AstrBot 4.5.7__py3-none-any.whl → 4.6.0__py3-none-any.whl

astrbot/core/agent/mcp_client.py

@@ -4,6 +4,14 @@ from contextlib import AsyncExitStack
 from datetime import timedelta
 from typing import Generic
 
+from tenacity import (
+    before_sleep_log,
+    retry,
+    retry_if_exception_type,
+    stop_after_attempt,
+    wait_exponential,
+)
+
 from astrbot import logger
 from astrbot.core.agent.run_context import ContextWrapper
 from astrbot.core.utils.log_pipe import LogPipe
@@ -12,21 +20,24 @@ from .run_context import TContext
 from .tool import FunctionTool
 
 try:
+    import anyio
     import mcp
     from mcp.client.sse import sse_client
 except (ModuleNotFoundError, ImportError):
-    logger.warning("警告: 缺少依赖库 'mcp'，将无法使用 MCP 服务。")
+    logger.warning(
+        "Warning: Missing 'mcp' dependency, MCP services will be unavailable."
+    )
 
 try:
     from mcp.client.streamable_http import streamablehttp_client
 except (ModuleNotFoundError, ImportError):
     logger.warning(
-        "警告: 缺少依赖库 'mcp' 或者 mcp 库版本过低，无法使用 Streamable HTTP 连接方式。",
+        "Warning: Missing 'mcp' dependency or MCP library version too old, Streamable HTTP connection unavailable.",
     )
 
 
 def _prepare_config(config: dict) -> dict:
-    """准备配置，处理嵌套格式"""
+    """Prepare configuration, handle nested format"""
     if config.get("mcpServers"):
         first_key = next(iter(config["mcpServers"]))
         config = config["mcpServers"][first_key]
@@ -35,7 +46,7 @@ def _prepare_config(config: dict) -> dict:
 
 
 async def _quick_test_mcp_connection(config: dict) -> tuple[bool, str]:
-    """快速测试 MCP 服务器可达性"""
+    """Quick test MCP server connectivity"""
     import aiohttp
 
     cfg = _prepare_config(config.copy())
@@ -50,7 +61,7 @@ async def _quick_test_mcp_connection(config: dict) -> tuple[bool, str]:
         elif "type" in cfg:
             transport_type = cfg["type"]
         else:
-            raise Exception("MCP 连接配置缺少 transport 或 type 字段")
+            raise Exception("MCP connection config missing transport or type field")
 
         async with aiohttp.ClientSession() as session:
             if transport_type == "streamable_http":
@@ -91,7 +102,7 @@ async def _quick_test_mcp_connection(config: dict) -> tuple[bool, str]:
                     return False, f"HTTP {response.status}: {response.reason}"
 
     except asyncio.TimeoutError:
-        return False, f"连接超时: {timeout}秒"
+        return False, f"Connection timeout: {timeout} seconds"
     except Exception as e:
         return False, f"{e!s}"
 
@@ -101,6 +112,7 @@ class MCPClient:
         # Initialize session and client objects
         self.session: mcp.ClientSession | None = None
         self.exit_stack = AsyncExitStack()
+        self._old_exit_stacks: list[AsyncExitStack] = []  # Track old stacks for cleanup
 
         self.name: str | None = None
         self.active: bool = True
@@ -108,22 +120,32 @@ class MCPClient:
         self.server_errlogs: list[str] = []
         self.running_event = asyncio.Event()
 
+        # Store connection config for reconnection
+        self._mcp_server_config: dict | None = None
+        self._server_name: str | None = None
+        self._reconnect_lock = asyncio.Lock()  # Lock for thread-safe reconnection
+        self._reconnecting: bool = False  # For logging and debugging
+
     async def connect_to_server(self, mcp_server_config: dict, name: str):
-        """连接到 MCP 服务器
+        """Connect to MCP server
 
-        如果 `url` 参数存在：
-            1. 当 transport 指定为 `streamable_http` 时，使用 Streamable HTTP 连接方式。
-            1. 当 transport 指定为 `sse` 时，使用 SSE 连接方式。
-            2. 如果没有指定，默认使用 SSE 的方式连接到 MCP 服务。
+        If `url` parameter exists:
+            1. When transport is specified as `streamable_http`, use Streamable HTTP connection.
+            2. When transport is specified as `sse`, use SSE connection.
+            3. If not specified, default to SSE connection to MCP service.
 
