feat: add agent loop protection - max iterations and tool result truncation

- Add max-tool-iterations config (default 16) to prevent runaway agent loops
- Add max-tool-result-chars config (default 8000) to truncate oversized tool results
- Both settings are configurable in pipeline UI under Local Agent settings
- Logs warnings when limits are hit for debugging

Closes #2051

pyproject.toml

@@ -61,7 +61,7 @@ dependencies = [
     "html2text>=2024.2.26",
     "langchain>=0.2.0",
     "langchain-text-splitters>=0.0.1",
-    "chromadb>=0.4.24",
+    "chromadb>=1.0.0,<2.0.0",
     "qdrant-client (>=1.15.1,<2.0.0)",
     "pyseekdb==1.1.0.post3",
     "langbot-plugin==0.3.0",

src/langbot/libs/wecom_customer_service_api/api.py

@@ -10,6 +10,7 @@ from typing import Callable
 from .wecomcsevent import WecomCSEvent
 import langbot_plugin.api.entities.builtin.platform.message as platform_message
 import aiofiles
+import time
 
 
 class WecomCSClient:
@@ -34,6 +35,10 @@ class WecomCSClient:
         self.unified_mode = unified_mode
         self.app = Quart(__name__)
 
+        # Customer info cache: {external_userid: (info_dict, timestamp)}
+        self._customer_cache: dict[str, tuple[dict, float]] = {}
+        self._cache_ttl = 60  # Cache TTL in seconds (1 minute)
+
         # 只有在非统一模式下才注册独立路由
         if not self.unified_mode:
             self.app.add_url_rule(
@@ -378,3 +383,53 @@ class WecomCSClient:
     async def get_media_id(self, image: platform_message.Image):
         media_id = await self.upload_to_work(image=image)
         return media_id
+
+    async def get_customer_info(self, external_userid: str) -> dict | None:
+        """
+        Get customer information by external_userid with caching.
+
+        Uses a 1-minute cache to avoid repeated API calls for the same user.
+
+        Args:
+            external_userid: The external user ID of the customer.
+
+        Returns:
+            Customer info dict with 'nickname', 'avatar', etc., or None if not found.
+        """
+        # Check cache first
+        current_time = time.time()
+        if external_userid in self._customer_cache:
+            cached_info, cached_time = self._customer_cache[external_userid]
+            if current_time - cached_time < self._cache_ttl:
+                return cached_info
+
+        # Cache miss or expired, fetch from API
+        if not await self.check_access_token():
+            self.access_token = await self.get_access_token(self.secret)
+
+        url = f'{self.base_url}/kf/customer/batchget?access_token={self.access_token}'
+
+        payload = {
+            'external_userid_list': [external_userid],
+        }
+
+        async with httpx.AsyncClient() as client:
+            response = await client.post(url, json=payload)
+            data = response.json()
+
+            if data.get('errcode') in [40014, 42001]:
+                self.access_token = await self.get_access_token(self.secret)
+                return await self.get_customer_info(external_userid)
+
+            if data.get('errcode', 0) != 0:
+                if self.logger:
+                    await self.logger.warning(f'Failed to get customer info: {data}')
+                return None
+
+            customer_list = data.get('customer_list', [])
+            if customer_list:
+                customer_info = customer_list[0]
+                # Store in cache
+                self._customer_cache[external_userid] = (customer_info, current_time)
+                return customer_info
+            return None

src/langbot/pkg/api/http/service/monitoring.py

@@ -30,6 +30,7 @@ class MonitoringService:
         level: str = 'info',
         platform: str | None = None,
         user_id: str | None = None,
+        user_name: str | None = None,
         runner_name: str | None = None,
         variables: str | None = None,
         role: str = 'user',
@@ -49,6 +50,7 @@ class MonitoringService:
             'level': level,
             'platform': platform,
             'user_id': user_id,
+            'user_name': user_name,
             'runner_name': runner_name,
             'variables': variables,
             'role': role,
@@ -152,6 +154,7 @@ class MonitoringService:
         pipeline_name: str,
         platform: str | None = None,
         user_id: str | None = None,
+        user_name: str | None = None,
     ) -> None:
         """Record a new session"""
         session_data = {
@@ -166,6 +169,7 @@ class MonitoringService:
             'is_active': True,
             'platform': platform,
             'user_id': user_id,
+            'user_name': user_name,
         }
 
         await self.ap.persistence_mgr.execute_async(

src/langbot/pkg/core/app.py

@@ -9,6 +9,7 @@ from ..platform import botmgr as im_mgr
 from ..platform.webhook_pusher import WebhookPusher
 from ..provider.session import sessionmgr as llm_session_mgr
 from ..provider.modelmgr import modelmgr as llm_model_mgr
+
 from langbot.pkg.provider.tools import toolmgr as llm_tool_mgr
 from ..config import manager as config_mgr
 from ..command import cmdmgr
@@ -30,6 +31,7 @@ from ..api.http.service import mcp as mcp_service
 from ..api.http.service import apikey as apikey_service
 from ..api.http.service import webhook as webhook_service
 from ..api.http.service import monitoring as monitoring_service
+
 from ..discover import engine as discover_engine
 from ..storage import mgr as storagemgr
 from ..utils import logcache

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

@@ -20,6 +20,7 @@ class MonitoringMessage(Base):
     level = sqlalchemy.Column(sqlalchemy.String(50), nullable=False)  # info, warning, error, debug
     platform = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
     user_id = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
+    user_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # User display name
     runner_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # Runner name for this query
     variables = sqlalchemy.Column(sqlalchemy.Text, nullable=True)  # Query variables as JSON string
     role = sqlalchemy.Column(sqlalchemy.String(50), nullable=True, default='user')  # user, assistant
@@ -64,6 +65,7 @@ class MonitoringSession(Base):
     is_active = sqlalchemy.Column(sqlalchemy.Boolean, nullable=False, default=True, index=True)
     platform = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
     user_id = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)
+    user_name = sqlalchemy.Column(sqlalchemy.String(255), nullable=True)  # User display name
 
 
 class MonitoringError(Base):

src/langbot/pkg/persistence/migrations/dbm021_merge_exception_handling.py

@@ -0,0 +1,74 @@
+from .. import migration
+
+import sqlalchemy
+import json
+
+
+@migration.migration_class(21)
+class DBMigrateMergeExceptionHandling(migration.DBMigration):
+    """Merge hide-exception and block-failed-request-output into a single exception-handling select option,
+    and add failure-hint field.
+
+    Conversion logic:
+    - block-failed-request-output=true  ->  exception-handling: hide
+    - hide-exception=true               ->  exception-handling: show-hint
+    - hide-exception=false              ->  exception-handling: show-error
+    """
+
+    async def upgrade(self):
+        """Upgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'output' not in config:
+                config['output'] = {}
+            if 'misc' not in config['output']:
+                config['output']['misc'] = {}
+
+            misc = config['output']['misc']
+
+            # Determine new exception-handling value from legacy fields
+            hide_exception = misc.get('hide-exception', True)
+            block_failed = misc.get('block-failed-request-output', False)
+
+            if block_failed:
+                exception_handling = 'hide'
+            elif hide_exception:
+                exception_handling = 'show-hint'
+            else:
+                exception_handling = 'show-error'
+
+            misc['exception-handling'] = exception_handling
+
+            # Add failure-hint with default value
+            misc['failure-hint'] = 'Request failed.'
+
+            # Remove legacy fields
+            misc.pop('hide-exception', None)
+
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+
+    async def downgrade(self):
+        """Downgrade"""
+        pass

src/langbot/pkg/persistence/migrations/dbm022_monitoring_user_name.py

@@ -0,0 +1,73 @@
+import sqlalchemy
+from .. import migration
+
+
+@migration.migration_class(22)
+class DBMigrateMonitoringUserId(migration.DBMigration):
+    """Add user_id and user_name columns to monitoring_sessions table
+
+    This migration adds the missing user_id column and also ensures user_name
+    column exists (in case migration 21 failed or was skipped).
+    """
+
+    async def _table_exists(self, table_name: str) -> bool:
+        """Check if a table exists (works for both SQLite and PostgreSQL)."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT EXISTS (SELECT FROM information_schema.tables WHERE table_name = :table_name);'
+                ).bindparams(table_name=table_name)
+            )
+            return bool(result.scalar())
+        else:
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text("SELECT name FROM sqlite_master WHERE type='table' AND name=:table_name;").bindparams(
+                    table_name=table_name
+                )
+            )
+            return result.first() is not None
+
+    async def _get_table_columns(self, table_name: str) -> list[str]:
+        """Get column names from a table (works for both SQLite and PostgreSQL)."""
+        if self.ap.persistence_mgr.db.name == 'postgresql':
+            result = await self.ap.persistence_mgr.execute_async(
+                sqlalchemy.text(
+                    'SELECT column_name FROM information_schema.columns WHERE table_name = :table_name;'
+                ).bindparams(table_name=table_name)
+            )
+            return [row[0] for row in result.fetchall()]
+        else:
+            if not table_name.isidentifier():
+                raise ValueError(f'Invalid table name: {table_name}')
+            result = await self.ap.persistence_mgr.execute_async(sqlalchemy.text(f'PRAGMA table_info({table_name});'))
+            return [row[1] for row in result.fetchall()]
+
+    async def _add_column_if_not_exists(self, table_name: str, column_name: str, column_type: str):
+        """Add a column to a table if it does not already exist."""
+        columns = await self._get_table_columns(table_name)
+        if column_name in columns:
+            self.ap.logger.debug('%s column already exists in %s.', column_name, table_name)
+            return
+        await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text(f'ALTER TABLE {table_name} ADD COLUMN {column_name} {column_type};')
+        )
+        self.ap.logger.info('Added %s column to %s table.', column_name, table_name)
+
+    async def upgrade(self):
+        # Check if monitoring_sessions table exists
+        if not await self._table_exists('monitoring_sessions'):
+            self.ap.logger.warning('monitoring_sessions table does not exist, skipping migration.')
+            return
+
+        # Add user_id column to monitoring_sessions table
+        await self._add_column_if_not_exists('monitoring_sessions', 'user_id', 'VARCHAR(255)')
+
+        # Add user_name column to monitoring_sessions table (in case migration 21 failed)
+        await self._add_column_if_not_exists('monitoring_sessions', 'user_name', 'VARCHAR(255)')
+
+        # Add user_name column to monitoring_messages table (in case migration 21 failed)
+        if await self._table_exists('monitoring_messages'):
+            await self._add_column_if_not_exists('monitoring_messages', 'user_name', 'VARCHAR(255)')
+
+    async def downgrade(self):
+        pass

