fix(runner): align SDK pin and complete real runtime verification (#2525)

* fix(runner): align SDK pin and workspace-aware integration fixtures

* fix(ci): format sources and resolve current migration head

* test(persistence): align standalone migration fixtures with current models

* test(web): align smoke fixtures with current processor UI

---------

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

pyproject.toml

@@ -232,4 +232,4 @@ line-ending = "auto"
 
 [tool.uv.sources]
 # Development contract: update to the matching SDK release before publishing.
-langbot-plugin = { git = "https://github.com/langbot-app/langbot-plugin-sdk", rev = "eac2c60509534512f9a373cd0c801e75985e0612" }
+langbot-plugin = { git = "https://github.com/langbot-app/langbot-plugin-sdk", rev = "92a9e03fa9c791f4ed30cc3f5f0602c13b800d28" }

src/langbot/libs/wecom_ai_bot_api/api.py

@@ -17,6 +17,7 @@ from quart import Quart, request, Response, jsonify
 
 from langbot.libs.wecom_ai_bot_api import wecombotevent
 from langbot.libs.wecom_ai_bot_api.WXBizMsgCrypt3 import WXBizMsgCrypt
+
 if TYPE_CHECKING:
     from langbot.pkg.platform.logger import EventLogger
 from langbot.pkg.utils import httpclient

src/langbot/pkg/agent/runner/event_log_store.py

@@ -1,4 +1,5 @@
 """EventLog store for writing and querying event records."""
+
 from __future__ import annotations
 
 import json
@@ -44,9 +45,7 @@ class EventLogStore:
 
     def __init__(self, engine: AsyncEngine):
         self.engine = engine
-        self._session_factory = sessionmaker(
+        self._session_factory = sessionmaker(engine, class_=AsyncSession, expire_on_commit=False)
-            engine, class_=AsyncSession, expire_on_commit=False
-        )
 
     async def append_event(
         self,
@@ -101,7 +100,7 @@ class EventLogStore:
 
         # Truncate input summary if too long
         if input_summary and len(input_summary) > self.MAX_INPUT_SUMMARY_LENGTH:
-            input_summary = input_summary[:self.MAX_INPUT_SUMMARY_LENGTH - 3] + "..."
+            input_summary = input_summary[: self.MAX_INPUT_SUMMARY_LENGTH - 3] + '...'
 
         async with self._session_factory() as session:
             event = EventLog(
@@ -144,9 +143,7 @@ class EventLogStore:
             Event record as dict, or None if not found
         """
         async with self._session_factory() as session:
-            result = await session.execute(
+            result = await session.execute(sqlalchemy.select(EventLog).where(EventLog.event_id == event_id))
-                sqlalchemy.select(EventLog).where(EventLog.event_id == event_id)
-            )
             row = result.scalars().first()
             if row is None:
                 return None
@@ -282,9 +279,7 @@ class EventLogStore:
     ) -> int:
         """Delete EventLog rows created before the supplied timestamp."""
         async with self._session_factory() as session:
-            result = await session.execute(
+            result = await session.execute(sqlalchemy.delete(EventLog).where(EventLog.created_at < before))
-                sqlalchemy.delete(EventLog).where(EventLog.created_at < before)
-            )
             await session.commit()
             return result.rowcount or 0
 

src/langbot/pkg/api/http/controller/groups/pipelines/embed.py

@@ -71,17 +71,8 @@ class EmbedRouterGroup(group.RouterGroup):
         ``web_page_bot``, is disabled, or has no Pipeline target for messages.
         """
         bot = await self.ap.platform_mgr.resolve_public_bot(bot_uuid)
-        pipeline_uuid = (
+        pipeline_uuid = bot.get_pipeline_target_for_event_type('message.received') if bot is not None else None
-            bot.get_pipeline_target_for_event_type('message.received')
+        if bot is not None and bot.bot_entity.adapter == 'web_page_bot' and bot.bot_entity.enable and pipeline_uuid:
-            if bot is not None
-            else None
-        )
-        if (
-            bot is not None
-            and bot.bot_entity.adapter == 'web_page_bot'
-            and bot.bot_entity.enable
-            and pipeline_uuid
-        ):
             return bot, pipeline_uuid
         return None, None
 

src/langbot/pkg/api/http/controller/groups/resources/tools.py

@@ -14,9 +14,7 @@ class ToolsRouterGroup(group.RouterGroup):
         self,
         request_context: RequestContext,
     ) -> list[dict] | None:
-        pipeline_uuid = quart.request.args.get(
+        pipeline_uuid = quart.request.args.get('pipeline_uuid') or quart.request.args.get('pipeline_id')
-            'pipeline_uuid'
-        ) or quart.request.args.get('pipeline_id')
         bound_plugins: list[str] | None = None
         bound_mcp_servers: list[str] | None = None
 
@@ -28,14 +26,9 @@ class ToolsRouterGroup(group.RouterGroup):
             if pipeline is None:
                 return None
 
-            extensions_prefs = normalize_extension_preferences(
+            extensions_prefs = normalize_extension_preferences(pipeline.get('extensions_preferences'))
-                pipeline.get('extensions_preferences')
-            )
             if not extensions_prefs['enable_all_plugins']:
-                bound_plugins = [
+                bound_plugins = [f'{plugin["author"]}/{plugin["name"]}' for plugin in extensions_prefs['plugins']]
-                    f'{plugin["author"]}/{plugin["name"]}'
-                    for plugin in extensions_prefs['plugins']
-                ]
             if not extensions_prefs['enable_all_mcp_servers']:
                 bound_mcp_servers = extensions_prefs['mcp_servers']
 

src/langbot/pkg/entity/persistence/transcript.py

@@ -1,4 +1,5 @@
 """Transcript persistence entity for conversation history projection."""
+
 from __future__ import annotations
 
 import sqlalchemy

src/langbot/pkg/persistence/alembic/versions/0017_repair_local_workspace_owner.py

@@ -48,12 +48,16 @@ def upgrade() -> None:
         sa.column('source', sa.String(32)),
         sa.column('created_by_account_uuid', sa.String(36)),
     )
-    workspace_uuids = conn.execute(
+    workspace_uuids = (
+        conn.execute(
             sa.select(workspaces.c.uuid).where(
                 workspaces.c.instance_uuid == instance_uuid.strip(),
                 workspaces.c.source == 'local',
             )
-    ).scalars().all()
+        )
+        .scalars()
+        .all()
+    )
     if not workspace_uuids:
         return
     if len(workspace_uuids) > 1:
@@ -99,10 +103,7 @@ def upgrade() -> None:
         sa.column('status', sa.String(32)),
     )
     owner_account_uuid = conn.execute(
-        sa.select(users.c.uuid)
+        sa.select(users.c.uuid).where(users.c.status == 'active').order_by(users.c.id).limit(1)
-        .where(users.c.status == 'active')
-        .order_by(users.c.id)
-        .limit(1)
     ).scalar_one_or_none()
     if owner_account_uuid is None:
         return

src/langbot/pkg/persistence/alembic/versions/58846a8d7a81_add_event_log_and_transcript_tables.py

@@ -4,6 +4,7 @@ Revision ID: 58846a8d7a81
 Revises: 0005_add_llm_context_length
 Create Date: 2026-05-23 15:41:47.030841
 """
+
 from alembic import op
 import sqlalchemy as sa
 

src/langbot/pkg/persistence/alembic/versions/7b2c1d9e4f30_add_transcript_scope_columns.py