         Args:
             mcp_server_config (dict): Configuration for the MCP server. See https://modelcontextprotocol.io/quickstart/server
 
         """
+        # Store config for reconnection
+        self._mcp_server_config = mcp_server_config
+        self._server_name = name
+
         cfg = _prepare_config(mcp_server_config.copy())
 
         def logging_callback(msg: str):
-            # 处理 MCP 服务的错误日志
+            # Handle MCP service error logs
             print(f"MCP Server {name} Error: {msg}")
             self.server_errlogs.append(msg)
 
@@ -137,7 +159,7 @@ class MCPClient:
             elif "type" in cfg:
                 transport_type = cfg["type"]
             else:
-                raise Exception("MCP 连接配置缺少 transport 或 type 字段")
+                raise Exception("MCP connection config missing transport or type field")
 
             if transport_type != "streamable_http":
                 # SSE transport method
@@ -193,7 +215,7 @@ class MCPClient:
             )
 
             def callback(msg: str):
-                # 处理 MCP 服务的错误日志
+                # Handle MCP service error logs
                 self.server_errlogs.append(msg)
 
             stdio_transport = await self.exit_stack.enter_async_context(
@@ -222,10 +244,120 @@ class MCPClient:
         self.tools = response.tools
         return response
 
+    async def _reconnect(self) -> None:
+        """Reconnect to the MCP server using the stored configuration.
+
+        Uses asyncio.Lock to ensure thread-safe reconnection in concurrent environments.
+
+        Raises:
+            Exception: raised when reconnection fails
+        """
+        async with self._reconnect_lock:
+            # Check if already reconnecting (useful for logging)
+            if self._reconnecting:
+                logger.debug(
+                    f"MCP Client {self._server_name} is already reconnecting, skipping"
+                )
+                return
+
+            if not self._mcp_server_config or not self._server_name:
+                raise Exception("Cannot reconnect: missing connection configuration")
+
+            self._reconnecting = True
+            try:
+                logger.info(
+                    f"Attempting to reconnect to MCP server {self._server_name}..."
+                )
+
+                # Save old exit_stack for later cleanup (don't close it now to avoid cancel scope issues)
+                if self.exit_stack:
+                    self._old_exit_stacks.append(self.exit_stack)
+
+                # Mark old session as invalid
+                self.session = None
+
+                # Create new exit stack for new connection
+                self.exit_stack = AsyncExitStack()
+
+                # Reconnect using stored config
+                await self.connect_to_server(self._mcp_server_config, self._server_name)
+                await self.list_tools_and_save()
+
+                logger.info(
+                    f"Successfully reconnected to MCP server {self._server_name}"
+                )
+            except Exception as e:
+                logger.error(
+                    f"Failed to reconnect to MCP server {self._server_name}: {e}"
+                )
+                raise
+            finally:
+                self._reconnecting = False
+
+    async def call_tool_with_reconnect(
+        self,
+        tool_name: str,
+        arguments: dict,
+        read_timeout_seconds: timedelta,
+    ) -> mcp.types.CallToolResult:
+        """Call MCP tool with automatic reconnection on failure, max 2 retries.
+
+        Args:
+            tool_name: tool name
+            arguments: tool arguments
+            read_timeout_seconds: read timeout
+
+        Returns:
+            MCP tool call result
+
+        Raises:
+            ValueError: MCP session is not available
+            anyio.ClosedResourceError: raised after reconnection failure
+        """
+
+        @retry(
+            retry=retry_if_exception_type(anyio.ClosedResourceError),
+            stop=stop_after_attempt(2),
+            wait=wait_exponential(multiplier=1, min=1, max=3),
+            before_sleep=before_sleep_log(logger, logging.WARNING),
+            reraise=True,
+        )
+        async def _call_with_retry():
+            if not self.session:
+                raise ValueError("MCP session is not available for MCP function tools.")
+
+            try:
+                return await self.session.call_tool(
+                    name=tool_name,
+                    arguments=arguments,
+                    read_timeout_seconds=read_timeout_seconds,
+                )
+            except anyio.ClosedResourceError:
+                logger.warning(
+                    f"MCP tool {tool_name} call failed (ClosedResourceError), attempting to reconnect..."
+                )
+                # Attempt to reconnect
+                await self._reconnect()
+                # Reraise the exception to trigger tenacity retry
+                raise
+
+        return await _call_with_retry()
+
     async def cleanup(self):
-        """Clean up resources"""
-        await self.exit_stack.aclose()
-        self.running_event.set()  # Set the running event to indicate cleanup is done
+        """Clean up resources including old exit stacks from reconnections"""
+        # Set running_event first to unblock any waiting tasks
+        self.running_event.set()
+
+        # Close current exit stack
+        try:
+            await self.exit_stack.aclose()
+        except Exception as e:
+            logger.debug(f"Error closing current exit stack: {e}")
+
+        # Don't close old exit stacks as they may be in different task contexts
+        # They will be garbage collected naturally
+        # Just clear the list to release references
+        self._old_exit_stacks.clear()
 