src/langbot/pkg/persistence/migrations/dbm023_model_fallback_config.py

@@ -0,0 +1,102 @@
+from .. import migration
+
+import sqlalchemy
+import json
+
+
+@migration.migration_class(23)
+class DBMigrateModelFallbackConfig(migration.DBMigration):
+    """Convert model field from plain UUID string to object with primary/fallbacks"""
+
+    async def upgrade(self):
+        """Upgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'ai' not in config or 'local-agent' not in config['ai']:
+                continue
+
+            local_agent = config['ai']['local-agent']
+            changed = False
+
+            # Convert model from string to object
+            model_value = local_agent.get('model', '')
+            if isinstance(model_value, str):
+                local_agent['model'] = {
+                    'primary': model_value,
+                    'fallbacks': [],
+                }
+                changed = True
+
+            # Remove leftover fallback-models field if present
+            if 'fallback-models' in local_agent:
+                del local_agent['fallback-models']
+                changed = True
+
+            if not changed:
+                continue
+
+            # Update using raw SQL with compatibility for both SQLite and PostgreSQL
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+
+    async def downgrade(self):
+        """Downgrade"""
+        result = await self.ap.persistence_mgr.execute_async(
+            sqlalchemy.text('SELECT uuid, config FROM legacy_pipelines')
+        )
+        pipelines = result.fetchall()
+
+        current_version = self.ap.ver_mgr.get_current_version()
+
+        for pipeline_row in pipelines:
+            uuid = pipeline_row[0]
+            config = json.loads(pipeline_row[1]) if isinstance(pipeline_row[1], str) else pipeline_row[1]
+
+            if 'ai' not in config or 'local-agent' not in config['ai']:
+                continue
+
+            local_agent = config['ai']['local-agent']
+
+            # Convert model from object back to string
+            model_value = local_agent.get('model', '')
+            if isinstance(model_value, dict):
+                local_agent['model'] = model_value.get('primary', '')
+            else:
+                continue
+
+            # Update using raw SQL with compatibility for both SQLite and PostgreSQL
+            if self.ap.persistence_mgr.db.name == 'postgresql':
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config::jsonb, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )
+            else:
+                await self.ap.persistence_mgr.execute_async(
+                    sqlalchemy.text(
+                        'UPDATE legacy_pipelines SET config = :config, for_version = :for_version WHERE uuid = :uuid'
+                    ),
+                    {'config': json.dumps(config), 'for_version': current_version, 'uuid': uuid},
+                )

src/langbot/pkg/pipeline/monitoring_helper.py

@@ -34,6 +34,15 @@ class MonitoringHelper:
             # Check if session exists, if not, record session start
             session_id = f'{query.launcher_type}_{query.launcher_id}'
 
+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
             # Try to record message
             # Use JSON serialization to preserve message chain structure (including image URLs, etc.)
             if hasattr(query, 'message_chain') and hasattr(query.message_chain, 'model_dump'):
@@ -57,6 +66,7 @@ class MonitoringHelper:
                 if hasattr(query.launcher_type, 'value')
                 else str(query.launcher_type),
                 user_id=query.sender_id,
+                user_name=sender_name,
                 runner_name=runner_name,
                 variables=None,  # Will be updated in record_query_success
             )
@@ -80,6 +90,7 @@ class MonitoringHelper:
                     if hasattr(query.launcher_type, 'value')
                     else str(query.launcher_type),
                     user_id=query.sender_id,
+                    user_name=sender_name,
                 )
 
             return message_id
@@ -128,6 +139,15 @@ class MonitoringHelper:
         try:
             session_id = f'{query.launcher_type}_{query.launcher_id}'
 
+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
             # Extract response content from resp_message_chain
             if hasattr(query, 'resp_message_chain') and query.resp_message_chain:
                 # Serialize the last response message chain
@@ -162,6 +182,7 @@ class MonitoringHelper:
                 if hasattr(query.launcher_type, 'value')
                 else str(query.launcher_type),
                 user_id=query.sender_id,
+                user_name=sender_name,
                 runner_name=runner_name,
                 role='assistant',
             )
@@ -183,6 +204,15 @@ class MonitoringHelper:
         try:
             session_id = f'{query.launcher_type}_{query.launcher_id}'
 
+            # Get sender name from message event
+            sender_name = None
+            if hasattr(query, 'message_event'):
+                if hasattr(query.message_event, 'sender'):
+                    if hasattr(query.message_event.sender, 'nickname'):
+                        sender_name = query.message_event.sender.nickname
+                    elif hasattr(query.message_event.sender, 'member_name'):
+                        sender_name = query.message_event.sender.member_name
+
             # Record error message
             message_id = await ap.monitoring_service.record_message(
                 bot_id=bot_id,
@@ -197,6 +227,7 @@ class MonitoringHelper:
                 if hasattr(query.launcher_type, 'value')
                 else str(query.launcher_type),
                 user_id=query.sender_id,
+                user_name=sender_name,
                 runner_name=runner_name,
             )
 

src/langbot/pkg/pipeline/preproc/preproc.py

@@ -36,17 +36,36 @@ class PreProcessor(stage.PipelineStage):
         session = await self.ap.sess_mgr.get_session(query)
 
         # When not local-agent, llm_model is None
-        try:
-            llm_model = (
-                await self.ap.model_mgr.get_model_by_uuid(query.pipeline_config['ai']['local-agent']['model'])
-                if selected_runner == 'local-agent'
-                else None
-            )
-        except ValueError:
-            self.ap.logger.warning(
-                f'LLM model {query.pipeline_config["ai"]["local-agent"]["model"] + " "}not found or not configured'
-            )
-            llm_model = None
+        llm_model = None
+        if selected_runner == 'local-agent':
+            # Read model config — new format is { primary: str, fallbacks: [str] },
+            # but handle legacy plain string for backward compatibility
+            model_config = query.pipeline_config['ai']['local-agent'].get('model', {})
+            if isinstance(model_config, str):
+                # Legacy format: plain UUID string
+                primary_uuid = model_config
+                fallback_uuids = []
+            else:
+                primary_uuid = model_config.get('primary', '')
+                fallback_uuids = model_config.get('fallbacks', [])
+
+            if primary_uuid:
+                try:
+                    llm_model = await self.ap.model_mgr.get_model_by_uuid(primary_uuid)
+                except ValueError:
+                    self.ap.logger.warning(f'LLM model {primary_uuid} not found or not configured')
+
+            # Resolve fallback model UUIDs
+            if fallback_uuids:
+                valid_fallbacks = []
+                for fb_uuid in fallback_uuids:
+                    try:
+                        await self.ap.model_mgr.get_model_by_uuid(fb_uuid)
+                        valid_fallbacks.append(fb_uuid)
+                    except ValueError:
+                        self.ap.logger.warning(f'Fallback model {fb_uuid} not found, skipping')
+                if valid_fallbacks:
+                    query.variables['_fallback_model_uuids'] = valid_fallbacks
 
         conversation = await self.ap.sess_mgr.get_conversation(
             query,
@@ -61,20 +80,28 @@ class PreProcessor(stage.PipelineStage):
         query.prompt = conversation.prompt.copy()
         query.messages = conversation.messages.copy()
 
-        if selected_runner == 'local-agent' and llm_model:
+        if selected_runner == 'local-agent':
             query.use_funcs = []
-            query.use_llm_model_uuid = llm_model.model_entity.uuid
+            if llm_model:
+                query.use_llm_model_uuid = llm_model.model_entity.uuid
 
-            if llm_model.model_entity.abilities.__contains__('func_call'):
-                # Get bound plugins and MCP servers for filtering tools
+                if llm_model.model_entity.abilities.__contains__('func_call'):
+                    # Get bound plugins and MCP servers for filtering tools
+                    bound_plugins = query.variables.get('_pipeline_bound_plugins', None)
+                    bound_mcp_servers = query.variables.get('_pipeline_bound_mcp_servers', None)
+                    query.use_funcs = await self.ap.tool_mgr.get_all_tools(bound_plugins, bound_mcp_servers)
+
+                    self.ap.logger.debug(f'Bound plugins: {bound_plugins}')
+                    self.ap.logger.debug(f'Bound MCP servers: {bound_mcp_servers}')
+                    self.ap.logger.debug(f'Use funcs: {query.use_funcs}')
+
+            # If primary model doesn't support func_call but fallback models exist,
+            # load tools anyway since fallback models may support them
+            if not query.use_funcs and query.variables.get('_fallback_model_uuids'):
                 bound_plugins = query.variables.get('_pipeline_bound_plugins', None)
                 bound_mcp_servers = query.variables.get('_pipeline_bound_mcp_servers', None)
                 query.use_funcs = await self.ap.tool_mgr.get_all_tools(bound_plugins, bound_mcp_servers)
 
-                self.ap.logger.debug(f'Bound plugins: {bound_plugins}')
-                self.ap.logger.debug(f'Bound MCP servers: {bound_mcp_servers}')
-                self.ap.logger.debug(f'Use funcs: {query.use_funcs}')
-
         sender_name = ''
 
         if isinstance(query.message_event, platform_events.GroupMessage):

src/langbot/pkg/pipeline/process/handlers/chat.py

@@ -149,12 +149,19 @@ class ChatMessageHandler(handler.MessageHandler):
                 self.ap.logger.error(f'Conversation({query.query_id}) Request Failed: {error_info}')
                 traceback.print_exc()
 
-                hide_exception_info = query.pipeline_config['output']['misc']['hide-exception']
+                exception_handling = query.pipeline_config['output']['misc'].get('exception-handling', 'show-hint')
+
+                if exception_handling == 'show-error':
+                    user_notice = f'{e}'
+                elif exception_handling == 'show-hint':
+                    user_notice = query.pipeline_config['output']['misc'].get('failure-hint', 'Request failed.')
+                else:  # hide
+                    user_notice = None
 
                 yield entities.StageProcessResult(
                     result_type=entities.ResultType.INTERRUPT,
                     new_query=query,
-                    user_notice='请求失败' if hide_exception_info else f'{e}',
+                    user_notice=user_notice,
                     error_notice=f'{e}',
                     debug_notice=traceback.format_exc(),
                 )

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

@@ -81,22 +81,33 @@ class WecomEventConverter(abstract_platform_adapter.AbstractEventConverter):
             return event.source_platform_object
 
     @staticmethod
-    async def target2yiri(event: WecomCSEvent):
+    async def target2yiri(event: WecomCSEvent, bot: WecomCSClient = None):
         """
         将 WecomEvent 转换为平台的 FriendMessage 对象。
 