@@ -4,6 +4,7 @@ Revision ID: 7b2c1d9e4f30
 Revises: 6dfd3dd7f0c7
 Create Date: 2026-06-12
 """
+
 from alembic import op
 import sqlalchemy as sa
 

src/langbot/pkg/persistence/alembic/versions/8d3a1f2c4b6e_add_agent_run_ledger.py

@@ -90,9 +90,7 @@ def upgrade() -> None:
         )
     else:
         _add_column_if_missing('agent_run', sa.Column('queue_name', sa.String(255), nullable=True))
-        _add_column_if_missing(
+        _add_column_if_missing('agent_run', sa.Column('priority', sa.Integer(), nullable=False, server_default='0'))
-            'agent_run', sa.Column('priority', sa.Integer(), nullable=False, server_default='0')
-        )
         _add_column_if_missing('agent_run', sa.Column('requested_runtime_id', sa.String(255), nullable=True))
         _add_column_if_missing('agent_run', sa.Column('claimed_by_runtime_id', sa.String(255), nullable=True))
         _add_column_if_missing('agent_run', sa.Column('claim_token', sa.String(255), nullable=True))

src/langbot/pkg/pipeline/extension_preferences.py

@@ -50,22 +50,16 @@ def normalize_extension_preferences(value: typing.Any) -> dict[str, typing.Any]:
     normalized['enable_all_plugins'] = value.get('enable_all_plugins', True) is True
     normalized['enable_all_mcp_servers'] = value.get('enable_all_mcp_servers', True) is True
     normalized['enable_all_skills'] = value.get('enable_all_skills', True) is True
-    normalized['mcp_resource_agent_read_enabled'] = (
+    normalized['mcp_resource_agent_read_enabled'] = value.get('mcp_resource_agent_read_enabled', True) is True
-        value.get('mcp_resource_agent_read_enabled', True) is True
-    )
 
     plugins = value.get('plugins', [])
-    plugins_are_valid = isinstance(plugins, list) and all(
+    plugins_are_valid = isinstance(plugins, list) and all(_valid_plugin_binding(plugin) for plugin in plugins)
-        _valid_plugin_binding(plugin) for plugin in plugins
-    )
     normalized['plugins'] = list(plugins) if plugins_are_valid else []
     if not plugins_are_valid:
         normalized['enable_all_plugins'] = False
 
     mcp_servers = value.get('mcp_servers', [])
-    mcp_servers_are_valid = isinstance(mcp_servers, list) and all(
+    mcp_servers_are_valid = isinstance(mcp_servers, list) and all(_valid_name(server) for server in mcp_servers)
-        _valid_name(server) for server in mcp_servers
-    )
     normalized['mcp_servers'] = list(mcp_servers) if mcp_servers_are_valid else []
     if not mcp_servers_are_valid:
         normalized['enable_all_mcp_servers'] = False
@@ -151,6 +145,4 @@ def _validate_list_field(
         raise ValueError(f"{context} field '{field_label}' must be a list")
     for index, item in enumerate(items):
         if not item_validator(item):
-            raise ValueError(
+            raise ValueError(f"{context} field '{field_label}[{index}]' must be {item_description}")
-                f"{context} field '{field_label}[{index}]' must be {item_description}"
-            )

src/langbot/pkg/platform/adapters/officialaccount/message_converter.py

@@ -65,8 +65,12 @@ class OfficialAccountMessageConverter(abstract_platform_adapter.AbstractMessageC
             else:
                 components.append(platform_message.Unknown(text='[officialaccount voice message without media id]'))
         elif event.type == 'event':
-            components.append(platform_message.Unknown(text=f'[officialaccount event: {event.detail_type or "unknown"}]'))
+            components.append(
+                platform_message.Unknown(text=f'[officialaccount event: {event.detail_type or "unknown"}]')
+            )
         else:
-            components.append(platform_message.Unknown(text=f'[unsupported officialaccount msgtype: {event.type or "unknown"}]'))
+            components.append(
+                platform_message.Unknown(text=f'[unsupported officialaccount msgtype: {event.type or "unknown"}]')
+            )
 
         return platform_message.MessageChain(components)

src/langbot/pkg/platform/adapters/qqofficial/__init__.py

@@ -3,4 +3,3 @@
 from langbot.pkg.platform.adapters.qqofficial.adapter import QQOfficialAdapter
 
 __all__ = ['QQOfficialAdapter']
-

src/langbot/pkg/platform/adapters/qqofficial/api_impl.py

@@ -54,7 +54,9 @@ class QQOfficialAPIMixin:
         self,
         group_id: typing.Union[int, str],
     ) -> list[platform_entities.UserGroupMember]:
-        return [member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)]
+        return [
+            member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)
+        ]
 
     async def edit_message(
         self,
@@ -100,4 +102,3 @@ class QQOfficialAPIMixin:
 
     async def leave_group(self, group_id: typing.Union[int, str]):
         raise NotSupportedError('leave_group')
-

src/langbot/pkg/platform/adapters/qqofficial/errors.py

@@ -8,4 +8,3 @@ except ModuleNotFoundError:
         def __init__(self, api_name: str, *args):
             super().__init__(f"API '{api_name}' is not supported by this adapter", *args)
             self.api_name = api_name
-

src/langbot/pkg/platform/adapters/qqofficial/platform_api.py

@@ -34,4 +34,3 @@ PLATFORM_API_MAP: dict[str, typing.Callable[[typing.Any, dict], typing.Awaitable
     'get_gateway_url': get_gateway_url,
     'get_mode': get_mode,
 }
-

src/langbot/pkg/platform/adapters/slack/api_impl.py

@@ -47,7 +47,9 @@ class SlackAPIMixin:
         self,
         group_id: typing.Union[int, str],
     ) -> list[platform_entities.UserGroupMember]:
-        return [member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)]
+        return [
+            member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)
+        ]
 
     async def get_group_member_info(
         self,

src/langbot/pkg/platform/adapters/slack/event_converter.py

@@ -19,7 +19,9 @@ class SlackEventConverter(abstract_platform_adapter.AbstractEventConverter):
     async def yiri2target(event: platform_events.Event) -> typing.Any:
         return getattr(event, 'source_platform_object', None)
 
-    async def target2legacy(self, event: SlackEvent) -> platform_events.FriendMessage | platform_events.GroupMessage | None:
+    async def target2legacy(
+        self, event: SlackEvent
+    ) -> platform_events.FriendMessage | platform_events.GroupMessage | None:
         eba_event = await self.target2yiri(event)
         if not isinstance(eba_event, platform_events.MessageReceivedEvent):
             return None

src/langbot/pkg/platform/adapters/telegram/__init__.py

@@ -1,3 +1,3 @@
 from langbot.pkg.platform.adapters.telegram.adapter import TelegramAdapter
 