 
 class MCPTool(FunctionTool, Generic[TContext]):
@@ -246,14 +378,8 @@ class MCPTool(FunctionTool, Generic[TContext]):
     async def call(
         self, context: ContextWrapper[TContext], **kwargs
     ) -> mcp.types.CallToolResult:
-        session = self.mcp_client.session
-        if not session:
-            raise ValueError("MCP session is not available for MCP function tools.")
-        res = await session.call_tool(
-            name=self.mcp_tool.name,
+        return await self.mcp_client.call_tool_with_reconnect(
+            tool_name=self.mcp_tool.name,
             arguments=kwargs,
-            read_timeout_seconds=timedelta(
-                seconds=context.tool_call_timeout,
-            ),
+            read_timeout_seconds=timedelta(seconds=context.tool_call_timeout),
         )
-        return res

astrbot/core/agent/message.py

@@ -76,7 +76,7 @@ class ImageURLPart(ContentPart):
         """The ID of the image, to allow LLMs to distinguish different images."""
 
     type: str = "image_url"
-    image_url: str
+    image_url: ImageURL
 
 
 class AudioURLPart(ContentPart):
@@ -119,6 +119,13 @@ class ToolCall(BaseModel):
     """The ID of the tool call."""
     function: FunctionBody
     """The function body of the tool call."""
+    extra_content: dict[str, Any] | None = None
+    """Extra metadata for the tool call."""
+
+    def model_dump(self, **kwargs: Any) -> dict[str, Any]:
+        if self.extra_content is None:
+            kwargs.setdefault("exclude", set()).add("extra_content")
+        return super().model_dump(**kwargs)
 
 
 class ToolCallPart(BaseModel):

astrbot/core/config/default.py

@@ -4,7 +4,7 @@ import os
 
 from astrbot.core.utils.astrbot_path import get_astrbot_data_path
 
-VERSION = "4.5.7"
+VERSION = "4.6.0"
 DB_PATH = os.path.join(get_astrbot_data_path(), "data_v4.db")
 
 # 默认配置
@@ -137,6 +137,7 @@ DEFAULT_CONFIG = {
     "kb_names": [],  # 默认知识库名称列表
     "kb_fusion_top_k": 20,  # 知识库检索融合阶段返回结果数量
     "kb_final_top_k": 5,  # 知识库检索最终返回结果数量
+    "kb_agentic_mode": False,
 }
 