         Args:
             event (WecomEvent): 企业微信客服事件。
+            bot (WecomCSClient): 企业微信客服客户端，用于获取用户信息。
 
         Returns:
             platform_events.FriendMessage: 转换后的 FriendMessage 对象。
         """
+        # Try to get customer nickname from WeChat API
+        nickname = str(event.user_id)
+        if bot and event.user_id:
+            try:
+                customer_info = await bot.get_customer_info(event.user_id)
+                if customer_info and customer_info.get('nickname'):
+                    nickname = customer_info.get('nickname')
+            except Exception:
+                pass  # Fall back to user_id as nickname
+
         # 转换消息链
         if event.type == 'text':
             yiri_chain = await WecomMessageConverter.target2yiri(event.message, event.message_id)
             friend = platform_entities.Friend(
                 id=f'u{event.user_id}',
-                nickname=str(event.user_id),
+                nickname=nickname,
                 remark='',
             )
 
@@ -106,7 +117,7 @@ class WecomEventConverter(abstract_platform_adapter.AbstractEventConverter):
         elif event.type == 'image':
             friend = platform_entities.Friend(
                 id=f'u{event.user_id}',
-                nickname=str(event.user_id),
+                nickname=nickname,
                 remark='',
             )
 
@@ -187,7 +198,7 @@ class WecomCSAdapter(abstract_platform_adapter.AbstractMessagePlatformAdapter):
         async def on_message(event: WecomCSEvent):
             self.bot_account_id = event.receiver_id
             try:
-                return await callback(await self.event_converter.target2yiri(event), self)
+                return await callback(await self.event_converter.target2yiri(event, self.bot), self)
             except Exception:
                 await self.logger.error(f'Error in wecomcs callback: {traceback.format_exc()}')
 

src/langbot/pkg/provider/runners/difysvapi.py

@@ -441,6 +441,7 @@ class DifyServiceAPIRunner(runner.RequestRunner):
         is_final = False
         think_start = False
         think_end = False
+        yielded_final = False
 
         remove_think = self.pipeline_config['output'].get('misc', '').get('remove-think')
 
@@ -493,13 +494,19 @@ class DifyServiceAPIRunner(runner.RequestRunner):
                     if answer:
                         basic_mode_pending_chunk = answer
 
-            if (is_final or message_idx % 8 == 0) and (basic_mode_pending_chunk != '' or is_final):
+            if (
+                not yielded_final
+                and (is_final or message_idx % 8 == 0)
+                and (basic_mode_pending_chunk != '' or is_final)
+            ):
                 # content, _ = self._process_thinking_content(basic_mode_pending_chunk)
                 yield provider_message.MessageChunk(
                     role='assistant',
                     content=basic_mode_pending_chunk,
                     is_final=is_final,
                 )
+                if is_final:
+                    yielded_final = True
 
         if chunk is None:
             raise errors.DifyAPIError('Dify API 没有返回任何响应，请检查网络连接和API配置')

src/langbot/pkg/provider/runners/localagent.py

@@ -4,6 +4,7 @@ import json
 import copy
 import typing
 from .. import runner
+from ..modelmgr import requester as modelmgr_requester
 import langbot_plugin.api.entities.builtin.pipeline.query as pipeline_query
 import langbot_plugin.api.entities.builtin.provider.message as provider_message
 import langbot_plugin.api.entities.builtin.rag.context as rag_context
@@ -26,21 +27,117 @@ Respond in the same language as the user's input.
 
 @runner.runner_class('local-agent')
 class LocalAgentRunner(runner.RequestRunner):
-    """本地Agent请求运行器"""
+    """Local agent request runner"""
 
-    class ToolCallTracker:
-        """工具调用追踪器"""
+    async def _get_model_candidates(
+        self,
+        query: pipeline_query.Query,
+    ) -> list[modelmgr_requester.RuntimeLLMModel]:
+        """Build ordered list of models to try: primary model + fallback models."""
+        candidates = []
 
-        def __init__(self):
-            self.active_calls: dict[str, dict] = {}
-            self.completed_calls: list[provider_message.ToolCall] = []
+        # Primary model
+        if query.use_llm_model_uuid:
+            try:
+                primary = await self.ap.model_mgr.get_model_by_uuid(query.use_llm_model_uuid)
+                candidates.append(primary)
+            except ValueError:
+                self.ap.logger.warning(f'Primary model {query.use_llm_model_uuid} not found')
+
+        # Fallback models
+        fallback_uuids = (query.variables or {}).get('_fallback_model_uuids', [])
+        for fb_uuid in fallback_uuids:
+            try:
+                fb_model = await self.ap.model_mgr.get_model_by_uuid(fb_uuid)
+                candidates.append(fb_model)
+            except ValueError:
+                self.ap.logger.warning(f'Fallback model {fb_uuid} not found, skipping')
+
+        return candidates
+
+    async def _invoke_with_fallback(
+        self,
+        query: pipeline_query.Query,
+        candidates: list[modelmgr_requester.RuntimeLLMModel],
+        messages: list,
+        funcs: list,
+        remove_think: bool,
+    ) -> tuple[provider_message.Message, modelmgr_requester.RuntimeLLMModel]:
+        """Try non-streaming invocation with sequential fallback. Returns (message, model_used)."""
+        last_error = None
+        for model in candidates:
+            try:
+                msg = await model.provider.invoke_llm(
+                    query,
+                    model,
+                    messages,
+                    funcs if model.model_entity.abilities.__contains__('func_call') else [],
+                    extra_args=model.model_entity.extra_args,
+                    remove_think=remove_think,
+                )
+                return msg, model
+            except Exception as e:
+                last_error = e
+                self.ap.logger.warning(f'Model {model.model_entity.name} failed: {e}, trying next fallback...')
+        raise last_error or RuntimeError('No model candidates available')
+
+    async def _invoke_stream_with_fallback(
+        self,
+        query: pipeline_query.Query,
+        candidates: list[modelmgr_requester.RuntimeLLMModel],
+        messages: list,
+        funcs: list,
+        remove_think: bool,
+    ) -> tuple[typing.AsyncGenerator, modelmgr_requester.RuntimeLLMModel]:
+        """Try streaming invocation with sequential fallback. Returns (stream_generator, model_used).
+
+        Fallback is only possible before any chunks have been yielded to the client.
+        Once streaming starts, the model is committed.
+        """
+        last_error = None
+        for model in candidates:
+            try:
+                stream = model.provider.invoke_llm_stream(
+                    query,
+                    model,
+                    messages,
+                    funcs if model.model_entity.abilities.__contains__('func_call') else [],
+                    extra_args=model.model_entity.extra_args,
+                    remove_think=remove_think,
+                )
+                # Attempt to get the first chunk to verify the stream works
+                first_chunk = await stream.__anext__()
+
+                async def _chain_stream(first, rest):
+                    yield first
+                    async for chunk in rest:
+                        yield chunk
+
+                return _chain_stream(first_chunk, stream), model
+            except StopAsyncIteration:
+                # Empty stream — treat as success (model returned nothing)
+                async def _empty_stream():
+                    return
+                    yield  # make it a generator
+
+                return _empty_stream(), model
+            except Exception as e:
+                last_error = e
+                self.ap.logger.warning(f'Model {model.model_entity.name} stream failed: {e}, trying next fallback...')
+        raise last_error or RuntimeError('No model candidates available')
 
     async def run(
         self, query: pipeline_query.Query
     ) -> typing.AsyncGenerator[provider_message.Message | provider_message.MessageChunk, None]:
-        """运行请求"""
+        """Run request"""
         pending_tool_calls = []
 
+        # Agent loop protection config
+        agent_config = query.pipeline_config['ai']['local-agent']
+        max_tool_iterations = agent_config.get('max-tool-iterations', 16)
+        max_tool_result_chars = agent_config.get('max-tool-result-chars', 8000)
+        iteration_count = 0
+
         # Get knowledge bases list (new field)
         kb_uuids = query.pipeline_config['ai']['local-agent'].get('knowledge-bases', [])
 
@@ -74,7 +171,13 @@ class LocalAgentRunner(runner.RequestRunner):
                     self.ap.logger.warning(f'Knowledge base {kb_uuid} not found, skipping')
                     continue
 
-                result = await kb.retrieve(user_message_text)
+                result = await kb.retrieve(
+                    user_message_text,
+                    settings={
+                        'sender_id': str(query.sender_id),
+                        'session_name': f'{query.session.launcher_type.value}_{query.session.launcher_id}',
+                    },
+                )
 
                 if result:
                     all_results.extend(result)
@@ -113,51 +216,51 @@ class LocalAgentRunner(runner.RequestRunner):
 
         remove_think = query.pipeline_config['output'].get('misc', '').get('remove-think')
 
-        use_llm_model = await self.ap.model_mgr.get_model_by_uuid(query.use_llm_model_uuid)
+        # Build ordered candidate list (primary + fallbacks)
+        candidates = await self._get_model_candidates(query)
+        if not candidates:
+            raise RuntimeError('No LLM model configured for local-agent runner')
 
         self.ap.logger.debug(
-            f'localagent req: query={query.query_id} req_messages={req_messages} use_llm_model={query.use_llm_model_uuid}'
+            f'localagent req: query={query.query_id} req_messages={req_messages} '
+            f'candidates={[m.model_entity.name for m in candidates]}'
         )
 
         if not is_stream:
-            # 非流式输出，直接请求
-
-            msg = await use_llm_model.provider.invoke_llm(
+            # Non-streaming: invoke with fallback
+            msg, use_llm_model = await self._invoke_with_fallback(
                 query,
-                use_llm_model,
+                candidates,
                 req_messages,
                 query.use_funcs,
-                extra_args=use_llm_model.model_entity.extra_args,
-                remove_think=remove_think,
+                remove_think,
             )
             yield msg
             final_msg = msg
         else:
-            # 流式输出，需要处理工具调用
+            # Streaming: invoke with fallback
             tool_calls_map: dict[str, provider_message.ToolCall] = {}
             msg_idx = 0
-            accumulated_content = ''  # 从开始累积的所有内容
+            accumulated_content = ''
             last_role = 'assistant'
             msg_sequence = 1
-            async for msg in use_llm_model.provider.invoke_llm_stream(
+
+            stream_src, use_llm_model = await self._invoke_stream_with_fallback(
                 query,
-                use_llm_model,
+                candidates,
                 req_messages,
                 query.use_funcs,
-                extra_args=use_llm_model.model_entity.extra_args,
-                remove_think=remove_think,
-            ):
+                remove_think,
+            )
+            async for msg in stream_src:
                 msg_idx = msg_idx + 1
 