-__all__ = ["TelegramAdapter"]
+__all__ = ['TelegramAdapter']

src/langbot/pkg/platform/adapters/telegram/platform_api.py

@@ -14,7 +14,7 @@ async def pin_message(bot: telegram.Bot, params: dict) -> dict:
         message_id=params['message_id'],
         disable_notification=params.get('disable_notification', False),
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def unpin_message(bot: telegram.Bot, params: dict) -> dict:
@@ -23,26 +23,26 @@ async def unpin_message(bot: telegram.Bot, params: dict) -> dict:
         chat_id=params['chat_id'],
         message_id=params.get('message_id'),
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def unpin_all_messages(bot: telegram.Bot, params: dict) -> dict:
     """Unpin all messages in a chat."""
     await bot.unpin_all_chat_messages(chat_id=params['chat_id'])
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def get_chat_administrators(bot: telegram.Bot, params: dict) -> dict:
     """Get chat administrator list."""
     admins = await bot.get_chat_administrators(chat_id=params['chat_id'])
     return {
-        "administrators": [
+        'administrators': [
             {
-                "user_id": a.user.id,
+                'user_id': a.user.id,
-                "username": a.user.username,
+                'username': a.user.username,
-                "first_name": a.user.first_name,
+                'first_name': a.user.first_name,
-                "status": a.status,
+                'status': a.status,
-                "custom_title": getattr(a, 'custom_title', None),
+                'custom_title': getattr(a, 'custom_title', None),
             }
             for a in admins
         ]
@@ -55,7 +55,7 @@ async def set_chat_title(bot: telegram.Bot, params: dict) -> dict:
         chat_id=params['chat_id'],
         title=params['title'],
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def set_chat_description(bot: telegram.Bot, params: dict) -> dict:
@@ -64,13 +64,13 @@ async def set_chat_description(bot: telegram.Bot, params: dict) -> dict:
         chat_id=params['chat_id'],
         description=params.get('description', ''),
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def get_chat_member_count(bot: telegram.Bot, params: dict) -> dict:
     """Get chat member count."""
     count = await bot.get_chat_member_count(chat_id=params['chat_id'])
-    return {"count": count}
+    return {'count': count}
 
 
 async def send_chat_action(bot: telegram.Bot, params: dict) -> dict:
@@ -79,7 +79,7 @@ async def send_chat_action(bot: telegram.Bot, params: dict) -> dict:
         chat_id=params['chat_id'],
         action=params.get('action', 'typing'),
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 async def create_chat_invite_link(bot: telegram.Bot, params: dict) -> dict:
@@ -91,10 +91,10 @@ async def create_chat_invite_link(bot: telegram.Bot, params: dict) -> dict:
         member_limit=params.get('member_limit'),
     )
     return {
-        "invite_link": link.invite_link,
+        'invite_link': link.invite_link,
-        "name": link.name,
+        'name': link.name,
-        "is_primary": link.is_primary,
+        'is_primary': link.is_primary,
-        "is_revoked": link.is_revoked,
+        'is_revoked': link.is_revoked,
     }
 
 
@@ -106,20 +106,20 @@ async def answer_callback_query(bot: telegram.Bot, params: dict) -> dict:
         show_alert=params.get('show_alert', False),
         url=params.get('url'),
     )
-    return {"ok": True}
+    return {'ok': True}
 
 
 # ---- Action dispatch table ----
 
 PLATFORM_API_MAP: dict[str, typing.Callable[[telegram.Bot, dict], typing.Awaitable[dict]]] = {
-    "pin_message": pin_message,
+    'pin_message': pin_message,
-    "unpin_message": unpin_message,
+    'unpin_message': unpin_message,
-    "unpin_all_messages": unpin_all_messages,
+    'unpin_all_messages': unpin_all_messages,
-    "get_chat_administrators": get_chat_administrators,
+    'get_chat_administrators': get_chat_administrators,
-    "set_chat_title": set_chat_title,
+    'set_chat_title': set_chat_title,
-    "set_chat_description": set_chat_description,
+    'set_chat_description': set_chat_description,
-    "get_chat_member_count": get_chat_member_count,
+    'get_chat_member_count': get_chat_member_count,
-    "send_chat_action": send_chat_action,
+    'send_chat_action': send_chat_action,
-    "create_chat_invite_link": create_chat_invite_link,
+    'create_chat_invite_link': create_chat_invite_link,
-    "answer_callback_query": answer_callback_query,
+    'answer_callback_query': answer_callback_query,
 }

src/langbot/pkg/platform/adapters/telegram/types.py

@@ -7,7 +7,8 @@ from enum import Enum
 
 class TelegramChatType(str, Enum):
     """Telegram chat type."""
-    PRIVATE = "private"
+
-    GROUP = "group"
+    PRIVATE = 'private'
-    SUPERGROUP = "supergroup"
+    GROUP = 'group'
-    CHANNEL = "channel"
+    SUPERGROUP = 'supergroup'
+    CHANNEL = 'channel'

src/langbot/pkg/platform/adapters/wecombot/api_impl.py

@@ -54,7 +54,9 @@ class WecomBotAPIMixin:
         self,
         group_id: typing.Union[int, str],
     ) -> list[platform_entities.UserGroupMember]:
-        return [member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)]
+        return [
+            member for (cached_group_id, _), member in self._member_cache.items() if cached_group_id == str(group_id)
+        ]
 
     async def upload_file(self, file_data: bytes, filename: str) -> str:
         raise NotSupportedError('upload_file')

src/langbot/pkg/platform/adapters/wecombot/event_converter.py

@@ -19,7 +19,9 @@ class WecomBotEventConverter(abstract_platform_adapter.AbstractEventConverter):
     async def yiri2target(event: platform_events.Event) -> typing.Any:
         return getattr(event, 'source_platform_object', None)
 
-    async def target2legacy(self, event: WecomBotEvent) -> platform_events.FriendMessage | platform_events.GroupMessage | None:
+    async def target2legacy(
+        self, event: WecomBotEvent
+    ) -> platform_events.FriendMessage | platform_events.GroupMessage | None:
         eba_event = await self.target2yiri(event)
         if not isinstance(eba_event, platform_events.MessageReceivedEvent):
             return None
@@ -50,7 +52,9 @@ class WecomBotEventConverter(abstract_platform_adapter.AbstractEventConverter):
     async def target2yiri(self, event: WecomBotEvent) -> platform_events.Event:
         if event.type in {'single', 'group'} and event.msgtype != 'event':
             return await self.message_to_eba(event)
-        return self.platform_specific(event, f'wecombot.{event.get("eventtype") or event.msgtype or event.type or "unknown"}')
+        return self.platform_specific(
+            event, f'wecombot.{event.get("eventtype") or event.msgtype or event.type or "unknown"}'
+        )
 
     async def message_to_eba(self, event: WecomBotEvent) -> platform_events.MessageReceivedEvent:
         sender = platform_entities.User(id=event.userid, nickname=event.username or event.userid)

src/langbot/pkg/platform/adapters/wecomcs/event_converter.py

@@ -17,7 +17,9 @@ class WecomCSEventConverter(abstract_platform_adapter.AbstractEventConverter):
         return getattr(event, 'source_platform_object', None)
 
     @staticmethod
-    async def target2legacy(event: WecomCSEvent, bot: WecomCSClient | None = None) -> platform_events.FriendMessage | None:
+    async def target2legacy(
+        event: WecomCSEvent, bot: WecomCSClient | None = None
+    ) -> platform_events.FriendMessage | None:
         eba_event = await WecomCSEventConverter.target2yiri(event, bot)
         if hasattr(eba_event, 'to_legacy_event'):
             return eba_event.to_legacy_event()
@@ -30,7 +32,9 @@ class WecomCSEventConverter(abstract_platform_adapter.AbstractEventConverter):
         return WecomCSEventConverter.platform_specific(event, f'wecomcs.{event.type or "unknown"}')
 