 
@@ -2146,6 +2147,7 @@ CONFIG_METADATA_2 = {
             "kb_names": {"type": "list", "items": {"type": "string"}},
             "kb_fusion_top_k": {"type": "int", "default": 20},
             "kb_final_top_k": {"type": "int", "default": 5},
+            "kb_agentic_mode": {"type": "bool"},
         },
     },
 }
@@ -2241,6 +2243,11 @@ CONFIG_METADATA_3 = {
                         "type": "int",
                         "hint": "从知识库中检索到的结果数量，越大可能获得越多相关信息，但也可能引入噪音。建议根据实际需求调整",
                     },
+                    "kb_agentic_mode": {
+                        "description": "Agentic 知识库检索",
+                        "type": "bool",
+                        "hint": "启用后，知识库检索将作为 LLM Tool，由模型自主决定何时调用知识库进行查询。需要模型支持函数调用能力。",
+                    },
                 },
             },
             "websearch": {

astrbot/core/core_lifecycle.py

@@ -22,6 +22,7 @@ from astrbot.core.config.default import VERSION
 from astrbot.core.conversation_mgr import ConversationManager
 from astrbot.core.db import BaseDatabase
 from astrbot.core.db.migration.migra_45_to_46 import migrate_45_to_46
+from astrbot.core.db.migration.migra_webchat_session import migrate_webchat_session
 from astrbot.core.knowledge_base.kb_mgr import KnowledgeBaseManager
 from astrbot.core.persona_mgr import PersonaManager
 from astrbot.core.pipeline.scheduler import PipelineContext, PipelineScheduler
@@ -103,6 +104,13 @@ class AstrBotCoreLifecycle:
             logger.error(f"Migration from version 4.5 to 4.6 failed: {e!s}")
             logger.error(traceback.format_exc())
 
+        # migration for webchat session
+        try:
+            await migrate_webchat_session(self.db)
+        except Exception as e:
+            logger.error(f"Migration for webchat session failed: {e!s}")
+            logger.error(traceback.format_exc())
+
         # 初始化事件队列
         self.event_queue = Queue()
 

astrbot/core/db/__init__.py

@@ -13,6 +13,7 @@ from astrbot.core.db.po import (
     ConversationV2,
     Persona,
     PlatformMessageHistory,
+    PlatformSession,
     PlatformStat,
     Preference,
     Stats,
@@ -183,7 +184,7 @@ class BaseDatabase(abc.ABC):
         user_id: str,
         offset_sec: int = 86400,
     ) -> None:
-        """Delete platform message history records older than the specified offset."""
+        """Delete platform message history records newer than the specified offset."""
         ...
 
     @abc.abstractmethod
@@ -313,3 +314,51 @@ class BaseDatabase(abc.ABC):
     ) -> tuple[list[dict], int]:
         """Get paginated session conversations with joined conversation and persona details, support search and platform filter."""
         ...
+
+    # ====
+    # Platform Session Management
+    # ====
+
+    @abc.abstractmethod
+    async def create_platform_session(
+        self,
+        creator: str,
+        platform_id: str = "webchat",
+        session_id: str | None = None,
+        display_name: str | None = None,
+        is_group: int = 0,
+    ) -> PlatformSession:
+        """Create a new Platform session."""
+        ...
+
+    @abc.abstractmethod
+    async def get_platform_session_by_id(
+        self, session_id: str
+    ) -> PlatformSession | None:
+        """Get a Platform session by its ID."""
+        ...
+
+    @abc.abstractmethod
+    async def get_platform_sessions_by_creator(
+        self,
+        creator: str,
+        platform_id: str | None = None,
+        page: int = 1,
+        page_size: int = 20,
+    ) -> list[PlatformSession]:
+        """Get all Platform sessions for a specific creator (username) and optionally platform."""
+        ...
+
+    @abc.abstractmethod
+    async def update_platform_session(
+        self,
+        session_id: str,
+        display_name: str | None = None,
+    ) -> None:
+        """Update a Platform session's updated_at timestamp and optionally display_name."""
+        ...
+
+    @abc.abstractmethod
+    async def delete_platform_session(self, session_id: str) -> None:
+        """Delete a Platform session by its ID."""
+        ...