-                # 记录角色
                 if msg.role:
                     last_role = msg.role
 
-                # 累积内容
                 if msg.content:
                     accumulated_content += msg.content
 
-                # 处理工具调用
                 if msg.tool_calls:
                     for tool_call in msg.tool_calls:
                         if tool_call.id not in tool_calls_map:
@@ -169,21 +272,18 @@ class LocalAgentRunner(runner.RequestRunner):
                                 ),
                             )
                         if tool_call.function and tool_call.function.arguments:
-                            # 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
                             tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments
-                # continue
-                # 每8个chunk或最后一个chunk时，输出所有累积的内容
+
                 if msg_idx % 8 == 0 or msg.is_final:
                     msg_sequence += 1
                     yield provider_message.MessageChunk(
                         role=last_role,
-                        content=accumulated_content,  # 输出所有累积内容
+                        content=accumulated_content,
                         tool_calls=list(tool_calls_map.values()) if (tool_calls_map and msg.is_final) else None,
                         is_final=msg.is_final,
                         msg_sequence=msg_sequence,
                     )
 
-            # 创建最终消息用于后续处理
             final_msg = provider_message.MessageChunk(
                 role=last_role,
                 content=accumulated_content,
@@ -198,8 +298,17 @@ class LocalAgentRunner(runner.RequestRunner):
 
         req_messages.append(final_msg)
 
-        # 持续请求，只要还有待处理的工具调用就继续处理调用
+        # Once a model succeeds, commit to it for the tool call loop
+        # (no fallback mid-conversation — different models may interpret tool results differently)
         while pending_tool_calls:
+            iteration_count += 1
+            if iteration_count > max_tool_iterations:
+                self.ap.logger.warning(
+                    f'localagent: query={query.query_id} agent loop exceeded max iterations ({max_tool_iterations}), '
+                    f'forcing termination'
+                )
+                break
+
             for tool_call in pending_tool_calls:
                 try:
                     func = tool_call.function
@@ -222,6 +331,14 @@ class LocalAgentRunner(runner.RequestRunner):
                     else:
                         tool_content = json.dumps(func_ret, ensure_ascii=False)
 
+                    # Truncate oversized tool results to prevent context overflow
+                    if isinstance(tool_content, str) and len(tool_content) > max_tool_result_chars:
+                        self.ap.logger.warning(
+                            f'localagent: tool {func.name} returned {len(tool_content)} chars, '
+                            f'truncating to {max_tool_result_chars}'
+                        )
+                        tool_content = tool_content[:max_tool_result_chars] + '\n...[result truncated]'
+
                     if is_stream:
                         msg = provider_message.MessageChunk(
                             role='tool',
@@ -239,7 +356,6 @@ class LocalAgentRunner(runner.RequestRunner):
 
                     req_messages.append(msg)
                 except Exception as e:
-                    # 工具调用出错，添加一个报错信息到 req_messages
                     err_msg = provider_message.Message(role='tool', content=f'err: {e}', tool_call_id=tool_call.id)
 
                     yield err_msg
@@ -247,39 +363,38 @@ class LocalAgentRunner(runner.RequestRunner):
                     req_messages.append(err_msg)
 
             self.ap.logger.debug(
-                f'localagent req: query={query.query_id} req_messages={req_messages} use_llm_model={query.use_llm_model_uuid}'
+                f'localagent req: query={query.query_id} req_messages={req_messages} '
+                f'use_llm_model={use_llm_model.model_entity.name}'
             )
 
             if is_stream:
                 tool_calls_map = {}
                 msg_idx = 0
-                accumulated_content = ''  # 从开始累积的所有内容
+                accumulated_content = ''
                 last_role = 'assistant'
                 msg_sequence = first_end_sequence
 
-                async for msg in use_llm_model.provider.invoke_llm_stream(
+                tool_stream_src = use_llm_model.provider.invoke_llm_stream(
                     query,
                     use_llm_model,
                     req_messages,
-                    query.use_funcs,
+                    query.use_funcs if use_llm_model.model_entity.abilities.__contains__('func_call') else [],
                     extra_args=use_llm_model.model_entity.extra_args,
                     remove_think=remove_think,
-                ):
+                )
+                async for msg in tool_stream_src:
                     msg_idx += 1
 
-                    # 记录角色
                     if msg.role:
                         last_role = msg.role
 
-                    # 第一次请求工具调用时的内容
+                    # Prepend first-round content on first chunk of tool-call round
                     if msg_idx == 1:
                         accumulated_content = first_content if first_content is not None else accumulated_content
 
-                    # 累积内容
                     if msg.content:
                         accumulated_content += msg.content
 
-                    # 处理工具调用
                     if msg.tool_calls:
                         for tool_call in msg.tool_calls:
                             if tool_call.id not in tool_calls_map:
@@ -291,15 +406,13 @@ class LocalAgentRunner(runner.RequestRunner):
                                     ),
                                 )
                             if tool_call.function and tool_call.function.arguments:
-                                # 流式处理中，工具调用参数可能分多个chunk返回，需要追加而不是覆盖
                                 tool_calls_map[tool_call.id].function.arguments += tool_call.function.arguments
 
-                    # 每8个chunk或最后一个chunk时，输出所有累积的内容
                     if msg_idx % 8 == 0 or msg.is_final:
                         msg_sequence += 1
                         yield provider_message.MessageChunk(
                             role=last_role,
-                            content=accumulated_content,  # 输出所有累积内容
+                            content=accumulated_content,
                             tool_calls=list(tool_calls_map.values()) if (tool_calls_map and msg.is_final) else None,
                             is_final=msg.is_final,
                             msg_sequence=msg_sequence,
@@ -312,12 +425,12 @@ class LocalAgentRunner(runner.RequestRunner):
                     msg_sequence=msg_sequence,
                 )
             else:
-                # 处理完所有调用，再次请求
+                # Non-streaming: use committed model directly (no fallback in tool loop)
                 msg = await use_llm_model.provider.invoke_llm(
                     query,
                     use_llm_model,
                     req_messages,
-                    query.use_funcs,
+                    query.use_funcs if use_llm_model.model_entity.abilities.__contains__('func_call') else [],
                     extra_args=use_llm_model.model_entity.extra_args,
                     remove_think=remove_think,
                 )

src/langbot/pkg/rag/knowledge/kbmgr.py

@@ -321,13 +321,19 @@ class RuntimeKnowledgeBase(KnowledgeBaseInterface):
         if not plugin_id:
             raise ValueError(f'No RAG plugin ID configured for KB {kb.uuid}. Retrieval failed.')
 
+        # Session context (e.g. session_name) stays in retrieval_settings
+        # for plugins that need it. Do NOT move them into filters, as filters
+        # are passed directly to vector_search by some plugins (e.g. LangRAG)
+        # and would cause empty results when the metadata field doesn't exist.
+        filters = settings.pop('filters', {})
+
         retrieval_context = {
             'query': query,
             'knowledge_base_id': kb.uuid,
             'collection_id': kb.collection_id or kb.uuid,
             'retrieval_settings': settings,
             'creation_settings': kb.creation_settings or {},
-            'filters': settings.pop('filters', {}),
+            'filters': filters,
         }
 
         result = await self.ap.plugin_connector.call_rag_retrieve(

src/langbot/pkg/utils/constants.py

@@ -2,7 +2,7 @@ import langbot
 
 semantic_version = f'v{langbot.__version__}'
 
-required_database_version = 20
+required_database_version = 23
 """Tag the version of the database schema, used to check if the database needs to be migrated"""
 
 debug_mode = False

src/langbot/pkg/vector/vdbs/chroma.py

@@ -2,11 +2,14 @@ from __future__ import annotations
 import asyncio
 from typing import Any
 from chromadb import PersistentClient
-from langbot.pkg.vector.vdb import VectorDatabase
+from langbot.pkg.vector.vdb import VectorDatabase, SearchType
 from langbot.pkg.core import app
 import chromadb
 import chromadb.errors
 
+# RRF smoothing constant (standard value from the literature)
+_RRF_K = 60
+
 
 class ChromaVectorDatabase(VectorDatabase):
     def __init__(self, ap: app.Application, base_path: str = './data/chroma'):
@@ -14,6 +17,10 @@ class ChromaVectorDatabase(VectorDatabase):
         self.client = PersistentClient(path=base_path)
         self._collections = {}
 
+    @classmethod
+    def supported_search_types(cls) -> list[SearchType]:
+        return [SearchType.VECTOR, SearchType.FULL_TEXT, SearchType.HYBRID]
+
     async def get_or_create_collection(self, collection: str) -> chromadb.Collection:
         if collection not in self._collections:
             self._collections[collection] = await asyncio.to_thread(
@@ -34,8 +41,8 @@ class ChromaVectorDatabase(VectorDatabase):
         kwargs: dict[str, Any] = dict(embeddings=embeddings_list, ids=ids, metadatas=metadatas)
         if documents is not None:
             kwargs['documents'] = documents
-        await asyncio.to_thread(col.add, **kwargs)
-        self.ap.logger.info(f"Added {len(ids)} embeddings to Chroma collection '{collection}'.")
+        await asyncio.to_thread(col.upsert, **kwargs)
+        self.ap.logger.info(f"Upserted {len(ids)} embeddings to Chroma collection '{collection}'.")
 
     async def search(
         self,
@@ -47,6 +54,23 @@ class ChromaVectorDatabase(VectorDatabase):
         filter: dict[str, Any] | None = None,
     ) -> dict[str, Any]:
         col = await self.get_or_create_collection(collection)
+
+        if search_type == SearchType.FULL_TEXT:
+            return await self._full_text_search(col, collection, k, query_text, filter)
+        elif search_type == SearchType.HYBRID:
+            return await self._hybrid_search(col, collection, query_embedding, k, query_text, filter)
+
+        # Default: vector search
+        return await self._vector_search(col, collection, query_embedding, k, filter)
+
+    async def _vector_search(
+        self,
+        col: chromadb.Collection,
+        collection: str,
+        query_embedding: list[float],
+        k: int,
+        filter: dict[str, Any] | None,
+    ) -> dict[str, Any]:
         query_kwargs: dict[str, Any] = dict(
             query_embeddings=query_embedding,
             n_results=k,
@@ -55,9 +79,137 @@ class ChromaVectorDatabase(VectorDatabase):
         if filter:
             query_kwargs['where'] = filter
         results = await asyncio.to_thread(col.query, **query_kwargs)
-        self.ap.logger.info(f"Chroma search in '{collection}' returned {len(results.get('ids', [[]])[0])} results.")
+        self.ap.logger.info(
+            f"Chroma vector search in '{collection}' returned {len(results.get('ids', [[]])[0])} results."
+        )
         return results
 