     @staticmethod
-    async def message_to_eba(event: WecomCSEvent, bot: WecomCSClient | None = None) -> platform_events.MessageReceivedEvent:
+    async def message_to_eba(
+        event: WecomCSEvent, bot: WecomCSClient | None = None
+    ) -> platform_events.MessageReceivedEvent:
         message_chain = await WecomCSMessageConverter.target2yiri(event)
         sender = await WecomCSEventConverter.user_from_event(event, bot)
         return platform_events.MessageReceivedEvent(

src/langbot/pkg/platform/human_input.py

@@ -89,10 +89,12 @@ def format_human_input_text(
     if actions:
         lines.append('')
         if input_defs:
-            lines.extend([
+            lines.extend(
+                [
                     'Reply with action plus field values to continue:',
                     '  action: <number or title>',
-            ])
+                ]
+            )
         else:
             lines.append('Reply with the number or title to continue:')
         for idx, action in enumerate(actions, start=1):

src/langbot/pkg/provider/tools/loaders/mcp.py

@@ -2327,15 +2327,9 @@ class MCPLoader(loader.ToolLoader):
 
         return items
 
-    def _session_by_source_id(
+    def _session_by_source_id(self, context: TenantContext, source_id: str) -> RuntimeMCPSession | None:
-        self, context: TenantContext, source_id: str
-    ) -> RuntimeMCPSession | None:
         return next(
-            (
+            (session for session in self._sessions_for_context(context) if session.server_uuid == source_id),
-                session
-                for session in self._sessions_for_context(context)
-                if session.server_uuid == source_id
-            ),
             None,
         )
 
@@ -2397,9 +2391,7 @@ class MCPLoader(loader.ToolLoader):
         source_id: str | None = None,
     ) -> typing.Any:
         """执行工具调用"""
-        execution_context = await self._assert_execution_active(
+        execution_context = await self._assert_execution_active(_execution_context_from_query(query))
-            _execution_context_from_query(query)
-        )
         if source_id is None and name == MCP_TOOL_LIST_RESOURCES:
             if getattr(query, 'variables', {}).get('_pipeline_mcp_resource_agent_read_enabled', True) is not True:
                 raise ToolExecutionDeniedError(name, 'MCP resource agent reads are disabled')

src/langbot/pkg/provider/tools/toolmgr.py

@@ -368,9 +368,7 @@ class ToolManager:
                 return None
             return await self.plugin_tool_loader.get_tool(name, source_id=source_id)
         if source == 'mcp':
-            return await self.mcp_tool_loader.get_tool(
+            return await self.mcp_tool_loader.get_tool(context, name, source_id=source_id)
-                context, name, source_id=source_id
-            )
         return None
 
     async def generate_tools_for_openai(self, use_funcs: list[resource_tool.LLMTool]) -> list:
@@ -488,9 +486,7 @@ class ToolManager:
         if source_ref is not None:
             execution_context = get_query_execution_context(query)
             await self._bind_plugin_workspace(execution_context)
-            sandbox_available = await self._workspace_sandbox_available(
+            sandbox_available = await self._workspace_sandbox_available(execution_context)
-                execution_context
-            )
             source = source_ref['source']
             source_id = source_ref.get('source_id')
             uses_source_id = False

tests/e2e/test_local_runner_fake_provider.py

@@ -57,7 +57,7 @@ def _package_local_agent_plugin(tmpdir: Path) -> Path:
     package_source = tmpdir / 'local-agent-package'
     ignore = shutil.ignore_patterns(
         '.git',
-        '.venv',
+        '.venv*',
         '__pycache__',
         '.pytest_cache',
         '.ruff_cache',
@@ -181,7 +181,20 @@ class _FakeToolManager:
     def __init__(self):
         self.calls: list[dict[str, Any]] = []
 
-    async def get_tool_schema(self, tool_name: str):
+    async def get_resolved_tool_catalog(
+        self,
+        context,
+        bound_plugins=None,
+        bound_mcp_servers=None,
+        include_skill_authoring=True,
+        include_mcp_resource_tools=False,
+    ):
+        assert context.workspace_uuid
+        return [{'name': E2E_TOOL_NAME, 'source': 'native', 'source_id': None}]
+
+    async def get_tool_schema(self, context, tool_name: str, source_ref=None):
+        assert context.workspace_uuid
+        assert source_ref == {'source': 'native', 'source_id': None}
         if tool_name != E2E_TOOL_NAME:
             return None, None
         return (
@@ -195,14 +208,15 @@ class _FakeToolManager:
             },
         )
 
-    async def get_tool_detail(self, tool_name: str):
+    async def get_tool_detail(self, context, tool_name: str, source_ref=None):
-        description, parameters = await self.get_tool_schema(tool_name)
+        description, parameters = await self.get_tool_schema(context, tool_name, source_ref=source_ref)
         if parameters is None:
             return None
         return {'name': tool_name, 'description': description, 'parameters': parameters}
 
-    async def execute_func_call(self, name: str, parameters: dict[str, Any], query: Any = None):
+    async def execute_func_call(self, name: str, parameters: dict[str, Any], query: Any = None, source_ref=None):
-        del query
+        assert query.workspace_uuid
+        assert source_ref == {'source': 'native', 'source_id': None}
         self.calls.append({'name': name, 'parameters': dict(parameters)})
         return {
             'value': f'tool-result:{parameters.get("query")}',
@@ -223,7 +237,8 @@ class _FakeKnowledgeBase:
     def get_name(self) -> str:
         return 'E2E Fake KB'
 
-    async def retrieve(self, query_text: str, settings: dict[str, Any]):
+    async def retrieve(self, context, query_text: str, settings: dict[str, Any]):
+        assert context.workspace_uuid
         self.retrieve_calls.append({'query_text': query_text, 'settings': settings})
         return [
             SimpleNamespace(
@@ -248,7 +263,8 @@ class _FakeRagManager:
         self.kb = kb
         self.knowledge_bases = {E2E_KB_UUID: kb}
 
-    async def get_knowledge_base_by_uuid(self, kb_uuid: str):
+    async def get_knowledge_base_by_uuid(self, context, kb_uuid: str):
+        assert context.workspace_uuid
         if kb_uuid == E2E_KB_UUID:
             return self.kb
         return None
@@ -443,10 +459,27 @@ def _scripted_tool_call(
 async def _boot_local_agent_app(tmpdir: Path):
     """Boot LangBot and wait until the Local Agent runner is discoverable."""
     from langbot.pkg.core import boot
-    from langbot_plugin.runtime.plugin.mgr import PluginInstallSource
 
     ap = await boot.make_app(asyncio.get_running_loop())
     run_task = asyncio.create_task(ap.run(), name='local-agent-e2e-app')
+    try:
+        await _wait_for_local_agent_runner(ap, tmpdir)
+    except BaseException:
+        # The caller has not received ap yet. Close it here so a boot failure
+        # cannot reconnect after the probe restores the global transport mode.
+        try:
+            await ap.shutdown()
+        finally:
+            run_task.cancel()
+            await asyncio.gather(run_task, return_exceptions=True)
+        raise
+    return ap, run_task
+
+
+async def _wait_for_local_agent_runner(ap, tmpdir: Path):
+    """Install and discover the runner on the application-owned connection."""
+    from langbot_plugin.runtime.plugin.mgr import PluginInstallSource
+
     for _ in range(60):
         handler = getattr(ap.plugin_connector, 'handler', None)
         if handler is not None:
@@ -478,8 +511,6 @@ async def _boot_local_agent_app(tmpdir: Path):
         else:
             raise AssertionError(f'{LOCAL_RUNNER_ID} was not discovered after installation')
 
