mirror of
https://github.com/zhenxun-org/zhenxun_bot.git
synced 2026-10-12 23:40:03 +08:00
✨ feat!(llm): 重构并升级大语言模型服务为全新 AI 智能体框架 (#2146)
* ✨ feat!(llm): 重构并升级大语言模型服务为全新 AI 智能体框架 - 【重构】将原 services/llm 重构并迁移至全新的 services/ai 架构，提供向下兼容垫片 - 【新增】引入 Agent、Team、Workflow 三大智能体与工作流编排范式 - 【新增】引入基于 RAG 的长期向量记忆与中期槽位记忆系统 - 【新增】引入基于 Docker 的安全代码执行沙箱环境 - 【新增】支持 MCP 协议，允许动态管理和调用 MCP 服务 - 【新增】引入输入输出安全合规护栏与自愈反思机制 - 【优化】重构并优化多厂商 API 适配器 (Gemini, OpenAI, DeepSeek, GLM 等) - 【优化】优化日志脱敏与 Token 预估机制 - 【移除】移除旧版 llm default 和 llm reset-key 命令，新增 llm mcp 管理命令 * 🔧 chore(deps): 更新项目依赖与配置 - 添加 mcp、jieba 和 aiodocker 依赖到配置文件及 requirements.txt - 在 pyright 配置中设置 reportMissingImports 为 none - 调整 .gitignore 中 resources 目录的忽略规则 * ♻️ refactor(tools): 重构工具终止机制并清理知识库日志输出 - 统一使用 `context.state["__end_run__"]` 替代 `EndRunResult` 控制任务结束 - 移除文件系统和向量知识库检索工具中 `ToolResult` 的 `.with_log` 调用 - 调整指令处理器（Directive）的返回值为 `tool_res.output` - 修复部分类型检查警告并优化联合类型判断语法 * ♻️ refactor(tools): 重构工具副作用指令与控制流熔断机制 - 引入 `DirectivePayload` 及 `ToolResult` 的子类以结构化表达工具副作用 - 移除通过 `context.state` 传递魔术变量的隐式控制流设计 - 重构 `DirectiveManager` 处理器接口，直接在处理器中修改 `AgentState` 并构建 `AgentRunResult` - 在 `StandardAgentExecutor` 中统一通过 `directive_manager` 调度工具返回的副作用指令 - 补全 `MessageBuilder` 中部分核心方法的文档注释 * 🐛 fix(sandbox): 修复 Docker 沙箱容器状态检测与会话清理逻辑 -【修复】修正 `is_alive` 中直接读取私有属性的问题，改用 `show()` 返回值 -【修复】解决 `execute_code` 中缓存的执行器与当前会话不一致的问题 -【优化】在清理工作区前增加容器存活检测，避免向已死容器发送请求 -【优化】创建容器时增加运行状态校验，若已停止则自动从缓存中移除并重建 -【优化】优化容器销毁和清理逻辑，静默处理容器不存在 (404) 的异常 * 📝 docs(core): 补充核心模块初始化方法的文档注释 * 🚨 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>
This commit is contained in:
Rumio
co-authored by
webjoin111
pre-commit-ci[bot]
webjoin111
pre-commit-ci[bot]
parent
bdc1374848
commit
80fc5b86a7
@@ -0,0 +1,213 @@
|
||||
"""
|
||||
模型自身设定域类型定义
|
||||
"""
|
||||
|
||||
import asyncio
|
||||
from dataclasses import dataclass
|
||||
from enum import Enum
|
||||
from typing import Any, Generic, Literal, TypeVar
|
||||
|
||||
from pydantic import BaseModel, ConfigDict, Field
|
||||
|
||||
from zhenxun.services.ai.core.options import GenerationConfig
|
||||
|
||||
ModelName = str | None
|
||||
|
||||
TReq = TypeVar("TReq", bound=BaseModel)
|
||||
"""泛型：LLM 请求体模型约束 (必须继承自 BaseModel)"""
|
||||
TRes = TypeVar("TRes", bound=BaseModel)
|
||||
"""泛型：LLM 响应体模型约束 (必须继承自 BaseModel)"""
|
||||
|
||||
|
||||
@dataclass
|
||||
class ModelIdentity:
|
||||
"""模型的身份标识与基础能力数据传输对象 (DTO)，剥离运行时状态"""
|
||||
|
||||
provider_name: str
|
||||
"""模型提供商名称"""
|
||||
model_name: str
|
||||
"""模型名称"""
|
||||
api_type: str
|
||||
"""API 适配器类型"""
|
||||
api_base: str | None
|
||||
"""API 基础请求地址/网关终结点"""
|
||||
path_prefix: str | None
|
||||
"""中转路由的 URL 前缀"""
|
||||
capabilities: "ModelCapabilities"