+    async def _full_text_search(
+        self,
+        col: chromadb.Collection,
+        collection: str,
+        k: int,
+        query_text: str,
+        filter: dict[str, Any] | None,
+    ) -> dict[str, Any]:
+        if not query_text:
+            return {'ids': [[]], 'metadatas': [[]], 'distances': [[]], 'documents': [[]]}
+
+        get_kwargs: dict[str, Any] = dict(
+            where_document={'$contains': query_text},
+            include=['metadatas', 'documents'],
+            limit=k,
+        )
+        if filter:
+            get_kwargs['where'] = filter
+        results = await asyncio.to_thread(col.get, **get_kwargs)
+
+        # col.get returns flat lists; wrap into column-major format.
+        # Distances are all 0.0 because Chroma's local $contains is a boolean
+        # filter with no relevance scoring.  Chroma's BM25 sparse embedding
+        # function (ChromaBm25EmbeddingFunction) can generate scored sparse
+        # vectors, but sparse vector *indexing* is only available on Chroma
+        # Cloud, not locally.  For ranked results, use hybrid mode or apply a
+        # reranker in a downstream stage.
+        ids = results.get('ids', [])
+        metadatas = results.get('metadatas', []) or [None] * len(ids)
+        documents = results.get('documents', []) or [None] * len(ids)
+        distances = [0.0] * len(ids)
+
+        self.ap.logger.info(f"Chroma full-text search in '{collection}' returned {len(ids)} results.")
+        return {'ids': [ids], 'metadatas': [metadatas], 'distances': [distances], 'documents': [documents]}
+
+    async def _hybrid_search(
+        self,
+        col: chromadb.Collection,
+        collection: str,
+        query_embedding: list[float],
+        k: int,
+        query_text: str,
+        filter: dict[str, Any] | None,
+    ) -> dict[str, Any]:
+        # Fall back to pure vector search when no text is provided
+        if not query_text:
+            return await self._vector_search(col, collection, query_embedding, k, filter)
+
+        # Run vector search and full-text search in parallel
+        vector_task = self._vector_search(col, collection, query_embedding, k, filter)
+        text_task = self._full_text_search(col, collection, k, query_text, filter)
+        vector_results, text_results = await asyncio.gather(vector_task, text_task)
+
+        vector_ids = vector_results.get('ids', [[]])[0]
+        text_ids = text_results.get('ids', [[]])[0]
+
+        if not vector_ids and not text_ids:
+            return {'ids': [[]], 'metadatas': [[]], 'distances': [[]], 'documents': [[]]}
+
+        # RRF fusion
+        fused = self._rrf_fuse([vector_ids, text_ids], k)
+        if not fused:
+            return {'ids': [[]], 'metadatas': [[]], 'distances': [[]], 'documents': [[]]}
+
+        fused_ids = [doc_id for doc_id, _ in fused]
+
+        # Fetch full metadata and documents for fused results
+        fetched = await asyncio.to_thread(col.get, ids=fused_ids, include=['metadatas', 'documents'])
+
+        # col.get returns results in arbitrary order; re-order to match fused ranking
+        fetched_map: dict[str, tuple] = {}
+        for i, fid in enumerate(fetched.get('ids', [])):
+            meta = (fetched.get('metadatas') or [None] * len(fetched['ids']))[i]
+            doc = (fetched.get('documents') or [None] * len(fetched['ids']))[i]
+            fetched_map[fid] = (meta, doc)
+
+        ordered_ids = []
+        ordered_metas = []
+        ordered_docs = []
+        ordered_dists = []
+
+        # Normalize RRF scores to 0~1 distances via min-max scaling.
+        # Raw RRF scores are tiny (e.g. 0.016~0.033 with k=60) so a naive
+        # ``1 - score`` would compress all distances into a narrow 0.96~0.98
+        # band with almost no discriminative power.  Min-max normalization
+        # spreads them across the full 0~1 range (0.0 = best match).
+        max_score = fused[0][1]
+        min_score = fused[-1][1]
+        score_range = max_score - min_score
+
+        for doc_id, score in fused:
+            if doc_id in fetched_map:
+                meta, doc = fetched_map[doc_id]
+                ordered_ids.append(doc_id)
+                ordered_metas.append(meta)
+                ordered_docs.append(doc)
+                if score_range > 0:
+                    ordered_dists.append(1.0 - (score - min_score) / score_range)
+                else:
+                    ordered_dists.append(0.0)
+
+        self.ap.logger.info(
+            f"Chroma hybrid search in '{collection}' returned {len(ordered_ids)} results "
+            f'(vector={len(vector_ids)}, text={len(text_ids)}).'
+        )
+        return {
+            'ids': [ordered_ids],
+            'metadatas': [ordered_metas],
+            'distances': [ordered_dists],
+            'documents': [ordered_docs],
+        }
+
+    @staticmethod
+    def _rrf_fuse(result_lists: list[list[str]], k: int) -> list[tuple[str, float]]:
+        """Reciprocal Rank Fusion over multiple ranked ID lists.
+
+        Returns a list of (doc_id, rrf_score) sorted by descending score,
+        truncated to *k* entries.
+        """
+        scores: dict[str, float] = {}
+        for ranked_ids in result_lists:
+            for rank, doc_id in enumerate(ranked_ids):
+                scores[doc_id] = scores.get(doc_id, 0.0) + 1.0 / (_RRF_K + rank + 1)
+        sorted_results = sorted(scores.items(), key=lambda x: x[1], reverse=True)
+        return sorted_results[:k]
+
     async def delete_by_file_id(self, collection: str, file_id: str) -> None:
         col = await self.get_or_create_collection(collection)
         await asyncio.to_thread(col.delete, where={'file_id': file_id})

src/langbot/templates/default-pipeline-config.json

@@ -95,11 +95,12 @@
             "max": 0
         },
         "misc": {
-            "hide-exception": true,
+            "exception-handling": "show-hint",
+            "failure-hint": "Request failed.",
             "at-sender": true,
             "quote-origin": true,
             "track-function-calls": false,
             "remove-think": false
         }
     }
-}
+}

src/langbot/templates/metadata/pipeline/ai.yaml

@@ -59,8 +59,11 @@ stages:
         label:
           en_US: Model
           zh_Hans: 模型
-        type: llm-model-selector
+        type: model-fallback-selector
         required: true
+        default:
+          primary: ''
+          fallbacks: []
       - name: max-round
         label:
           en_US: Max Round
@@ -90,6 +93,26 @@ stages:
         type: knowledge-base-multi-selector
         required: false
         default: []
+      - name: max-tool-iterations
+        label:
+          en_US: Max Tool Iterations
+          zh_Hans: 最大工具调用轮次
+        description:
+          en_US: Maximum number of tool call iterations in a single agent loop to prevent runaway loops
+          zh_Hans: 单次 Agent 循环中工具调用的最大轮次，防止无限循环
+        type: integer
+        required: false
+        default: 16
+      - name: max-tool-result-chars
+        label:
+          en_US: Max Tool Result Length
+          zh_Hans: 工具返回最大字符数
+        description:
+          en_US: Maximum character length of a single tool call result, longer results will be truncated
+          zh_Hans: 单次工具调用返回结果的最大字符数，超出部分将被截断
+        type: integer
+        required: false
+        default: 8000
   - name: tbox-app-api
     label:
       en_US: Tbox App API

src/langbot/templates/metadata/pipeline/output.yaml

@@ -78,13 +78,39 @@ stages:
       en_US: Misc
       zh_Hans: 杂项
     config:
-      - name: hide-exception
+      - name: exception-handling
         label:
-          en_US: Hide Exception
-          zh_Hans: 不输出异常信息给用户
-        type: boolean
+          en_US: Exception Handling Strategy
+          zh_Hans: 异常处理策略
+        description:
+          en_US: Controls how error messages are displayed to the user when an AI request fails
+          zh_Hans: 控制 AI 请求失败时向用户展示错误信息的方式
+        type: select
         required: true
-        default: true
+        default: show-hint
+        options:
+          - name: show-error
+            label:
+              en_US: Show Full Error
+              zh_Hans: 显示完整报错信息
+          - name: show-hint
+            label:
+              en_US: Show Failure Hint
+              zh_Hans: 仅文字提示
+          - name: hide
+            label:
+              en_US: Hide All
+              zh_Hans: 不显示任何异常信息
+      - name: failure-hint
+        label:
+          en_US: Failure Hint Text
+          zh_Hans: 失败提示文本
+        description:
+          en_US: The text to display when a request fails. Only effective when Exception Handling Strategy is set to "Show Failure Hint"
+          zh_Hans: 请求失败时显示的提示文本，仅在异常处理策略设置为"仅文字提示"时生效
+        type: string
+        required: false
+        default: 'Request failed.'
       - name: at-sender
         label:
           en_US: At Sender
@@ -119,3 +145,4 @@ stages:
         type: boolean
         required: true
         default: false
+

web/src/app/home/bots/components/bot-form/BotForm.tsx

@@ -124,12 +124,6 @@ export default function BotForm({
   const currentAdapter = form.watch('adapter');
   const currentAdapterConfig = form.watch('adapter_config');
 
-  // Serialize adapter_config to a stable string so it can be used as a
-  // useEffect dependency without triggering on every render.  form.watch()
-  // returns a new object reference each time, which would otherwise cause
-  // the filtering effect below to loop indefinitely.
-  const adapterConfigJson = JSON.stringify(currentAdapterConfig);
-
   useEffect(() => {
     setBotFormValues();
   }, []);
@@ -153,7 +147,7 @@ export default function BotForm({
       // For non-Lark adapters, show all fields
       setFilteredDynamicFormConfigList(dynamicFormConfigList);
     }
-  }, [currentAdapter, adapterConfigJson, dynamicFormConfigList]);
+  }, [currentAdapter, currentAdapterConfig, dynamicFormConfigList]);
 
