': 'message-1'})
+    adapter.bot.send_private_text_msg = AsyncMock()
+    adapter.bot.send_group_text_msg = AsyncMock()
+    adapter.bot.send_private_markdown_msg = AsyncMock()
+    adapter.bot.send_group_markdown_msg = AsyncMock()
+    adapter.bot.send_channle_group_text_msg = AsyncMock()
+    adapter.bot.send_channle_private_text_msg = AsyncMock()
     adapter.ap = None
     adapter._stream_ctx = {}
     adapter._stream_ctx_ts = {}
@@ -108,7 +116,7 @@ def _stream_test_adapter():
 
 
 @pytest.mark.asyncio
-async def test_qq_stream_uses_cumulative_chunks_as_snapshots():
+async def test_qq_stream_replace_mode_sends_complete_snapshots():
     adapter = _stream_test_adapter()
     adapter._stream_ctx['message-1'] = {
         'user_openid': 'user-1',
@@ -138,10 +146,109 @@ async def test_qq_stream_uses_cumulative_chunks_as_snapshots():
 
     assert [call.kwargs['content'] for call in adapter.bot.send_stream_msg.await_args_list] == [
         '<think>one',
-        ' two',
+        '<think>one two',
     ]
 
 
+@pytest.mark.asyncio
+async def test_qq_markdown_messages_use_markdown_payloads():
+    requests = []
+
+    def capture_request(request: httpx.Request) -> httpx.Response:
+        requests.append((str(request.url), json.loads(request.content)))
+        return httpx.Response(200, json={})
+
+    client = QQOfficialClient('secret', 'token', 'app-id', AsyncMock())
+    client.access_token = 'access-token'
+    client.access_token_expiry_time = time.time() + 3600
+    client._http_clients[None] = httpx.AsyncClient(transport=httpx.MockTransport(capture_request))
+
+    try:
+        await client.send_private_markdown_msg('user-1', '# Hello', msg_id='message-1', msg_seq=2)
+        await client.send_group_markdown_msg('group-1', '* Hello', event_id='event-1', msg_seq=3)
+    finally:
+        await client.close()
+
+    assert requests == [
+        (
+            'https://api.sgroup.qq.com/v2/users/user-1/messages',
+            {'msg_type': 2, 'markdown': {'content': '# Hello'}, 'msg_seq': 2, 'msg_id': 'message-1'},
+        ),
+        (
+            'https://api.sgroup.qq.com/v2/groups/group-1/messages',
+            {'msg_type': 2, 'markdown': {'content': '* Hello'}, 'msg_seq': 3, 'event_id': 'event-1'},
+        ),
+    ]
+
+
+@pytest.mark.asyncio
+async def test_qq_markdown_rendering_switches_c2c_and_group_text_replies():
+    adapter = _stream_test_adapter()
+    adapter.config = {'enable-markdown-rendering': True}
+
+    await adapter._send_c2c_or_group_text_reply('c2c', 'user-1', '# Hello', msg_id='message-1')
+    await adapter._send_c2c_or_group_text_reply('group', 'group-1', '* Hello', event_id='event-1')
+
+    adapter.bot.send_private_markdown_msg.assert_awaited_once_with(
+        user_openid='user-1',
+        content='# Hello',
+        msg_id='message-1',
+        event_id=None,
+        msg_seq=1,
+    )
+    adapter.bot.send_group_markdown_msg.assert_awaited_once_with(
+        group_openid='group-1',
+        content='* Hello',
+        msg_id=None,
+        event_id='event-1',
+        msg_seq=1,
+    )
+    adapter.bot.send_private_text_msg.assert_not_awaited()
+    adapter.bot.send_group_text_msg.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_qq_markdown_rendering_defaults_to_plain_text_replies():
+    adapter = _stream_test_adapter()
+    adapter.config = {}
+
+    await adapter._send_c2c_or_group_text_reply('c2c', 'user-1', 'Hello')
+    await adapter._send_c2c_or_group_text_reply('group', 'group-1', 'Hello')
+
+    adapter.bot.send_private_text_msg.assert_awaited_once()
+    adapter.bot.send_group_text_msg.assert_awaited_once()
+    adapter.bot.send_private_markdown_msg.assert_not_awaited()
+    adapter.bot.send_group_markdown_msg.assert_not_awaited()
+
+
+@pytest.mark.asyncio
+async def test_qq_markdown_rendering_does_not_affect_channel_messages():
+    adapter = _stream_test_adapter()
+    adapter.config = {'enable-markdown-rendering': True}
+    message = platform_message.MessageChain([platform_message.Plain(text='# Hello')])
+
+    channel_source = MagicMock()
+    channel_source.t = 'AT_MESSAGE_CREATE'
+    channel_source.channel_id = 'channel-1'
+    channel_source.d_id = 'message-1'
+    channel_event = MagicMock()
+    channel_event.source_platform_object = channel_source
+    await adapter.reply_message(channel_event, message)
+
+    dm_source = MagicMock()
+    dm_source.t = 'DIRECT_MESSAGE_CREATE'
+    dm_source.guild_id = 'guild-1'
+    dm_source.d_id = 'message-2'
+    dm_event = MagicMock()
+    dm_event.source_platform_object = dm_source
+    await adapter.reply_message(dm_event, message)
+
+    adapter.bot.send_channle_group_text_msg.assert_awaited_once_with('channel-1', '# Hello', 'message-1')
+    adapter.bot.send_channle_private_text_msg.assert_awaited_once_with('guild-1', '# Hello', 'message-2')
+    adapter.bot.send_private_markdown_msg.assert_not_awaited()
+    adapter.bot.send_group_markdown_msg.assert_not_awaited()
+
+
 @pytest.mark.asyncio
 async def test_qq_non_streaming_fallback_keeps_latest_snapshot_only():
     from langbot.pkg.platform.sources.qqofficial import QQOfficialAdapter