astrbot/core/db/migration/migra_webchat_session.py

@@ -0,0 +1,131 @@
+"""Migration script for WebChat sessions.
+
+This migration creates PlatformSession from existing platform_message_history records.
+
+Changes:
+- Creates platform_sessions table
+- Adds platform_id field (default: 'webchat')
+- Adds display_name field
+- Session_id format: {platform_id}_{uuid}
+"""
+
+from sqlalchemy import func, select
+from sqlmodel import col
+
+from astrbot.api import logger, sp
+from astrbot.core.db import BaseDatabase
+from astrbot.core.db.po import ConversationV2, PlatformMessageHistory, PlatformSession
+
+
+async def migrate_webchat_session(db_helper: BaseDatabase):
+    """Create PlatformSession records from platform_message_history.
+
+    This migration extracts all unique user_ids from platform_message_history
+    where platform_id='webchat' and creates corresponding PlatformSession records.
+    """
+    # 检查是否已经完成迁移
+    migration_done = await db_helper.get_preference(
+        "global", "global", "migration_done_webchat_session"
+    )
+    if migration_done:
+        return
+
+    logger.info("开始执行数据库迁移（WebChat 会话迁移）...")
+
+    try:
+        async with db_helper.get_db() as session:
+            # 从 platform_message_history 创建 PlatformSession
+            query = (
+                select(
+                    col(PlatformMessageHistory.user_id),
+                    col(PlatformMessageHistory.sender_name),
+                    func.min(PlatformMessageHistory.created_at).label("earliest"),
+                    func.max(PlatformMessageHistory.updated_at).label("latest"),
+                )
+                .where(col(PlatformMessageHistory.platform_id) == "webchat")
+                .where(col(PlatformMessageHistory.sender_id) == "astrbot")
+                .group_by(col(PlatformMessageHistory.user_id))
+            )
+
+            result = await session.execute(query)
+            webchat_users = result.all()
+
+            if not webchat_users:
+                logger.info("没有找到需要迁移的 WebChat 数据")
+                await sp.put_async(
+                    "global", "global", "migration_done_webchat_session", True
+                )
+                return
+
+            logger.info(f"找到 {len(webchat_users)} 个 WebChat 会话需要迁移")
+
+            # 检查已存在的会话
+            existing_query = select(col(PlatformSession.session_id))
+            existing_result = await session.execute(existing_query)
+            existing_session_ids = {row[0] for row in existing_result.fetchall()}
+
+            # 查询 Conversations 表中的 title，用于设置 display_name
+            # 对于每个 user_id，对应的 conversation user_id 格式为: webchat:FriendMessage:webchat!astrbot!{user_id}
+            user_ids_to_query = [
+                f"webchat:FriendMessage:webchat!astrbot!{user_id}"
+                for user_id, _, _, _ in webchat_users
+            ]
+            conv_query = select(
+                col(ConversationV2.user_id), col(ConversationV2.title)
+            ).where(col(ConversationV2.user_id).in_(user_ids_to_query))
+            conv_result = await session.execute(conv_query)
+            # 创建 user_id -> title 的映射字典
+            title_map = {
+                user_id.replace("webchat:FriendMessage:webchat!astrbot!", ""): title
+                for user_id, title in conv_result.fetchall()
+            }
+
+            # 批量创建 PlatformSession 记录
+            sessions_to_add = []
+            skipped_count = 0
+
+            for user_id, sender_name, created_at, updated_at in webchat_users:
+                # user_id 就是 webchat_conv_id (session_id)
+                session_id = user_id
+
+                # sender_name 通常是 username，但可能为 None
+                creator = sender_name if sender_name else "guest"
+
+                # 检查是否已经存在该会话
+                if session_id in existing_session_ids:
+                    logger.debug(f"会话 {session_id} 已存在，跳过")
+                    skipped_count += 1
+                    continue
+
+                # 从 Conversations 表中获取 display_name
+                display_name = title_map.get(user_id)
+
+                # 创建新的 PlatformSession（保留原有的时间戳）
+                new_session = PlatformSession(
+                    session_id=session_id,
+                    platform_id="webchat",
+                    creator=creator,
+                    is_group=0,
+                    created_at=created_at,
+                    updated_at=updated_at,
+                    display_name=display_name,
+                )
+                sessions_to_add.append(new_session)
+
+            # 批量插入
+            if sessions_to_add:
+                session.add_all(sessions_to_add)
+                await session.commit()
+
+                logger.info(
+                    f"WebChat 会话迁移完成！成功迁移: {len(sessions_to_add)}, 跳过: {skipped_count}",
+                )
+            else:
+                logger.info("没有新会话需要迁移")
+
+        # 标记迁移完成
+        await sp.put_async("global", "global", "migration_done_webchat_session", True)
+
+    except Exception as e:
+        logger.error(f"迁移过程中发生错误: {e}", exc_info=True)
+        raise

astrbot/core/db/po.py

@@ -3,13 +3,7 @@ from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from typing import TypedDict
 