   // 复制到剪贴板的辅助函数 - 使用页面上的真实input元素
   const copyToClipboard = () => {

web/src/app/home/bots/components/bot-session/BotSessionMonitor.tsx

@@ -6,6 +6,7 @@ import { httpClient } from '@/app/infra/http/HttpClient';
 import { ScrollArea } from '@/components/ui/scroll-area';
 import { Button } from '@/components/ui/button';
 import { cn } from '@/lib/utils';
+import { Copy, Check } from 'lucide-react';
 import {
   MessageChainComponent,
   Plain,
@@ -27,6 +28,7 @@ interface SessionInfo {
   is_active: boolean;
   platform?: string | null;
   user_id?: string | null;
+  user_name?: string | null;
 }
 
 interface SessionMessage {
@@ -60,8 +62,29 @@ export default function BotSessionMonitor({ botId }: BotSessionMonitorProps) {
   const [messages, setMessages] = useState<SessionMessage[]>([]);
   const [loadingSessions, setLoadingSessions] = useState(false);
   const [loadingMessages, setLoadingMessages] = useState(false);
+  const [copiedUserId, setCopiedUserId] = useState(false);
   const messagesContainerRef = useRef<HTMLDivElement>(null);
 
+  const parseSessionType = (sessionId: string): string | null => {
+    const idx = sessionId.indexOf('_');
+    if (idx === -1) return null;
+    const type = sessionId.slice(0, idx);
+    if (type === 'person' || type === 'group') return type;
+    return null;
+  };
+
+  const abbreviateId = (id: string): string => {
+    if (id.length <= 10) return id;
+    return `${id.slice(0, 4)}..${id.slice(-4)}`;
+  };
+
+  const copyUserId = (userId: string) => {
+    navigator.clipboard.writeText(userId).then(() => {
+      setCopiedUserId(true);
+      setTimeout(() => setCopiedUserId(false), 2000);
+    });
+  };
+
   const loadSessions = useCallback(async () => {
     setLoadingSessions(true);
     try {
@@ -338,24 +361,36 @@ export default function BotSessionMonitor({ botId }: BotSessionMonitorProps) {
                   >
                     <div className="flex items-center justify-between mb-0.5">
                       <span className="text-sm font-medium truncate mr-2">
-                        {session.user_id || session.session_id.slice(0, 12)}
+                        {session.user_name ||
+                          session.user_id ||
+                          session.session_id.slice(0, 12)}
                       </span>
                       <span className="text-[11px] text-muted-foreground tabular-nums flex-shrink-0">
                         {formatRelativeTime(session.last_activity)}
                       </span>
                     </div>
                     <div className="flex items-center gap-1.5 text-xs text-muted-foreground">
+                      {parseSessionType(session.session_id) && (
+                        <span className="px-1 py-0.5 rounded bg-muted text-[10px]">
+                          {parseSessionType(session.session_id)}
+                        </span>
+                      )}
                       {session.platform && (
                         <span className="px-1 py-0.5 rounded bg-muted text-[10px]">
                           {session.platform}
                         </span>
                       )}
+                      {session.user_id && (
+                        <span className="truncate text-[10px]">
+                          {abbreviateId(session.user_id)}
+                        </span>
+                      )}
                       {session.is_active && (
                         <span className="flex items-center gap-0.5 text-green-600 dark:text-green-400">
                           <span className="w-1.5 h-1.5 rounded-full bg-green-500 inline-block" />
                         </span>
                       )}
-                      <span>{session.pipeline_name}</span>
+                      <span className="truncate">{session.pipeline_name}</span>
                     </div>
                   </button>
                 );
@@ -377,15 +412,42 @@ export default function BotSessionMonitor({ botId }: BotSessionMonitorProps) {
             <div className="px-6 py-3 border-b shrink-0 flex items-center justify-between">
               <div className="min-w-0">
                 <div className="text-sm font-medium truncate">
-                  {selectedSession?.user_id || selectedSessionId.slice(0, 20)}
+                  {selectedSession?.user_name ||
+                    selectedSession?.user_id ||
+                    selectedSessionId.slice(0, 20)}
                 </div>
                 <div className="flex items-center gap-2 text-xs text-muted-foreground">
+                  {parseSessionType(selectedSessionId) && (
+                    <span>{parseSessionType(selectedSessionId)}</span>
+                  )}
                   {selectedSession?.platform && (
-                    <span>{selectedSession.platform}</span>
+                    <>
+                      {parseSessionType(selectedSessionId) && <span>·</span>}
+                      <span>{selectedSession.platform}</span>
+                    </>
+                  )}
+                  {selectedSession?.user_id && (
+                    <>
+                      <span>·</span>
+                      <span className="font-mono">
+                        {selectedSession.user_id}
+                      </span>
+                      <button
+                        onClick={() => copyUserId(selectedSession.user_id!)}
+                        className="inline-flex items-center text-muted-foreground hover:text-foreground transition-colors"
+                        title={t('common.copy')}
+                      >
+                        {copiedUserId ? (
+                          <Check className="w-3 h-3 text-green-600" />
+                        ) : (
+                          <Copy className="w-3 h-3" />
+                        )}
+                      </button>
+                    </>
                   )}
                   {selectedSession?.pipeline_name && (
                     <>
-                      {selectedSession?.platform && <span>·</span>}
+                      <span>·</span>
                       <span>{selectedSession.pipeline_name}</span>
                     </>
                   )}

web/src/app/home/components/dynamic-form/DynamicFormComponent.tsx

@@ -11,7 +11,7 @@ import {
   FormMessage,
 } from '@/components/ui/form';
 import DynamicFormItemComponent from '@/app/home/components/dynamic-form/DynamicFormItemComponent';
-import { useCallback, useEffect, useRef } from 'react';
+import { useEffect, useRef } from 'react';
 import { extractI18nObject } from '@/i18n/I18nProvider';
 import { useTranslation } from 'react-i18next';
 
@@ -73,6 +73,12 @@ export default function DynamicFormComponent({
           case 'bot-selector':
             fieldSchema = z.string();
             break;
+          case 'model-fallback-selector':
+            fieldSchema = z.object({
+              primary: z.string(),
+              fallbacks: z.array(z.string()),
+            });
+            break;
           case 'prompt-editor':
             fieldSchema = z.array(
               z.object({
@@ -160,39 +166,34 @@ export default function DynamicFormComponent({
   const onSubmitRef = useRef(onSubmit);
   onSubmitRef.current = onSubmit;
 
-  // Track the last emitted values to avoid emitting identical snapshots,
-  // which would cause the parent to call setValue with an equivalent object,
-  // triggering a re-render loop.
-  const lastEmittedRef = useRef<string>('');
-
-  const emitValues = useCallback(() => {
+  // 监听表单值变化
+  useEffect(() => {
+    // Emit initial form values immediately so the parent always has a valid snapshot,
+    // even if the user saves without modifying any field.
+    // form.watch(callback) only fires on subsequent changes, not on mount.
     const formValues = form.getValues();
-    const finalValues = itemConfigList.reduce(
+    const initialFinalValues = itemConfigList.reduce(
       (acc, item) => {
         acc[item.name] = formValues[item.name] ?? item.default;
         return acc;
       },
       {} as Record<string, object>,
     );
-    const serialized = JSON.stringify(finalValues);
-    if (serialized !== lastEmittedRef.current) {
-      lastEmittedRef.current = serialized;
-      onSubmitRef.current?.(finalValues);
-    }
-  }, [form, itemConfigList]);
-
-  // 监听表单值变化
-  useEffect(() => {
-    // Emit initial form values immediately so the parent always has a valid snapshot,
-    // even if the user saves without modifying any field.
-    // form.watch(callback) only fires on subsequent changes, not on mount.
-    emitValues();
+    onSubmitRef.current?.(initialFinalValues);
 
     const subscription = form.watch(() => {
-      emitValues();
+      const formValues = form.getValues();
+      const finalValues = itemConfigList.reduce(
+        (acc, item) => {
+          acc[item.name] = formValues[item.name] ?? item.default;
+          return acc;
+        },
+        {} as Record<string, object>,
+      );
+      onSubmitRef.current?.(finalValues);
     });
     return () => subscription.unsubscribe();
-  }, [form, itemConfigList, emitValues]);
+  }, [form, itemConfigList]);
 
   return (
     <Form {...form}>
@@ -231,6 +232,7 @@ export default function DynamicFormComponent({
 
           // All fields are disabled when editing (creation_settings are immutable)
           const isFieldDisabled = !!isEditing;
+
           return (
             <FormField
               key={config.id}

web/src/app/home/components/dynamic-form/DynamicFormItemComponent.tsx

@@ -124,6 +124,28 @@ export default function DynamicFormItemComponent({
     }
   }, [config.type]);
 
+  useEffect(() => {
+    if (config.type === DynamicFormItemType.MODEL_FALLBACK_SELECTOR) {
+      httpClient
+        .getProviderLLMModels()
+        .then((resp) => {
+          let models = resp.models;
+          if (
+            systemInfo.disable_models_service ||
+            userInfo?.account_type !== 'space'
+          ) {
+            models = models.filter(
+              (m) => m.provider?.requester !== 'space-chat-completions',
+            );
+          }
+          setLlmModels(models);
+        })
+        .catch((err) => {
+          toast.error('Failed to get LLM model list: ' + err.msg);
+        });
+    }
+  }, [config.type]);
+
   useEffect(() => {
     if (
       config.type === DynamicFormItemType.KNOWLEDGE_BASE_SELECTOR ||
@@ -171,12 +193,7 @@ export default function DynamicFormItemComponent({
       return <Textarea {...field} className="min-h-[120px]" />;
 
     case DynamicFormItemType.BOOLEAN:
-      return (
-        <Switch
-          checked={field.value ?? false}
-          onCheckedChange={field.onChange}
-        />
-      );
+      return <Switch checked={field.value} onCheckedChange={field.onChange} />;
 
     case DynamicFormItemType.STRING_ARRAY:
       return (
@@ -227,7 +244,7 @@ export default function DynamicFormItemComponent({
 
     case DynamicFormItemType.SELECT:
       return (
-        <Select value={field.value ?? ''} onValueChange={field.onChange}>
+        <Select value={field.value} onValueChange={field.onChange}>
           <SelectTrigger className="bg-[#ffffff] dark:bg-[#2a2a2e]">
             <SelectValue placeholder={t('common.select')} />
           </SelectTrigger>
@@ -318,6 +335,172 @@ export default function DynamicFormItemComponent({
         </Select>
       );
 