|
||||
"""模型的能力配置定义描述"""
|
||||
generation_config: GenerationConfig | None
|
||||
"""模型的默认生成配置"""
|
||||
|
||||
|
||||
class CancellationToken:
|
||||
"""全局取消令牌，用于在异步链路中传递中止信号"""
|
||||
|
||||
def __init__(self):
|
||||
self._cancelled = False
|
||||
self._futures: list[asyncio.Future] = []
|
||||
|
||||
def cancel(self) -> None:
|
||||
self._cancelled = True
|
||||
for f in self._futures:
|
||||
if not f.done():
|
||||
f.cancel()
|
||||
|
||||
def is_cancelled(self) -> bool:
|
||||
return self._cancelled
|
||||
|
||||
def raise_if_cancelled(self) -> None:
|
||||
if self._cancelled:
|
||||
raise asyncio.CancelledError(
|
||||
"任务已被主动取消 (CancellationToken triggered)"
|
||||
)
|
||||
|
||||
def link_future(self, future: asyncio.Future) -> None:
|
||||
if self._cancelled:
|
||||
future.cancel()
|
||||
else:
|
||||
self._futures.append(future)
|
||||
|
||||
|
||||
class LLMContext(BaseModel, Generic[TReq, TRes]):
|
||||
"""LLM 执行上下文，用于在中间件管道中传递请求状态"""
|
||||
|
||||
request: TReq
|
||||
"""强类型的各模态请求对象实体。"""
|
||||
|
||||
runtime_state: dict[str, Any] = Field(default_factory=dict)
|
||||
"""中间件运行时的临时状态存储。"""
|
||||
cancellation_token: CancellationToken | None = Field(default=None)
|
||||
"""全局取消令牌。"""
|
||||
|
||||
model_config = ConfigDict(arbitrary_types_allowed=True)
|
||||
|
||||
|
||||
class ToolDefinition(BaseModel):
|
||||
"""结构化的工具定义模型"""
|
||||
|
||||
name: str = Field(...)
|
||||
"""工具名称"""
|
||||
description: str = Field(...)
|
||||
"""工具描述"""
|
||||
parameters: dict[str, Any] = Field(default_factory=dict)
|
||||
"""JSON Schema 参数"""
|
||||
metadata: dict[str, Any] = Field(default_factory=dict)
|
||||
"""元数据"""
|
||||
|
||||
|
||||
class ToolChoice(BaseModel):
|
||||
"""工具选择配置"""
|
||||
|
||||
mode: Literal["auto", "none", "any", "required"] = Field(default="auto")
|
||||
"""工具选择模式"""
|
||||
allowed_function_names: list[str] | None = Field(default=None)
|
||||
"""允许调用的函数名称列表"""
|
||||
|
||||
|
||||
class ModelModality(str, Enum):
|
||||
TEXT = "text"
|
||||
IMAGE = "image"
|
||||
AUDIO = "audio"
|
||||
VIDEO = "video"
|
||||
FILE = "file"
|
||||
EMBEDDING = "embedding"
|
||||
|
||||
|
||||
class ReasoningMode(str, Enum):
|
||||
"""推理/思考模式类型"""
|
||||
|
||||
NONE = "none"
|
||||
BUDGET = "budget"
|
||||
LEVEL = "level"
|
||||
EFFORT = "effort"
|
||||
|
||||
|
||||
class ModelCapabilities(BaseModel):
|
||||
"""定义一个模型的核心能力。"""
|
||||
|
||||
input_modalities: set[ModelModality] = Field(default={ModelModality.TEXT})
|
||||
"""模型支持的输入模态集合。"""
|
||||
output_modalities: set[ModelModality] = Field(default={ModelModality.TEXT})
|
||||
"""模型支持的输出模态集合。"""
|
||||
supports_tool_calling: bool = False
|
||||
"""是否支持工具调用能力。"""
|
||||
