♻️ refactor(core): 重构 AI 编排框架与记忆及 RAG 子系统 (#2149)

* ♻️ refactor(core): 重构 AI 编排框架与记忆及 RAG 子系统

- 【重构】重构 `BaseRunnable` 并引入统一的 `RunIntent` 意图载体，规范 Agent、Team 和 Workflow 的执行流
- 【解耦】将中期记忆槽和长期向量记忆从 `MemoryConfig` 中解耦，转为独立的能力组件与工具箱进行管理
- 【记忆】移除 `MemoryReader` 和 `MemoryWriter`，统一封装为 `SessionMemoryContext` 会话记忆门面
- 【RAG】重构检索器与存储后端接口，统一采用 `QueryRequest` 进行多维度联合检索，并引入 `InMemoryScorer` 提升打分性能
- 【事件】优化 `EventBus` 异步事件分发机制，引入队列机制确保事件按序处理，避免并发竞态问题
- 【依赖注入】移除 `memory` 注入项，优化 `DependencyInjector` 的签名解析缓存以提升性能

* 🚨 auto fix by pre-commit hooks

---------

Co-authored-by: webjoin111 <455457521@qq.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

zhenxun/services/ai/run/__init__.py

@@ -12,6 +12,7 @@ from .hooks import Hooks
 from .models import (
     AgentRunResult,
     AgentTask,
+    RunIntent,
     StreamedRunResult,
 )
 from .session import session_manager
@@ -28,6 +29,7 @@ __all__ = [
     "Inject",
     "NoneBotDeps",
     "RunContext",
+    "RunIntent",
     "StreamedRunResult",
     "UIController",
     "get_current_run_context",

zhenxun/services/ai/run/context.py

@@ -8,6 +8,9 @@ from typing import TYPE_CHECKING, Any, Generic, cast, get_origin
 from typing_extensions import TypeVar
 import uuid
 
+if TYPE_CHECKING:
+    from zhenxun.services.ai.context.memory.types import SessionMetadata
+
 from nonebot.adapters import Bot, Event
 from nonebot.matcher import Matcher, current_bot, current_event, current_matcher
 from pydantic import BaseModel, ConfigDict, Field
@@ -19,16 +22,12 @@ from zhenxun.services.ai.core.models import CancellationToken
 from zhenxun.services.ai.core.protocols.tool import ToolExecutable
 from zhenxun.services.ai.core.stream_events import AgentStreamEvent, EventBus
 from zhenxun.services.ai.utils import ContextUtils
-from zhenxun.services.ai.utils.scope import ScopeSelector
 from zhenxun.services.scheduler.types import ScheduleContext
 from zhenxun.utils.platform import PlatformUtils
 from zhenxun.utils.utils import infer_plugin_namespace
 
 from .blackboard import BlackboardManager
 
-if TYPE_CHECKING:
-    from zhenxun.services.ai.context.memory.facades import AgentSessionFacade
-
 AgentDepsT = TypeVar("AgentDepsT", default=Any)
 """泛型类型变量：外部环境依赖对象 (Agent Dependencies)。"""
 ProviderFunc = Callable[["RunContext"], Any | Awaitable[Any]]
@@ -114,31 +113,8 @@ class SessionContext(Generic[AgentDepsT]):
     """触发事件的插件命名空间"""
     append_only_manager: Any = dataclasses.field(default=None)
     """用于大模型前缀缓存命中优化的追加写入管理器。"""
-
-    @property
-    def memory(self) -> "AgentSessionFacade":
-        """
-        获取当前会话的持久化记忆访问门面 (AgentSessionFacade)。
-        提供极简的 history 和 slots 操作 API。
-        """
-        from zhenxun.services.ai.context.memory.facades import AgentSessionFacade
-        from zhenxun.services.ai.context.memory.manager import memory_manager
-        from zhenxun.services.ai.context.memory.types import SessionMetadata
-
-        user_id = ContextUtils.extract_user_id(self.deps)
-        group_id = ContextUtils.extract_group_id(self.deps)
-        platform = ContextUtils.extract_platform(self.deps)
-
-        meta = SessionMetadata(
-            session_id=self.session_id,
-            selector=ScopeSelector(
-                user_id=user_id,
-                group_id=group_id,
-                platform=platform,
-                namespace=self.namespace,
-            ),
-        )
-        return AgentSessionFacade(memory_manager, meta)
+    session_meta: "SessionMetadata | None" = dataclasses.field(default=None)
+    """隔离会话的元信息(Session ID, 命名空间, 权限等)，在状态流转中生成"""
 