-from sqlmodel import (
-    JSON,
-    Field,
-    SQLModel,
-    Text,
-    UniqueConstraint,
-)
+from sqlmodel import JSON, Field, SQLModel, Text, UniqueConstraint
 
 
 class PlatformStat(SQLModel, table=True):
@@ -18,7 +12,7 @@ class PlatformStat(SQLModel, table=True):
     Note: In astrbot v4, we moved `platform` table to here.
     """
 
-    __tablename__ = "platform_stats"
+    __tablename__ = "platform_stats"  # type: ignore
 
     id: int = Field(primary_key=True, sa_column_kwargs={"autoincrement": True})
     timestamp: datetime = Field(nullable=False)
@@ -37,7 +31,7 @@ class PlatformStat(SQLModel, table=True):
 
 
 class ConversationV2(SQLModel, table=True):
-    __tablename__ = "conversations"
+    __tablename__ = "conversations"  # type: ignore
 
     inner_conversation_id: int = Field(
         primary_key=True,
@@ -74,7 +68,7 @@ class Persona(SQLModel, table=True):
     It can be used to customize the behavior of LLMs.
     """
 
-    __tablename__ = "personas"
+    __tablename__ = "personas"  # type: ignore
 
     id: int | None = Field(
         primary_key=True,
@@ -104,7 +98,7 @@ class Persona(SQLModel, table=True):
 class Preference(SQLModel, table=True):
     """This class represents preferences for bots."""
 
-    __tablename__ = "preferences"
+    __tablename__ = "preferences"  # type: ignore
 
     id: int | None = Field(
         default=None,
@@ -140,7 +134,7 @@ class PlatformMessageHistory(SQLModel, table=True):
     or platform-specific messages.
     """
 
-    __tablename__ = "platform_message_history"
+    __tablename__ = "platform_message_history"  # type: ignore
 
     id: int | None = Field(
         primary_key=True,
@@ -161,13 +155,55 @@ class PlatformMessageHistory(SQLModel, table=True):
     )
 
 
+class PlatformSession(SQLModel, table=True):
+    """Platform session table for managing user sessions across different platforms.
+
+    A session represents a chat window for a specific user on a specific platform.
+    Each session can have multiple conversations (对话) associated with it.
+    """
+
+    __tablename__ = "platform_sessions"  # type: ignore
+
+    inner_id: int | None = Field(
+        primary_key=True,
+        sa_column_kwargs={"autoincrement": True},
+        default=None,
+    )
+    session_id: str = Field(
+        max_length=100,
+        nullable=False,
+        unique=True,
+        default_factory=lambda: f"webchat_{uuid.uuid4()}",
+    )
+    platform_id: str = Field(default="webchat", nullable=False)
+    """Platform identifier (e.g., 'webchat', 'qq', 'discord')"""
+    creator: str = Field(nullable=False)
+    """Username of the session creator"""
+    display_name: str | None = Field(default=None, max_length=255)
+    """Display name for the session"""
+    is_group: int = Field(default=0, nullable=False)
+    """0 for private chat, 1 for group chat (not implemented yet)"""
+    created_at: datetime = Field(default_factory=lambda: datetime.now(timezone.utc))
+    updated_at: datetime = Field(
+        default_factory=lambda: datetime.now(timezone.utc),
+        sa_column_kwargs={"onupdate": datetime.now(timezone.utc)},
+    )
+
+    __table_args__ = (
+        UniqueConstraint(
+            "session_id",
+            name="uix_platform_session_id",
+        ),
+    )
+
+
 class Attachment(SQLModel, table=True):
     """This class represents attachments for messages in AstrBot.
 
     Attachments can be images, files, or other media types.
     """
 
-    __tablename__ = "attachments"
+    __tablename__ = "attachments"  # type: ignore
 
     inner_attachment_id: int | None = Field(
         primary_key=True,