tests/unit_tests/utils/test_safe_regex.py

@@ -53,6 +53,45 @@ async def test_matches_any_rejects_pattern_and_input_amplification():
         )
 
 
+@pytest.mark.asyncio
+async def test_mask_patterns_honors_explicit_pattern_count_cap():
+    patterns = ['a'] * (safe_regex.MAX_PATTERN_COUNT + 6)
+    found, masked = await safe_regex.mask_patterns(
+        patterns,
+        'hello',
+        mask='*',
+        mask_word='',
+        max_pattern_count=len(patterns),
+    )
+    assert found is False
+    assert masked == 'hello'
+
+    with pytest.raises(safe_regex.SafeRegexLimitError):
+        await safe_regex.mask_patterns(
+            patterns,
+            'hello',
+            mask='*',
+            mask_word='',
+        )
+
+
+@pytest.mark.asyncio
+async def test_mask_patterns_rejects_oversized_sequence_before_copying_it():
+    class OversizedPatterns(list):
+        def __iter__(self):
+            raise AssertionError('oversized patterns must not be materialized')
+
+    patterns = OversizedPatterns(['a'] * (safe_regex.MAX_PATTERN_COUNT + 1))
+
+    with pytest.raises(safe_regex.SafeRegexLimitError):
+        await safe_regex.mask_patterns(
+            patterns,
+            'hello',
+            mask='*',
+            mask_word='',
+        )
+
+
 @pytest.mark.asyncio
 async def test_mask_patterns_bounds_replacement_growth_and_masks_matches():
     found, masked = await safe_regex.mask_patterns(

tests/unit_tests/workspace/test_invitation_delivery.py

@@ -88,14 +88,15 @@ async def test_environment_mapping_enables_provider_without_leaking_secret(monke
     assert service.capability() == {'enabled': True, 'provider': 'smtp'}
 
 
-async def test_cloud_invitation_email_has_branded_html_plain_fallback_and_expiry_copy():
+async def test_invitation_email_has_generic_langbot_brand_plain_fallback_and_expiry_copy():
     service = InvitationDeliveryService(_app({}))
     link = 'https://cloud.langbot.app/invitations/accept#token=lbi_secret&next=<unsafe>'
 
     text = service._plain_text('Research & Development', link)
     html = service._html('Research & Development', link)
 
-    assert 'LangBot Cloud' in text
+    assert 'LangBot' in text
+    assert 'LangBot Cloud' not in text
     assert 'Research & Development' in text
     assert '7 days' in text
     assert link in text
@@ -103,3 +104,55 @@ async def test_cloud_invitation_email_has_branded_html_plain_fallback_and_expiry
     assert 'Research &amp; Development' in html
     assert 'expires in 7 days' in html
     assert 'lbi_secret&amp;next=&lt;unsafe&gt;' in html
+    assert 'LangBot Cloud' not in html
+
+
+async def test_invitation_email_uses_quiet_brand_lockup_and_compact_fallback_link():
+    service = InvitationDeliveryService(_app({}))
+    link = 'https://cloud.langbot.app/invitations/accept#token=lbi_secret'
+
+    html = service._html("RockChinQ's Workspace", link)
+
+    assert 'https://docs.langbot.app/langbot-logo.png' in html
+    assert '>LangBot<' in html
+    assert 'Workspace invitation' in html
+    assert 'Open invitation link' in html
+    assert 'linear-gradient' not in html
+    assert 'box-shadow' not in html
+    assert 'border-top:4px solid' not in html
+    assert 'border:1px solid #dfe6f0' not in html
+    assert 'height="28"' in html
+    assert 'height="32"' in html
+    assert 'margin-top:32px' not in html
+    assert f'>{link}<' not in html
+
+
+async def test_oss_smtp_configuration_delivers_the_generic_invitation_email():
+    service = InvitationDeliveryService(
+        _app(
+            {
+                'workspace': {
+                    'invitations': {
+                        'email': {
+                            'provider': 'smtp',
+                            'from': 'LangBot <noreply@example.com>',
+                            'smtp': {'host': 'smtp.example.com'},
+                        }
+                    }
+                }
+            }
+        )
+    )
+    service._send_smtp = AsyncMock(return_value=True)
+    link = 'https://self-hosted.example/invitations/accept#token=lbi_secret'
+
+    result = await service.deliver_invitation(
+        recipient_email='member@example.com',
+        workspace_name='Self-hosted Workspace',
+        invitation_link=link,
+    )
+
+    assert result == InvitationDeliveryResult(status='sent', provider='smtp')
+    service._send_smtp.assert_awaited_once()
+    assert 'LangBot Cloud' not in service._plain_text('Self-hosted Workspace', link)
+    assert 'LangBot Cloud' not in service._html('Self-hosted Workspace', link)

web/src/app/infra/http/BackendClient.ts

@@ -710,11 +710,32 @@ export class BackendClient extends BaseHttpClient {
     );
   }
 