 
 @dataclasses.dataclass
@@ -324,7 +300,7 @@ class RunContext(Generic[AgentDepsT]):
                 self.deps.get("namespace") if isinstance(self.deps, dict) else None
             )
         if not ns:
-            ns = infer_plugin_namespace(default="global")
+            ns = infer_plugin_namespace()
 
         self.session = SessionContext(
             session_id=self.session_id or "default_session",

zhenxun/services/ai/run/di.py

@@ -1,14 +1,27 @@
 from collections import defaultdict
 from collections.abc import Awaitable, Callable
+from functools import lru_cache
 import inspect
 from typing import Annotated, Any, ClassVar, cast
 
+
+@lru_cache(maxsize=2048)
+def _get_signature_cached(func: Callable) -> inspect.Signature:
+    return inspect.signature(func)
+
+
+def get_signature(func: Callable) -> inspect.Signature:
+    try:
+        return _get_signature_cached(func)
+    except TypeError:
+        return inspect.signature(func)
+
+
 from nonebot.adapters import Bot, Event
 from nonebot.matcher import Matcher
 from nonebot.utils import is_coroutine_callable
 from nonebot_plugin_session import EventSession, extract_session
 
-from zhenxun.services.ai.context.memory.facades import AgentSessionFacade
 from zhenxun.utils.utils import infer_plugin_namespace
 
 from .blackboard import BlackboardManager
@@ -59,7 +72,7 @@ CurrentEventPayload = Annotated[Any, Hidden(), _InjectMarker("stream_event")]
 CurrentBlackboard = Annotated[
     BlackboardManager | None, Hidden(), _InjectMarker("blackboard")
 ]
-CurrentMemory = Annotated[AgentSessionFacade, Hidden(), _InjectMarker("memory")]
+
 CurrentSandbox = Annotated[Any, Hidden(), _InjectMarker("sandbox")]
 
 
@@ -160,9 +173,6 @@ class Inject:
     Blackboard = CurrentBlackboard
     """自动注入：当前工作流/团队挂载的强类型黑板 (BlackboardManager) 实例"""
 
-    Memory = CurrentMemory
-    """自动注入：当前会话的持久化记忆存取门面 (AgentSessionFacade) 实例"""
-
     Sandbox = CurrentSandbox
     """自动注入：当前沙箱环境管理器实例"""
 
@@ -314,7 +324,7 @@ class DependencyInjector:
         context: RunContext,
     ) -> Any:
         """统一执行带有依赖注入的函数 (支持同步/异步)"""
-        sig = inspect.signature(func)
+        sig = get_signature(func)
         resolved_kwargs = await cls.resolve_all(sig, call_kwargs, context)
         filtered_kwargs = {
             k: v for k, v in resolved_kwargs.items() if k in sig.parameters
@@ -354,7 +364,6 @@ Inject.register_provider(
 Inject.register_provider(
     "shared_state", lambda ctx: ctx.session.shared_state, scope="global"
 )
-Inject.register_provider("memory", lambda ctx: ctx.session.memory, scope="global")
 
 
 def _resolve_blackboard(ctx: RunContext):

zhenxun/services/ai/run/models.py

@@ -4,16 +4,15 @@ from __future__ import annotations
 运行时（Run）相关核心类型定义
 """
 