+    case DynamicFormItemType.MODEL_FALLBACK_SELECTOR: {
+      // Group models by provider
+      const groupedModelsForFallback = llmModels.reduce(
+        (acc, model) => {
+          const providerName =
+            model.provider?.name || model.provider?.requester || 'Unknown';
+          if (!acc[providerName]) acc[providerName] = [];
+          acc[providerName].push(model);
+          return acc;
+        },
+        {} as Record<string, LLMModel[]>,
+      );
+
+      const modelValue = field.value as {
+        primary: string;
+        fallbacks: string[];
+      };
+
+      const renderModelSelect = (
+        value: string,
+        onChange: (val: string) => void,
+        placeholder: string,
+      ) => (
+        <Select value={value} onValueChange={onChange}>
+          <SelectTrigger className="bg-[#ffffff] dark:bg-[#2a2a2e]">
+            <SelectValue placeholder={placeholder} />
+          </SelectTrigger>
+          <SelectContent>
+            {Object.entries(groupedModelsForFallback).map(
+              ([providerName, models]) => (
+                <SelectGroup key={providerName}>
+                  <SelectLabel>{providerName}</SelectLabel>
+                  {models.map((model) => (
+                    <SelectItem key={model.uuid} value={model.uuid}>
+                      <span className="inline-flex items-center gap-1">
+                        {model.name}
+                        {model.abilities?.includes('vision') && (
+                          <Eye className="h-3 w-3 text-muted-foreground" />
+                        )}
+                        {model.abilities?.includes('func_call') && (
+                          <Wrench className="h-3 w-3 text-muted-foreground" />
+                        )}
+                      </span>
+                    </SelectItem>
+                  ))}
+                </SelectGroup>
+              ),
+            )}
+          </SelectContent>
+        </Select>
+      );
+
+      const updateValue = (patch: Partial<typeof modelValue>) => {
+        field.onChange({ ...modelValue, ...patch });
+      };
+
+      const addFallbackModel = () => {
+        updateValue({ fallbacks: [...modelValue.fallbacks, ''] });
+      };
+
+      const updateFallbackModel = (index: number, value: string) => {
+        const updated = [...modelValue.fallbacks];
+        updated[index] = value;
+        updateValue({ fallbacks: updated });
+      };
+
+      const removeFallbackModel = (index: number) => {
+        const updated = [...modelValue.fallbacks];
+        updated.splice(index, 1);
+        updateValue({ fallbacks: updated });
+      };
+
+      const moveFallbackModel = (index: number, direction: 'up' | 'down') => {
+        const updated = [...modelValue.fallbacks];
+        const newIndex = direction === 'up' ? index - 1 : index + 1;
+        if (newIndex < 0 || newIndex >= updated.length) return;
+        [updated[index], updated[newIndex]] = [
+          updated[newIndex],
+          updated[index],
+        ];
+        updateValue({ fallbacks: updated });
+      };
+
+      return (
+        <div className="space-y-3">
+          {/* Primary model selector */}
+          <div>
+            <p className="text-xs text-muted-foreground mb-1">
+              {t('models.fallback.primary')}
+            </p>
+            {renderModelSelect(
+              modelValue.primary,
+              (val) => updateValue({ primary: val }),
+              t('models.selectModel'),
+            )}
+          </div>
+
+          {/* Fallback models */}
+          {modelValue.fallbacks.length > 0 && (
+            <div className="space-y-2">
+              <p className="text-xs text-muted-foreground">
+                {t('models.fallback.fallbackList')}
+              </p>
+              {modelValue.fallbacks.map((fbUuid: string, index: number) => (
+                <div key={index} className="flex items-center gap-2">
+                  <span className="text-xs text-muted-foreground w-4 shrink-0">
+                    {index + 1}.
+                  </span>
+                  <div className="flex-1">
+                    {renderModelSelect(
+                      fbUuid,
+                      (val) => updateFallbackModel(index, val),
+                      t('models.selectModel'),
+                    )}
+                  </div>
+                  <div className="flex gap-1 shrink-0">
+                    <Button
+                      type="button"
+                      variant="ghost"
+                      size="sm"
+                      className="h-8 w-8 p-0"
+                      onClick={() => moveFallbackModel(index, 'up')}
+                      disabled={index === 0}
+                    >
+                      ↑
+                    </Button>
+                    <Button
+                      type="button"
+                      variant="ghost"
+                      size="sm"
+                      className="h-8 w-8 p-0"
+                      onClick={() => moveFallbackModel(index, 'down')}
+                      disabled={index === modelValue.fallbacks.length - 1}
+                    >
+                      ↓
+                    </Button>
+                    <Button
+                      type="button"
+                      variant="ghost"
+                      size="sm"
+                      className="h-8 w-8 p-0 text-destructive"
+                      onClick={() => removeFallbackModel(index)}
+                    >
+                      <X className="h-4 w-4" />
+                    </Button>
+                  </div>
+                </div>
+              ))}
+            </div>
+          )}
+
+          {/* Add fallback button */}
+          <Button
+            type="button"
+            variant="outline"
+            size="sm"
+            className="w-full"
+            onClick={addFallbackModel}
+          >
+            <Plus className="h-4 w-4 mr-1" />
+            {t('models.fallback.addFallback')}
+          </Button>
+        </div>
+      );
+    }
+
     case DynamicFormItemType.KNOWLEDGE_BASE_SELECTOR:
       // Group KBs by Knowledge Engine name
       const kbsByEngine = knowledgeBases.reduce(

web/src/app/home/components/models-dialog/ModelsDialog.tsx

@@ -463,14 +463,16 @@ export default function ModelsDialog({
                     )
                   : t('models.providerCount', { count: otherProviders.length })}
               </span>
-              <Button
-                size="sm"
-                variant="outline"
-                onClick={handleCreateProvider}
-              >
-                <Plus className="h-4 w-4 mr-1" />
-                {t('models.addProvider')}
-              </Button>
+              <div className="flex gap-2">
+                <Button
+                  size="sm"
+                  variant="outline"
+                  onClick={handleCreateProvider}
+                >
+                  <Plus className="h-4 w-4 mr-1" />
+                  {t('models.addProvider')}
+                </Button>
+              </div>
             </div>
 
             {/* Provider List */}

web/src/app/home/plugins/components/plugin-market/PluginMarketComponent.tsx

@@ -1,13 +1,6 @@
 'use client';
 
-import {
-  useState,
-  useEffect,
-  useCallback,
-  useRef,
-  Suspense,
-  useMemo,
-} from 'react';
+import { useState, useEffect, useCallback, useRef, Suspense } from 'react';
 import { Input } from '@/components/ui/input';
 import {
   Select,
@@ -70,7 +63,7 @@ function MarketPageContent({
     RecommendationList[]
   >([]);
 
-  const pageSize = 16; // 每页16个，4行x4列
+  const pageSize = 12; // 每页12个
   const searchTimeoutRef = useRef<NodeJS.Timeout | null>(null);
   const scrollContainerRef = useRef<HTMLDivElement | null>(null);
 
@@ -330,38 +323,7 @@ function MarketPageContent({
     };
   }, []);
 
-  // 计算所有推荐插件的 ID 集合
-  const recommendedPluginIds = useMemo(() => {
-    const ids = new Set<string>();
-    recommendationLists.forEach((list) => {
-      list.plugins.forEach((plugin) => {
-        ids.add(`${plugin.author} / ${plugin.name}`);
-      });
-    });
-    return ids;
-  }, [recommendationLists]);
-
-  // 过滤掉已在推荐列表中展示的插件
-  // 仅在显示推荐列表的条件下（无搜索、无筛选、第一页或后续页的累积数据中）进行过滤
-  // 注意：如果用户翻页，我们希望一直保持去重，否则推荐过的插件会在第二页出现
-  // 但是推荐列表只在第一页且无筛选时显示。
-  // 如果用户进行了筛选/搜索，推荐列表不显示，此时不需要去重。
-  const visiblePlugins = useMemo(() => {
-    const showRecommendations =
-      !searchQuery && componentFilter === 'all' && selectedTags.length === 0;
-
-    if (!showRecommendations) {
-      return plugins;
-    }
-
-    return plugins.filter((p) => !recommendedPluginIds.has(p.pluginId));
-  }, [
-    plugins,
-    recommendedPluginIds,
-    searchQuery,
-    componentFilter,
-    selectedTags,
-  ]);
+  const visiblePlugins = plugins;
 
   // 加载更多
   const loadMore = useCallback(() => {

web/src/app/home/plugins/components/plugin-market/RecommendationLists.tsx

@@ -47,10 +47,12 @@ function RecommendationListRow({
   list,
   tagNames,
   onInstall,
+  isLast,
 }: {
   list: RecommendationList;
   tagNames: Record<string, string>;
   onInstall: (author: string, pluginName: string) => void;
+  isLast: boolean;
 }) {
   const { t } = useTranslation();
   const [page, setPage] = useState(0);
@@ -143,7 +145,9 @@ function RecommendationListRow({
           />
         ))}
       </div>
-      {totalPages > 1 && <div className="border-b border-border mt-6" />}
+      {totalPages > 1 && !isLast && (
+        <div className="border-b border-border mt-6" />
+      )}
     </div>
   );
 }
@@ -161,12 +165,13 @@ export function RecommendationLists({
 
   return (
     <div className="mt-6">
-      {lists.map((list) => (
+      {lists.map((list, index) => (
         <RecommendationListRow
           key={list.uuid}
           list={list}
           tagNames={tagNames}
           onInstall={onInstall}
+          isLast={index === lists.length - 1}
         />
       ))}
       <div className="border-b border-border mb-6" />

web/src/app/home/plugins/components/plugin-market/plugin-market-card/PluginMarketCardComponent.tsx

@@ -17,7 +17,7 @@ import {
   FileText,
   Info,
 } from 'lucide-react';
-import { useState } from 'react';
+import { useState, useRef, useEffect } from 'react';
 import { Button } from '@/components/ui/button';
 
 export default function PluginMarketCardComponent({
@@ -31,6 +31,43 @@ export default function PluginMarketCardComponent({
 }) {
   const { t } = useTranslation();
   const [isHovered, setIsHovered] = useState(false);
+  const bottomRef = useRef<HTMLDivElement>(null);
+  const [visibleTags, setVisibleTags] = useState(2);
+
+  // Measure how many tags fit in the bottom row
+  useEffect(() => {
+    const tags = cardVO.tags;
+    if (!bottomRef.current || !tags || tags.length === 0) return;
+
+    const measure = () => {
+      const container = bottomRef.current;
+      if (!container) return;
+      const width = container.offsetWidth;
+      const availableForTags = width - 140 - 80;
+      if (availableForTags <= 0) {
+        setVisibleTags(0);
+        return;
+      }
+      const tagWidth = 80;
+      const plusBadgeWidth = 40;
+      const maxTags = Math.max(
+        0,
+        Math.floor((availableForTags - plusBadgeWidth) / tagWidth),
+      );
+      if (maxTags >= tags.length) {
+        setVisibleTags(tags.length);
+      } else {
+        setVisibleTags(Math.max(1, maxTags));
+      }
+    };
+
+    measure();
+    const observer = new ResizeObserver(measure);
+    observer.observe(bottomRef.current);
+    return () => observer.disconnect();
+  }, [cardVO.tags]);
+
+  const remainingTags = cardVO.tags ? cardVO.tags.length - visibleTags : 0;
 
   function handleInstallClick(e: React.MouseEvent) {
     e.stopPropagation();
@@ -135,10 +172,13 @@ export default function PluginMarketCardComponent({
         </div>
 