-  private async getAuthenticatedObjectURL(path: string): Promise<string> {
+  private async getAuthenticatedObjectURL(
+    path: string,
+    rewritePluginPageSdk = false,
+  ): Promise<string> {
     const response = await this.instance.get<Blob>(path, {
       responseType: 'blob',
     });
-    return URL.createObjectURL(response.data);
+    let blob = response.data;
+    if (rewritePluginPageSdk && blob.type.startsWith('text/html')) {
+      const apiBase =
+        this.instance.defaults.baseURL === '/'
+          ? window.location.origin
+          : this.instance.defaults.baseURL?.replace(/\/$/, '');
+      const pageSdkUrl = `${apiBase}/api/v1/plugins/_sdk/page-sdk.js`;
+      const html = await blob.text();
+      blob = new Blob(
+        [
+          html.replace(
+            /(<script\b[^>]*\bsrc\s*=\s*)(["'])\/api\/v1\/plugins\/_sdk\/page-sdk\.js\2/gi,
+            `$1$2${pageSdkUrl}$2`,
+          ),
+        ],
+        { type: blob.type },
+      );
+    }
+    return URL.createObjectURL(blob);
   }
 
   public getAuthenticatedPluginAssetURL(
@@ -724,6 +745,7 @@ export class BackendClient extends BaseHttpClient {
   ): Promise<string> {
     return this.getAuthenticatedObjectURL(
       `/api/v1/plugins/${author}/${name}/authenticated-assets/${filepath}`,
+      true,
     );
   }
 

web/tests/e2e/plugin-page-auth.spec.ts

@@ -69,7 +69,55 @@ test('loads a Cloud plugin page through the authenticated asset route', async ({
       await route.fulfill({
         status: 200,
         contentType: 'text/html',
-        body: '<!doctype html><html><body><h1>LangRAG Observability</h1></body></html>',
+        body: `<!doctype html>
+          <html>
+            <body>
+              <h1>LangRAG Observability</h1>
+              <button id="save">Save</button>
+              <script src="/api/v1/plugins/_sdk/page-sdk.js"></script>
+              <script>
+                document.querySelector('#save').addEventListener('click', async () => {
+                  await window.langbot.api('/settings', { enabled: true }, 'POST');
+                  document.body.dataset.saved = 'true';
+                });
+              </script>
+            </body>
+          </html>`,
+      });
+    },
+  );
+  let pageSdkRequests = 0;
+  await page.route('**/api/v1/plugins/_sdk/page-sdk.js', async (route) => {
+    pageSdkRequests += 1;
+    await route.fulfill({
+      status: 200,
+      contentType: 'application/javascript',
+      body: `window.langbot = {
+        api(endpoint, body, method) {
+          return new Promise((resolve) => {
+            const requestId = 'request-' + Date.now();
+            const handler = (event) => {
+              if (event.data?.type === 'langbot:api:response' && event.data.requestId === requestId) {
+                window.removeEventListener('message', handler);
+                resolve(event.data.data);
+              }
+            };
+            window.addEventListener('message', handler);
+            window.parent.postMessage({ type: 'langbot:api', requestId, endpoint, body, method }, '*');
+          });
+        },
+      };`,
+    });
+  });
+  let pageApiRequests = 0;
+  await page.route(
+    '**/api/v1/plugins/langbot-team/LangRAG/page-api',
+    async (route) => {
+      pageApiRequests += 1;
+      await route.fulfill({
+        status: 200,
+        contentType: 'application/json',
+        body: wrapped({ saved: true }),
       });
     },
   );
@@ -78,11 +126,17 @@ test('loads a Cloud plugin page through the authenticated asset route', async ({
     '/home/plugin-pages?id=langbot-team%2FLangRAG%2Fobservability',
   );
 
+  const pluginFrame = page.frameLocator('iframe');
   await expect(
-    page
+    pluginFrame.getByRole('heading', { name: 'LangRAG Observability' }),
-      .frameLocator('iframe')
-      .getByRole('heading', { name: 'LangRAG Observability' }),
   ).toBeVisible();
+  await pluginFrame.getByRole('button', { name: 'Save' }).click();
+  await expect(pluginFrame.locator('body')).toHaveAttribute(
+    'data-saved',
+    'true',
+  );
   expect(authenticatedAssetRequests).toBeGreaterThan(0);
+  expect(pageSdkRequests).toBe(1);
+  expect(pageApiRequests).toBe(1);
   await expect(page.getByText('Loading...')).toHaveCount(0);
 });