-    return ap, run_task
-
 
 def _run_local_agent_probe(tmpdir: Path, probe):
     """Run one Local Agent probe inside the temporary LangBot app."""
@@ -689,7 +720,13 @@ def test_local_runner_retrieves_authorized_rag_context_through_host_action(
     assert retrieve_calls == [
         {
             'query_text': 'Answer with the retrieved RAG sentinel.',
-            'settings': {'top_k': 1, 'filters': {}},
+            'settings': {
+                'top_k': 1,
+                'filters': {},
+                'session_name': 'person_e2e-local-agent-rag-conversation',
+                'bot_uuid': '',
+                'sender_id': 'user-001',
+            },
         }
     ]
     assert any(
@@ -738,11 +775,13 @@ def test_local_runner_compacts_history_and_persists_checkpoint(
         )
 
         store = TranscriptStore(ap.persistence_mgr.get_db_engine())
+        execution_context = await ap.plugin_connector._current_execution_context()
         for index in range(12):
             await store.append_transcript(
                 transcript_id=None,
                 event_id=f'e2e-local-agent-history-{index}',
                 conversation_id='e2e-local-agent-compaction-conversation',
+                workspace_id=execution_context.workspace_uuid,
                 role='user' if index % 2 == 0 else 'assistant',
                 content=(
                     f'HIST_SENTINEL-{index} This is intentionally long deterministic history for compaction. ' * 10
@@ -847,11 +886,13 @@ def test_local_runner_combines_rag_compaction_and_multi_turn_tool_loop(
         ap.rag_mgr = _FakeRagManager(fake_kb)
 
         store = TranscriptStore(ap.persistence_mgr.get_db_engine())
+        execution_context = await ap.plugin_connector._current_execution_context()
         for index in range(16):
             await store.append_transcript(
                 transcript_id=None,
                 event_id=f'e2e-local-agent-combo-history-{index}',
                 conversation_id='e2e-local-agent-combo-conversation',
+                workspace_id=execution_context.workspace_uuid,
                 role='user' if index % 2 == 0 else 'assistant',
                 content=(
                     f'HIST_COMBO_SENTINEL-{index} RAG_TOOL_COMBO_GOAL '
@@ -907,7 +948,13 @@ def test_local_runner_combines_rag_compaction_and_multi_turn_tool_loop(
     assert retrieve_calls == [
         {
             'query_text': 'current combo request must survive; use RAG and tools before answering.',
-            'settings': {'top_k': 1, 'filters': {}},
+            'settings': {
+                'top_k': 1,
+                'filters': {},
+                'session_name': 'person_e2e-local-agent-combo-conversation',
+                'bot_uuid': '',
+                'sender_id': 'user-001',
+            },
         }
     ]
     assert invoke_count >= 4

tests/integration/persistence/test_migrations.py

@@ -16,7 +16,9 @@ import sqlalchemy
 from sqlalchemy import text
 from sqlalchemy.ext.asyncio import create_async_engine
 
+from langbot.pkg.entity import persistence
 from langbot.pkg.entity.persistence.base import Base
+from langbot.pkg.utils import importutil
 from langbot.pkg.persistence.alembic_runner import (
     run_alembic_downgrade,
     run_alembic_upgrade,
@@ -28,6 +30,11 @@ from alembic.config import Config
 from alembic.script import ScriptDirectory
 
 
+# Match PersistenceManager's model registration before create_all, including
+# workspace foreign-key targets, without relying on other tests being collected.
+importutil.import_modules_in_pkg(persistence)
+
+
 def _get_script_head() -> str:
     """Resolve the current Alembic head revision from the script directory.
 
@@ -108,7 +115,6 @@ class TestSQLiteMigrationUpgrade:
         await run_alembic_upgrade(sqlite_engine, 'head')
 
         assert await get_alembic_current(sqlite_engine) == _get_script_head()
-        assert _get_script_head() == '0023_drop_agent_enabled'
 
     @pytest.mark.asyncio
     async def test_upgrade_from_development_workspace_head_to_merged_head(self, sqlite_engine):
@@ -120,7 +126,6 @@ class TestSQLiteMigrationUpgrade:
         await run_alembic_upgrade(sqlite_engine, 'head')
 
         assert await get_alembic_current(sqlite_engine) == _get_script_head()
-        assert _get_script_head() == '0023_drop_agent_enabled'
 
     @pytest.mark.asyncio
     async def test_upgrade_from_reasoning_config_head_to_merged_head(self, sqlite_engine):
@@ -131,7 +136,7 @@ class TestSQLiteMigrationUpgrade:
         await run_alembic_stamp(sqlite_engine, '0018_llm_reasoning_config')
         await run_alembic_upgrade(sqlite_engine, 'head')
 
-        assert await get_alembic_current(sqlite_engine) == '0023_drop_agent_enabled'
+        assert await get_alembic_current(sqlite_engine) == _get_script_head()
 
     @pytest.mark.asyncio
     async def test_upgrade_removes_agent_enabled_column(self, sqlite_engine):

tests/integration/persistence/test_migrations_postgres.py

@@ -856,7 +856,7 @@ class TestPostgreSQLTenantRuntime:
                                 adapter='capacity-probe',
                                 adapter_config={},
                                 enable=False,
-                                pipeline_routing_rules=[],
+                                event_bindings=[],
                             ),
                             persistence_pipeline.LegacyPipeline(
                                 uuid=f'capacity-pipeline-{suffix}',