         {/* 下部分：下载量、标签和组件列表 */}
-        <div className="w-full flex flex-row items-center justify-between gap-2 px-0 sm:px-[0.4rem] flex-shrink-0">
-          <div className="flex flex-row items-center justify-start gap-2 flex-wrap">
+        <div
+          ref={bottomRef}
+          className="w-full flex flex-row items-center justify-between gap-2 px-0 sm:px-[0.4rem] flex-shrink-0 overflow-hidden"
+        >
+          <div className="flex flex-row items-center justify-start gap-2 min-w-0 overflow-hidden">
             {/* 下载数量 */}
-            <div className="flex flex-row items-center gap-[0.3rem] sm:gap-[0.4rem]">
+            <div className="flex flex-row items-center gap-[0.3rem] sm:gap-[0.4rem] flex-shrink-0">
               <svg
                 className="w-4 h-4 sm:w-[1.2rem] sm:h-[1.2rem] text-[#2563eb] dark:text-[#5b8def] flex-shrink-0"
                 xmlns="http://www.w3.org/2000/svg"
@@ -156,14 +196,14 @@ export default function PluginMarketCardComponent({
               </div>
             </div>
 
-            {/* Tags */}
-            {cardVO.tags && cardVO.tags.length > 0 && (
-              <div className="flex flex-wrap gap-1.5">
-                {cardVO.tags.slice(0, 2).map((tag) => (
+            {/* Tags - adaptive */}
+            {cardVO.tags && cardVO.tags.length > 0 && visibleTags > 0 && (
+              <div className="flex flex-row items-center gap-1.5 overflow-hidden flex-shrink min-w-0">
+                {cardVO.tags.slice(0, visibleTags).map((tag) => (
                   <Badge
                     key={tag}
                     variant="secondary"
-                    className="text-[0.65rem] sm:text-[0.7rem] px-2 py-0.5 h-5 flex items-center gap-1 flex-shrink-0"
+                    className="text-[0.65rem] sm:text-[0.7rem] px-2 py-0.5 h-5 flex items-center gap-1 flex-shrink-0 whitespace-nowrap"
                   >
                     <svg
                       className="w-2.5 h-2.5 flex-shrink-0"
@@ -178,15 +218,17 @@ export default function PluginMarketCardComponent({
                       <path d="M20.59 13.41l-7.17 7.17a2 2 0 0 1-2.83 0L2 12V2h10l8.59 8.59a2 2 0 0 1 0 2.82z" />
                       <line x1="7" y1="7" x2="7.01" y2="7" />
                     </svg>
-                    <span className="truncate">{tagNames[tag] || tag}</span>
+                    <span className="truncate max-w-[5rem]">
+                      {tagNames[tag] || tag}
+                    </span>
                   </Badge>
                 ))}
-                {cardVO.tags.length > 2 && (
+                {remainingTags > 0 && (
                   <Badge
                     variant="outline"
-                    className="text-[0.65rem] sm:text-[0.7rem] px-2 py-0.5 h-5 flex items-center flex-shrink-0"
+                    className="text-[0.65rem] sm:text-[0.7rem] px-1.5 py-0.5 h-5 flex items-center flex-shrink-0 whitespace-nowrap"
                   >
-                    +{cardVO.tags.length - 2}
+                    +{remainingTags}
                   </Badge>
                 )}
               </div>

web/src/app/infra/entities/form/dynamic.ts

@@ -35,6 +35,7 @@ export enum DynamicFormItemType {
   SELECT = 'select',
   LLM_MODEL_SELECTOR = 'llm-model-selector',
   EMBEDDING_MODEL_SELECTOR = 'embedding-model-selector',
+  MODEL_FALLBACK_SELECTOR = 'model-fallback-selector',
   PROMPT_EDITOR = 'prompt-editor',
   UNKNOWN = 'unknown',
   KNOWLEDGE_BASE_SELECTOR = 'knowledge-base-selector',

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

@@ -356,6 +356,7 @@ export class BackendClient extends BaseHttpClient {
       is_active: boolean;
       platform: string | null;
       user_id: string | null;
+      user_name: string | null;
     }>;
     total: number;
   }> {
@@ -384,6 +385,7 @@ export class BackendClient extends BaseHttpClient {
       level: string;
       platform: string | null;
       user_id: string | null;
+      user_name: string | null;
       runner_name: string | null;
       variables: string | null;
       role: string | null;

web/src/app/login/page.tsx

@@ -284,6 +284,27 @@ export default function Login() {
               </form>
             </Form>
           )}
+
+          <p className="text-xs text-center text-muted-foreground">
+            {t('common.agreementNotice')}{' '}
+            <a
+              href="https://langbot.app/privacy"
+              target="_blank"
+              rel="noopener noreferrer"
+              className="underline hover:text-foreground transition-colors"
+            >
+              {t('common.privacyPolicy')}
+            </a>{' '}
+            {t('common.and')}{' '}
+            <a
+              href={t('common.dataCollectionPolicyUrl')}
+              target="_blank"
+              rel="noopener noreferrer"
+              className="underline hover:text-foreground transition-colors"
+            >
+              {t('common.dataCollectionPolicy')}
+            </a>
+          </p>
         </CardContent>
       </Card>
     </div>

web/src/app/register/page.tsx

@@ -253,6 +253,27 @@ export default function Register() {
               </Button>
             </form>
           </Form>
+
+          <p className="text-xs text-center text-muted-foreground">
+            {t('common.agreementNotice')}{' '}
+            <a
+              href="https://langbot.app/privacy"
+              target="_blank"
+              rel="noopener noreferrer"
+              className="underline hover:text-foreground transition-colors"
+            >
+              {t('common.privacyPolicy')}
+            </a>{' '}
+            {t('common.and')}{' '}
+            <a
+              href={t('common.dataCollectionPolicyUrl')}
+              target="_blank"
+              rel="noopener noreferrer"
+              className="underline hover:text-foreground transition-colors"
+            >
+              {t('common.dataCollectionPolicy')}
+            </a>
+          </p>
         </CardContent>
       </Card>
     </div>

web/src/i18n/locales/en-US.ts

@@ -47,6 +47,12 @@ const enUS = {
     copyFailed: 'Copy Failed',
     test: 'Test',
     forgotPassword: 'Forgot Password?',
+    agreementNotice: 'By continuing, you agree to our',
+    privacyPolicy: 'Privacy Policy',
+    and: 'and',
+    dataCollectionPolicy: 'Data Collection Policy',
+    dataCollectionPolicyUrl:
+      'https://docs.langbot.app/en/insight/data-collection-policy',
     loading: 'Loading...',
     fieldRequired: 'This field is required',
     or: 'or',
@@ -230,6 +236,11 @@ const enUS = {
     modelsCount: '{{count}} model(s)',
     expandModels: 'Expand',
     collapseModels: 'Collapse',
+    fallback: {
+      primary: 'Primary Model',
+      fallbackList: 'Fallback Models',
+      addFallback: 'Add Fallback Model',
+    },
   },
   bots: {
     title: 'Bots',

web/src/i18n/locales/ja-JP.ts

@@ -1,4 +1,4 @@
-const jaJP = {
+const jaJP = {
   common: {
     login: 'ログイン',
     logout: 'ログアウト',
@@ -48,6 +48,12 @@ const jaJP = {
     copyFailed: 'コピーに失敗しました',
     test: 'テスト',
     forgotPassword: 'パスワードを忘れた？',
+    agreementNotice: '続行することで、以下に同意したものとみなされます：',
+    privacyPolicy: 'プライバシーポリシー',
+    and: 'および',
+    dataCollectionPolicy: 'データ収集ポリシー',
+    dataCollectionPolicyUrl:
+      'https://docs.langbot.app/ja/insight/data-collection-policy',
     loading: '読み込み中...',
     fieldRequired: 'この項目は必須です',
     or: 'または',
@@ -235,6 +241,11 @@ const jaJP = {
     modelsCount: '{{count}} 個のモデル',
     expandModels: '展開',
     collapseModels: '折りたたむ',
+    fallback: {
+      primary: 'プライマリモデル',
+      fallbackList: 'フォールバックモデル',
+      addFallback: 'フォールバックモデルを追加',
+    },
   },
   bots: {
     title: 'ボット',

web/src/i18n/locales/zh-Hans.ts

@@ -47,6 +47,12 @@ const zhHans = {
     copyFailed: '复制失败',
     test: '测试',
     forgotPassword: '忘记密码？',
+    agreementNotice: '继续即表示您同意我们的',
+    privacyPolicy: '隐私政策',
+    and: '和',
+    dataCollectionPolicy: '数据收集政策',
+    dataCollectionPolicyUrl:
+      'https://docs.langbot.app/zh/insight/data-collection-policy',
     loading: '加载中...',
     fieldRequired: '此字段为必填项',
     or: '或',
@@ -221,6 +227,11 @@ const zhHans = {
     modelsCount: '{{count}} 个模型',
     expandModels: '展开',
     collapseModels: '收起',
+    fallback: {
+      primary: '主模型',
+      fallbackList: '备用模型',
+      addFallback: '添加备用模型',
+    },
   },
   bots: {
     title: '机器人',

web/src/i18n/locales/zh-Hant.ts

@@ -47,6 +47,12 @@ const zhHant = {
     copyFailed: '複製失敗',
     test: '測試',
     forgotPassword: '忘記密碼？',
+    agreementNotice: '繼續即表示您同意我們的',
+    privacyPolicy: '隱私政策',
+    and: '和',
+    dataCollectionPolicy: '數據收集政策',
+    dataCollectionPolicyUrl:
+      'https://docs.langbot.app/zh/insight/data-collection-policy',
     loading: '載入中...',
     fieldRequired: '此欄位為必填',
     or: '或',
@@ -220,6 +226,11 @@ const zhHant = {
     modelsCount: '{{count}} 個模型',
     expandModels: '展開',
     collapseModels: '收起',
+    fallback: {
+      primary: '主模型',
+      fallbackList: '備用模型',
+      addFallback: '新增備用模型',
+    },
   },
   bots: {
     title: '機器人',