is_embedding_model: bool = False
|
||||
"""是否为嵌入模型。"""
|
||||
is_rerank_model: bool = False
|
||||
"""是否为重排序模型。"""
|
||||
reasoning_mode: ReasoningMode = ReasoningMode.NONE
|
||||
"""推理模式类型。"""
|
||||
reasoning_visibility: Literal["visible", "hidden", "none"] = "none"
|
||||
"""推理过程可见性设置。"""
|
||||
reasoning_effort_map: dict[str, str] = Field(default_factory=dict)
|
||||
"""思考强度参数(reasoning_effort)的降级映射矩阵，如 {"max": "xhigh"}。"""
|
||||
max_input_tokens: int = Field(default=256000)
|
||||
"""最大输入 Token 数量（用于触发上下文压缩策略，未显式声明则默认为 256K）。"""
|
||||
supported_native_tools: set[str] = Field(default_factory=set)
|
||||
"""该模型实际支持的云端原生能力/内置工具。"""
|
||||
default_voice_id: str | None = None
|
||||
"""默认的语音合成音色 ID（TTS模型专用）。"""
|
||||
|
||||
features: set[str] = Field(default_factory=set)
|
||||
"""用于第三方插件动态注入的自定义能力标签。"""
|
||||
|
||||
def supports_task(self, task: str) -> bool:
|
||||
"""判断模型是否支持指定的底层任务类型"""
|
||||
if task == "embedding":
|
||||
return self.is_embedding_model
|
||||
elif task == "rerank":
|
||||
return self.is_rerank_model
|
||||
elif task == "tts":
|
||||
return ModelModality.AUDIO in self.output_modalities
|
||||
elif task == "image":
|
||||
return ModelModality.IMAGE in self.output_modalities
|
||||
elif task == "chat":
|
||||
return ModelModality.TEXT in self.output_modalities
|
||||
return False
|
||||
|
||||
def accepts_input(self, modality: ModelModality) -> bool:
|
||||
"""检查模型是否支持某种输入模态"""
|
||||
return modality in self.input_modalities
|
||||
|
||||
def accepts_output(self, modality: ModelModality) -> bool:
|
||||
"""检查模型是否支持某种输出模态"""
|
||||
return modality in self.output_modalities
|
||||
|
||||
def has_feature(self, feature: str) -> bool:
|
||||
"""检查模型是否具备某个扩展特性"""
|
||||
return feature in self.features
|
||||
|
||||
|
||||
class ModelDetail(BaseModel):
|
||||
"""模型详细信息"""
|
||||
|
||||
model_name: str
|
||||
"""模型名称。"""
|
||||
is_available: bool = True
|
||||
"""模型是否可用。"""
|
||||
temperature: float | None = None
|
||||
"""采样温度参数。"""
|
||||
generation_max_tokens: int | None = None
|
||||
"""单次生成最大 Token 数。"""
|
||||
api_type: str | None = None
|
||||
"""API 类型标识。"""
|
||||
endpoint: str | None = None
|
||||
"""模型服务端点地址。"""
|
||||
task_type: str | None = Field(default=None)
|
||||
"""显式声明的主任务类型 (如 'image_generation')。"""
|
||||
path_prefix: str | None = Field(default=None)
|
||||
"""中转路由前缀，例如 '/cogvideox' 或 '/minimax'。"""
|
||||
|
||||
|
||||
__all__ = [
|
||||
"CancellationToken",
|
||||
"LLMContext",
|
||||
"ModelCapabilities",
|
||||
"ModelDetail",
|
||||
"ModelIdentity",
|
||||
"ModelModality",
|
||||
"ModelName",
|
||||
"ReasoningMode",
|
||||
"ToolChoice",
|
||||
"ToolDefinition",
|
||||
]
|
||||
Reference in New Issue
Block a user