tests/unit_tests/agent/test_local_runner_e2e_fixtures.py

@@ -0,0 +1,123 @@
+"""Keep E2E fake resources compatible with the real Host contract."""
+
+from types import SimpleNamespace
+from unittest.mock import Mock
+import pytest
+from langbot.pkg.agent.runner.descriptor import RunnerDescriptor
+from langbot.pkg.agent.runner.resource_builder import AgentResourceBuilder
+from langbot.pkg.api.http.context import ExecutionContext
+from tests.e2e.test_local_runner_fake_provider import (
+    E2E_KB_UUID,
+    E2E_TOOL_NAME,
+    LOCAL_RUNNER_ID,
+    _FakeKnowledgeBase,
+    _FakeRagManager,
+    _FakeToolManager,
+    _binding,
+    _event,
+)
+
+CONTEXT = ExecutionContext(instance_uuid='instance-test', workspace_uuid='workspace-test', placement_generation=1)
+SOURCE = {'source': 'native', 'source_id': None}
+
+
+async def _resources(tool_mgr, rag_mgr, binding):
+    descriptor = RunnerDescriptor(
+        id=LOCAL_RUNNER_ID,
+        source='plugin',
+        label={'en_US': 'Local Agent'},
+        plugin_author='langbot-team',
+        plugin_name='LocalAgent',
+        runner_name='default',
+        capabilities={'tool_calling': True, 'knowledge_retrieval': True},
+        permissions={'tools': ['detail', 'call'], 'knowledge_bases': ['list', 'retrieve']},
+    )
+    app = SimpleNamespace(logger=Mock(), skill_mgr=None, tool_mgr=tool_mgr, rag_mgr=rag_mgr)
+    return await AgentResourceBuilder(app).build_resources_from_binding(
+        CONTEXT,
+        _event(event_id='evt', conversation_id='conv', text='test'),
+        binding,
+        descriptor,
+    )
+
+
+@pytest.mark.asyncio
+async def test_fake_tool_is_projected_with_frozen_source_and_executed():
+    manager = _FakeToolManager()
+    resources = await _resources(
+        manager, _FakeRagManager(_FakeKnowledgeBase()), _binding(allowed_tool_names=[E2E_TOOL_NAME])
+    )
+    assert len(resources['tools']) == 1
+    tool = resources['tools'][0]
+    assert tool['tool_name'] == E2E_TOOL_NAME
+    assert {key: tool[key] for key in SOURCE} == SOURCE
+    assert tool['parameters']['required'] == ['query']
+    detail = await manager.get_tool_detail(CONTEXT, E2E_TOOL_NAME, source_ref=SOURCE)
+    assert detail['parameters'] == tool['parameters']
+    result = await manager.execute_func_call(
+        name=E2E_TOOL_NAME,
+        parameters={'query': 'alpha'},
+        query=SimpleNamespace(workspace_uuid=CONTEXT.workspace_uuid),
+        source_ref=SOURCE,
+    )
+    assert result['value'] == 'tool-result:alpha'
+    assert manager.calls == [{'name': E2E_TOOL_NAME, 'parameters': {'query': 'alpha'}}]
+
+
+@pytest.mark.asyncio
+async def test_fake_kb_is_projected_and_retrieved_with_execution_context():
+    kb = _FakeKnowledgeBase()
+    manager = _FakeRagManager(kb)
+    resources = await _resources(_FakeToolManager(), manager, _binding(allowed_kb_uuids=[E2E_KB_UUID]))
+    assert [item['kb_id'] for item in resources['knowledge_bases']] == [E2E_KB_UUID]
+    resolved = await manager.get_knowledge_base_by_uuid(CONTEXT, E2E_KB_UUID)
+    entries = await resolved.retrieve(CONTEXT, 'test', settings={'top_k': 1, 'filters': {}})
+    assert 'RAG_SENTINEL' in entries[0].content
+    assert kb.retrieve_calls == [{'query_text': 'test', 'settings': {'top_k': 1, 'filters': {}}}]
+
+
+def test_local_agent_package_excludes_alternate_virtualenvs(tmp_path, monkeypatch):
+    import zipfile
+    from tests.e2e import test_local_runner_fake_provider as fixtures
+
+    source = tmp_path / 'source'
+    source.mkdir()
+    (source / 'manifest.yaml').write_text('metadata: {author: langbot-team, name: LocalAgent}')
+    for name in ('.venv', '.venv311'):
+        (source / name).mkdir()
+        (source / name / 'not-plugin.py').write_text('pass')
+    monkeypatch.setattr(fixtures, '_local_agent_repo', lambda: source)
+    package = fixtures._package_local_agent_plugin(tmp_path / 'package')
+    with zipfile.ZipFile(package) as archive:
+        assert archive.namelist() == ['manifest.yaml']
+
+
+@pytest.mark.asyncio
+async def test_failed_local_agent_boot_shuts_down_before_restoring_transport(monkeypatch, tmp_path):
+    import asyncio
+    from unittest.mock import AsyncMock
+    from langbot.pkg.core import boot
+    from tests.e2e import test_local_runner_fake_provider as fixtures
+
+    async def running():
+        await asyncio.Event().wait()
+
+    app = SimpleNamespace(
+        run=running,
+        shutdown=AsyncMock(),
+        plugin_connector=SimpleNamespace(
+            handler=SimpleNamespace(ping=AsyncMock()), _current_execution_context=AsyncMock(return_value=CONTEXT)
+        ),
+        runner_registry=SimpleNamespace(list_runners=AsyncMock(side_effect=ValueError('probe boot failure'))),
+    )
+    monkeypatch.setattr(boot, 'make_app', AsyncMock(return_value=app))
+    try:
+        with pytest.raises(ValueError, match='probe boot failure'):
+            await fixtures._boot_local_agent_app(tmp_path)
+        app.shutdown.assert_awaited_once()
+        assert not any(task.get_name() == 'local-agent-e2e-app' for task in asyncio.all_tasks())
+    finally:
+        tasks = [task for task in asyncio.all_tasks() if task.get_name() == 'local-agent-e2e-app']
+        for task in tasks:
+            task.cancel()
+        await asyncio.gather(*tasks, return_exceptions=True)

uv.lock

@@ -2119,7 +2119,7 @@ requires-dist = [
     { name = "ebooklib", specifier = ">=0.18" },
     { name = "gewechat-client", specifier = ">=0.1.5" },
     { name = "html2text", specifier = ">=2024.2.26" },
-    { name = "langbot-plugin", git = "https://github.com/langbot-app/langbot-plugin-sdk?rev=eac2c60509534512f9a373cd0c801e75985e0612" },
+    { name = "langbot-plugin", git = "https://github.com/langbot-app/langbot-plugin-sdk?rev=92a9e03fa9c791f4ed30cc3f5f0602c13b800d28" },
     { name = "langchain", specifier = ">=1.3.9" },
     { name = "langchain-core", specifier = ">=1.3.3" },
     { name = "langchain-text-splitters", specifier = ">=1.1.2" },
@@ -2186,7 +2186,7 @@ dev = [
 [[package]]
 name = "langbot-plugin"
 version = "0.5.5"
-source = { git = "https://github.com/langbot-app/langbot-plugin-sdk?rev=eac2c60509534512f9a373cd0c801e75985e0612#eac2c60509534512f9a373cd0c801e75985e0612" }
+source = { git = "https://github.com/langbot-app/langbot-plugin-sdk?rev=92a9e03fa9c791f4ed30cc3f5f0602c13b800d28#92a9e03fa9c791f4ed30cc3f5f0602c13b800d28" }
 dependencies = [
     { name = "aiofiles" },
     { name = "aiohttp" },

web/tests/e2e/crud-smoke.spec.ts

@@ -89,7 +89,7 @@ test.describe('frontend CRUD smoke flows', () => {
     await expect(
       page.locator('[data-processor-kind="pipeline"]'),
     ).toContainText(
-      '流水线即为经典的“收到消息、请求AI、回复用户”流程，并辅以常用的配置功能。仅处理消息事件，适合步骤明确、需要稳定控制处理过程的场景。',
+      '流水线按“接收消息、调用 AI、回复用户”的固定流程运行，可配置知识库和插件扩展。仅处理消息事件，适合步骤明确、需要控制处理过程的场景。',
     );
   });
 
@@ -1034,7 +1034,7 @@ test.describe('agent runner resource selectors', () => {
     await expect(
       page.getByText('No Runner extension is installed yet.'),
     ).toBeVisible();
-    await expect(page.getByText('Runner Marketplace')).toBeVisible();
+    await expect(page.getByText('Runner plugins in Marketplace')).toBeVisible();
     const selectorPopup = page.locator('[data-slot="select-content"]');
     await expect(
       selectorPopup.getByText('Runner used by the grouped selector test.', {
@@ -1055,20 +1055,35 @@ test.describe('agent runner resource selectors', () => {
       });
 
     await page
-      .getByRole('option')
+      .getByRole('button', { name: 'Install Marketplace Runner', exact: true })
-      .filter({ hasText: 'Marketplace Runner' })
       .click();
 
     await expect.poll(() => installRequests).toBe(1);
     await expect.poll(() => taskPolls).toBeGreaterThan(0);
+    await expect(
+      page.getByRole('option', {
+        name: 'Marketplace Runner Runner used by the grouped selector test.',
+        exact: true,
+      }),
+    ).toBeVisible();
+    await page
+      .getByRole('option', {
+        name: 'Marketplace Runner Runner used by the grouped selector test.',
+        exact: true,
+      })
+      .click();
     await expect(runnerSelect).toContainText('Marketplace Runner');
 
     await runnerSelect.click();
     await expect(
-      page.getByRole('option').filter({ hasText: runnerId }),
+      page.getByRole('option', {
+        name: 'Marketplace Runner Runner used by the grouped selector test.',
+        exact: true,
+      }),
     ).toBeVisible();
     await expect(
-      page.getByText('Runner used by the grouped selector test.', {
+      page.getByRole('button', {
+        name: 'Install Marketplace Runner',
         exact: true,
       }),
     ).toHaveCount(0);
@@ -1130,7 +1145,7 @@ test.describe('agent runner resource selectors', () => {
     const runnerSelect = page.getByRole('combobox', { name: 'Runner' });
     await runnerSelect.click();
     await expect(page.getByText('Installed Runners')).toBeVisible();
-    await expect(page.getByText('Runner Marketplace')).toBeVisible();
+    await expect(page.getByText('Runner plugins in Marketplace')).toBeVisible();
     await expect(
       page
         .locator('[data-slot="select-content"]')
@@ -1139,9 +1154,10 @@ test.describe('agent runner resource selectors', () => {
         }),
     ).toBeVisible();
     await expect(
-      page
+      page.getByRole('button', {
-        .getByRole('option')
+        name: 'Install Pipeline Marketplace Runner',
-        .filter({ hasText: 'Pipeline Marketplace Runner' }),
+        exact: true,
+      }),
     ).toBeVisible();
   });
 