-from collections.abc import AsyncIterator, Callable
+from collections.abc import AsyncIterator
 import json
 from typing import Any, Generic, cast
-from typing_extensions import TypeVar
 
 from pydantic import BaseModel, ConfigDict, Field, PrivateAttr
 
 from zhenxun.services.ai.core.messages import AgentMessage, LLMMessage, UsageInfo
+from zhenxun.services.ai.core.messages.types import OutputDataT
 from zhenxun.services.ai.core.options import BaseOutputDefinition
-from zhenxun.services.ai.core.protocols.tool import ToolResolvable
 from zhenxun.services.ai.core.stream_events import (
     AgentStreamEvent,
     EventBus,
@@ -21,7 +20,6 @@ from zhenxun.services.ai.core.stream_events import (
     ToolStreamChunkEvent,
 )
 from zhenxun.services.ai.guardrails import BaseGuardrail, GuardrailSource
-from zhenxun.services.ai.tools.core.tool import BaseTool
 from zhenxun.utils.pydantic_compat import model_dump, model_validator
 
 
@@ -71,9 +69,6 @@ class HandoffPayload(BaseModel):
     """随移交传递的上下文数据"""
 
 
-OutputDataT = TypeVar("OutputDataT", default=str)
-
-
 class AgentRunResult(BaseModel, Generic[OutputDataT]):
     """Agent 单次无状态运行的结果"""
 
@@ -238,9 +233,7 @@ class AgentTask(BaseModel):
     """强制要求返回的强类型结构 (Pydantic Model) 或
     OutputDefinition，为空则返回普通文本"""
 
-    tools: list[str | Callable | dict[str, Any] | BaseTool | ToolResolvable] | None = (
-        None
-    )
+    tools: list[Any] | None = None
     """针对此特定任务动态追加或覆盖的工具列表"""
 
     guardrails: list[GuardrailSource] | None = None
@@ -259,9 +252,74 @@ class AgentTask(BaseModel):
         return self
 
 
+class RunIntent(BaseModel):
+    """标准化且归一化的运行时意图载体"""
+
+    text: str = ""
+    """提取出的纯文本指令（用于路由、日志和并发控制判断）"""
+    original_input: Any = None
+    """用户最原始的输入对象（如 UniMessage 等，用于多模态图像/音频数据提取）"""
+    payload_to_render: Any = None
+    """将要被压入 MessageBuilder 渲染为大模型 Prompt 的实际载体"""
+    task_obj: AgentTask | None = None
+    """如果是强类型任务契约，存储其原始对象引用"""
+    response_model: type[BaseModel] | BaseOutputDefinition | None = None
+    """提取出的强类型输出约束"""
+    extra_tools: list[Any] = Field(default_factory=list)
+    """提取出的附加工具集"""
+    guardrails: list[BaseGuardrail] = Field(default_factory=list)
+    """提取出的安全护栏集"""
+
+    model_config = ConfigDict(arbitrary_types_allowed=True)
+
+    @classmethod
+    def from_input(cls, prompt: Any) -> "RunIntent":
+        task_obj = None
+        text_content = ""
+        extra_tools = []
+        response_model = None
+        guardrails = []
+        payload_to_render = prompt
+
+        if isinstance(prompt, AgentTask):
+            task_obj = prompt
+            response_model = task_obj.response_model
+            if task_obj.tools:
+                extra_tools.extend(task_obj.tools)
+            if hasattr(task_obj, "_parsed_guardrails"):
+                guardrails.extend(task_obj._parsed_guardrails)
+
+            prompt_parts = [
+                f"### 📋 [任务指令]\n{task_obj.description}",
+                f"### 🎯 [预期产出要求]\n{task_obj.expected_output}",
+            ]
+            text_content = "\n\n".join(prompt_parts)
+            payload_to_render = text_content
+        elif prompt is not None:
+            text_content = getattr(prompt, "description", None) or (
+                getattr(prompt, "extract_plain_text", lambda: str(prompt))()
+                if prompt
+                else str(prompt)
+            )
+        else:
+            text_content = ""
+            payload_to_render = None
+
+        return cls(
+            text=text_content,
+            original_input=prompt,
+            payload_to_render=payload_to_render,
+            task_obj=task_obj,
+            response_model=response_model,
+            extra_tools=extra_tools,
+            guardrails=guardrails,
+        )
+
+
 __all__ = [
     "AgentRunResult",
     "AgentTask",
     "OutputDataT",
+    "RunIntent",
     "StreamedRunResult",
 ]