@@ -1214,8 +1230,11 @@ test.describe('agent and pipeline save concurrency', () => {
 
     await page.goto('/home/agents?id=agent-save-race');
     const saveButton = page.getByRole('button', { name: /^Save$/ });
-    await page.getByRole('tab', { name: 'Bindable Event Range' }).click();
+    await page.getByRole('tab', { name: 'Events & tools' }).click();
-    const eventPatterns = page.getByLabel('Event Range');
+    const eventPatterns = page.getByRole('button', {
+      name: 'Add event',
+      exact: true,
+    });
     await expect(eventPatterns).toBeVisible();
 
     await eventPatterns.click();
@@ -1231,8 +1250,8 @@ test.describe('agent and pipeline save concurrency', () => {
     await eventPatterns.click();
     await page.getByRole('option').filter({ hasText: 'group.*' }).click();
     await page
-      .getByRole('option')
+      .getByRole('button', { name: 'Remove event', exact: true })
-      .filter({ hasText: 'message.received' })
+      .first()
       .click();
     await page.keyboard.press('Escape');
     await forceFormSubmit(page, '#agent-form');

web/tests/e2e/fixtures/langbot-api.ts

@@ -725,10 +725,12 @@ async function handleBackendApi(route: Route, state: LangBotApiMockState) {
     return fulfillJson(route, { agents: state.pipelines });
   }
 
-  const agentDebugMatch = path.match(/^\/api\/v1\/agents\/([^/]+)\/debug$/);
+  const agentDebugMatch = path.match(
+    /^\/api\/v1\/agents\/([^/]+)\/debug(?:\/stream)?$/,
+  );
   if (agentDebugMatch) {
     const payload = parseJsonBody(route);
-    return fulfillJson(route, {
+    const result = {
       event_id: nextId(state, 'event'),
       event_type: String(payload.event_type || 'message.received'),
       conversation_id: String(payload.conversation_id || 'debug-session'),
@@ -740,8 +742,16 @@ async function handleBackendApi(route: Route, state: LangBotApiMockState) {
           text: 'Mock Agent response',
         },
       ],
+    };
+    if (path.endsWith('/stream')) {
+      return route.fulfill({
+        status: 200,
+        contentType: 'application/x-ndjson',
+        body: JSON.stringify({ kind: 'completed', data: result }) + '\n',
       });
     }
+    return fulfillJson(route, result);
+  }
 
   const agentMatch = path.match(/^\/api\/v1\/agents\/([^/]+)$/);
   if (agentMatch) {

web/tests/e2e/home-smoke.spec.ts

@@ -11,7 +11,7 @@ const appRoutes = [
   {
     path: '/home/agents',
     heading: 'Processors',
-    bodyText: 'Select an Agent or Pipeline from the sidebar',
+    bodyText: 'Select a processor from the sidebar',
   },
   {
     path: '/home/extensions',

web/tests/e2e/processor-detail-workbench.spec.ts

@@ -60,8 +60,8 @@ test.describe('processor detail workbench', () => {
 
     const appShell = page.locator('[class*="group/sidebar-wrapper"]');
     const sidebarInset = page.locator('[data-slot="sidebar-inset"]');
-    await expect(appShell).toHaveCSS('overflow', 'clip');
+    await expect(appShell).toHaveCSS('overflow', 'hidden');
-    await expect(sidebarInset).toHaveCSS('overflow', 'clip');
+    await expect(sidebarInset).toHaveCSS('overflow', 'hidden');
     await appShell.evaluate((element) => {
       element.scrollTop = 300;
     });
@@ -79,9 +79,7 @@ test.describe('processor detail workbench', () => {
     const flow = configPanel.getByRole('tablist');
     await expect(flow.getByRole('tab').nth(0)).toContainText('Runner');
     await expect(flow.getByRole('tab').nth(1)).toContainText('Local Agent');
-    await expect(flow.getByRole('tab').nth(2)).toContainText(
+    await expect(flow.getByRole('tab').nth(2)).toContainText('Events & tools');
-      'Bindable Event Range',
-    );
     await expect(flow.getByRole('tab')).toHaveCount(3);
     await expect(flow.getByText('Management')).toHaveCount(0);
 
@@ -126,19 +124,32 @@ test.describe('processor detail workbench', () => {
 
     await flow.getByRole('tab').nth(2).click();
     await expect(
-      configPanel.getByText('Bindable Event Range', { exact: true }).last(),
+      configPanel.getByText('Events & tools', { exact: true }).last(),
     ).toBeVisible();
-    const eventPicker = configPanel.getByRole('combobox', {
+    const eventPicker = configPanel.getByRole('button', {
-      name: 'Event Range',
+      name: 'Add event',
+      exact: true,
     });
     await expect(eventPicker).toBeVisible();
-    await expect(configPanel.getByRole('textbox')).toHaveCount(0);
+    await expect(configPanel.getByRole('textbox')).toHaveCount(1);
+    await expect(
+      configPanel.getByRole('textbox', { name: 'Search tools…' }),
+    ).toBeVisible();
     await eventPicker.click();
     await expect(
       page.getByRole('option').filter({ hasText: 'message.received' }),
     ).toBeVisible();
-    await expect(page.getByRole('group', { name: 'Messages' })).toHaveCount(1);
+    await expect(
-    await expect(page.getByRole('group', { name: 'Groups' })).toHaveCount(1);
+      page.getByRole('option').filter({ hasText: 'group.*' }),
+    ).toBeVisible();
+    await expect(
+      page
+        .locator('[cmdk-group-heading]')
+        .getByText('Messages', { exact: true }),
+    ).toBeVisible();
+    await expect(
+      page.locator('[cmdk-group-heading]').getByText('Groups', { exact: true }),
+    ).toBeVisible();
     await page
       .getByRole('option')
       .filter({ hasText: 'message.*' })
@@ -189,7 +200,7 @@ test.describe('processor detail workbench', () => {
       }
       if (
         request.method() === 'POST' &&
-        path === '/api/v1/agents/agent-workbench/debug'
+        path === '/api/v1/agents/agent-workbench/debug/stream'
       ) {
         requests.push('debug');
       }
@@ -237,15 +248,17 @@ test.describe('processor detail workbench', () => {
   }) => {
     await installLangBotApiMocks(page, { authenticated: true });
     await page.route(
-      '**/api/v1/agents/agent-workbench/debug',
+      '**/api/v1/agents/agent-workbench/debug/stream',
       async (route) => {
         await route.fulfill({
-          status: 422,
+          status: 200,
-          contentType: 'application/json',
+          contentType: 'application/x-ndjson',
-          body: JSON.stringify({
+          body:
+            JSON.stringify({
+              kind: 'error',
               code: 'dify.config_invalid',
               msg: 'api-key is required',
-          }),
+            }) + '\n',
         });
       },
     );
@@ -454,8 +467,8 @@ test.describe('processor detail workbench', () => {
 
     const appShell = page.locator('[class*="group/sidebar-wrapper"]');
     const sidebarInset = page.locator('[data-slot="sidebar-inset"]');
-    await expect(appShell).toHaveCSS('overflow', 'clip');
+    await expect(appShell).toHaveCSS('overflow', 'hidden');
-    await expect(sidebarInset).toHaveCSS('overflow', 'clip');
+    await expect(sidebarInset).toHaveCSS('overflow', 'hidden');
     await expect
       .poll(() => appShell.evaluate((element) => element.scrollTop))
       .toBe(0);

web/tests/unit/agent-debug-execution.test.mjs

@@ -2,27 +2,64 @@ import assert from 'node:assert/strict';
 import fs from 'node:fs';
 import test from 'node:test';
 import ts from 'typescript';
-const source = fs.readFileSync(new URL('../../src/app/home/agents/components/debug-execution.ts', import.meta.url), 'utf8');
+const source = fs.readFileSync(
+  new URL(
+    '../../src/app/home/agents/components/debug-execution.ts',
+    import.meta.url,
+  ),
+  'utf8',
+);
 const module = { exports: {} };
-new Function('exports', ts.transpileModule(source, { compilerOptions: { module: ts.ModuleKind.CommonJS, target: ts.ScriptTarget.ES2022 } }).outputText)(module.exports);
+new Function(
+  'exports',
+  ts.transpileModule(source, {
+    compilerOptions: {
+      module: ts.ModuleKind.CommonJS,
+      target: ts.ScriptTarget.ES2022,
+    },
+  }).outputText,
+)(module.exports);
 const { executionSteps } = module.exports;
 const event = (type, data) => ({ type, data });
 test('separates streamed thinking and text, replaces final snapshot without duplication', () => {
-  assert.deepEqual(executionSteps([
+  assert.deepEqual(
+    executionSteps([
       event('message.delta', { chunk: { content: '<think>plan' } }),
       event('message.delta', { chunk: { content: '</think>hello' } }),
-    event('message.completed', {message: {content:'<think>plan</think>hello'}}),
+      event('message.completed', {
+        message: { content: '<think>plan</think>hello' },
+      }),
       event('run.completed', { message: { content: 'hello' } }),
-  ]), [{kind:'message', text:'hello', reasoning:'plan'}]);
+    ]),
+    [{ kind: 'message', text: 'hello', reasoning: 'plan' }],
+  );
 });
 test('retains structured reasoning and tool parameters/results in order', () => {
   const steps = executionSteps([
-    event('message.delta', {chunk: {provider_specific_fields:{reasoning_content:'plan'}}}),
+    event('message.delta', {
+      chunk: { provider_specific_fields: { reasoning_content: 'plan' } },
+    }),
     event('message.completed', { message: { content: '' } }),
-    event('tool.call.started', {tool_call_id:'1',tool_name:'exec', parameters:{command:'echo hi'}}),
+    event('tool.call.started', {
-    event('tool.call.started', {tool_call_id:'2',tool_name:'exec', parameters:{command:'bad'}}),
+      tool_call_id: '1',
-    event('tool.call.completed', {tool_call_id:'2',tool_name:'exec', error:'failed'}),
+      tool_name: 'exec',
-    event('tool.call.completed', {tool_call_id:'1',tool_name:'exec', result:{stdout:'hi'}}),
+      parameters: { command: 'echo hi' },
+    }),
+    event('tool.call.started', {
+      tool_call_id: '2',
+      tool_name: 'exec',
+      parameters: { command: 'bad' },
+    }),
+    event('tool.call.completed', {
+      tool_call_id: '2',
+      tool_name: 'exec',
+      error: 'failed',
+    }),
+    event('tool.call.completed', {
+      tool_call_id: '1',
+      tool_name: 'exec',
+      result: { stdout: 'hi' },
+    }),
     event('run.failed', {}),
   ]);
   assert.equal(steps[0].reasoning, 'plan');
@@ -33,17 +70,32 @@ test('retains structured reasoning and tool parameters/results in order', () =>
 });
 
 test('replaces LocalAgent cumulative snapshots instead of repeating text', () => {
-  assert.deepEqual(executionSteps([
+  assert.deepEqual(
+    executionSteps([
       event('message.delta', { chunk: { content: 'hello', msg_sequence: 1 } }),
-    event('message.delta', {chunk: {content:'hello world', msg_sequence:2}}),
+      event('message.delta', {
-    event('message.delta', {chunk: {content:'hello world', msg_sequence:3, is_final:true}}),
+        chunk: { content: 'hello world', msg_sequence: 2 },
-  ]), [{kind:'message', text:'hello world', reasoning:''}]);
+      }),
+      event('message.delta', {
+        chunk: { content: 'hello world', msg_sequence: 3, is_final: true },
+      }),
+    ]),
+    [{ kind: 'message', text: 'hello world', reasoning: '' }],
+  );
 });
 
 test('shows failed tool results even when the call transport completed', () => {
   const steps = executionSteps([
-    event('tool.call.started', {tool_call_id:'exit7', tool_name:'exec', parameters:{command:'exit 7'}}),
+    event('tool.call.started', {
-    event('tool.call.completed', {tool_call_id:'exit7', tool_name:'exec', result:{ok:false, exit_code:7, stderr:'expected'}}),
+      tool_call_id: 'exit7',
+      tool_name: 'exec',
+      parameters: { command: 'exit 7' },
+    }),
+    event('tool.call.completed', {
+      tool_call_id: 'exit7',
+      tool_name: 'exec',
+      result: { ok: false, exit_code: 7, stderr: 'expected' },
+    }),
   ]);
   assert.equal(steps[0].status, 'failed');
   assert.equal(steps[0].result.exit_code, 7);
@@ -53,15 +105,31 @@ test('does not repeat prior thinking across LocalAgent tool turns', () => {
   const prefix = '<think>first thought</think>';
   const steps = executionSteps([
     event('message.delta', { chunk: { content: prefix, msg_sequence: 1 } }),
-    event('tool.call.started', {tool_call_id:'w',tool_name:'write',parameters:{path:'/workspace/a'}}),
+    event('tool.call.started', {
-    event('tool.call.completed', {tool_call_id:'w',tool_name:'write',result:{ok:true}}),
+      tool_call_id: 'w',
-    event('message.delta', {chunk:{content:prefix+'now read',msg_sequence:1}}),
+      tool_name: 'write',
+      parameters: { path: '/workspace/a' },
+    }),
+    event('tool.call.completed', {
+      tool_call_id: 'w',
+      tool_name: 'write',
+      result: { ok: true },
+    }),
+    event('message.delta', {
+      chunk: { content: prefix + 'now read', msg_sequence: 1 },
+    }),
     event('tool.call.started', { tool_call_id: 'r', tool_name: 'read' }),
-    event('tool.call.completed', {tool_call_id:'r',tool_name:'read',result:{ok:true}}),
+    event('tool.call.completed', {
-    event('message.delta', {chunk:{content:prefix+'now read'+'done',msg_sequence:1}}),
+      tool_call_id: 'r',
+      tool_name: 'read',
+      result: { ok: true },
+    }),
+    event('message.delta', {
+      chunk: { content: prefix + 'now read' + 'done', msg_sequence: 1 },
+    }),
     event('message.completed', { message: { content: 'done' } }),
   ]);
-  const messages = steps.filter(s=>s.kind==='message');
+  const messages = steps.filter((s) => s.kind === 'message');
   assert.deepEqual(messages, [
     { kind: 'message', text: '', reasoning: 'first thought' },
     { kind: 'message', text: 'now read', reasoning: '' },