mirror of
https://github.com/zhenxun-org/zhenxun_bot.git
synced 2026-10-11 06:39:59 +08:00
80fc5b86a7
* ✨ 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>
95 lines
3.5 KiB
Python
95 lines
3.5 KiB
Python
from collections.abc import AsyncGenerator
|
||
from contextlib import asynccontextmanager
|
||
import json
|
||
from typing import Any
|
||
|
||
from anyio import create_memory_object_stream, create_task_group
|
||
from mcp.shared.message import SessionMessage
|
||
from mcp.types import JSONRPCMessage
|
||
|
||
from zhenxun.services.ai.sandbox.addons.base import BaseMcpProxyExtension
|
||
from zhenxun.services.ai.sandbox.protocols import SupportsStreamExecution
|
||
from zhenxun.services.ai.sandbox.registry import SandboxRegistry
|
||
from zhenxun.services.log import logger
|
||
from zhenxun.utils.pydantic_compat import model_dump_json, model_validate
|
||
|
||
|
||
class UniversalMcpExtension(BaseMcpProxyExtension):
|
||
@property
|
||
def extension_name(self) -> str:
|
||
return "universal_mcp"
|
||
|
||
@asynccontextmanager
|
||
async def connect_mcp(
|
||
self, command: str, args: list[str], env: dict[str, str] | None = None
|
||
) -> AsyncGenerator[tuple[Any, Any], None]:
|
||
if not isinstance(self.session, SupportsStreamExecution):
|
||
raise RuntimeError(
|
||
"当前沙箱驱动不支持流式后台进程执行 (SupportsStreamExecution)，"
|
||
"无法启动原生 MCP 代理。"
|
||
)
|
||
|
||
logger.info(
|
||
"[UniversalMcpExtension] 正在沙箱内原生启动 MCP 服务器: "
|
||
f"{command} {' '.join(args)}"
|
||
)
|
||
|
||
cmd_list = [command, *args]
|
||
|
||
async with self.session.create_stream_process(
|
||
command=cmd_list, cwd=self.session.workspace_path, env=env
|
||
) as process_stream:
|
||
read_prod, read_cons = create_memory_object_stream(10)
|
||
write_prod, write_cons = create_memory_object_stream(10)
|
||
|
||
async def stream_reader():
|
||
buffer = b""
|
||
try:
|
||
while True:
|
||
msg = await process_stream.read()
|
||
if msg is None:
|
||
break
|
||
if msg.stream_type == 1:
|
||
buffer += msg.data
|
||
while b"\n" in buffer:
|
||
line, buffer = buffer.split(b"\n", 1)
|
||
if not line.strip():
|
||
continue
|
||
try:
|
||
msg = model_validate(
|
||
JSONRPCMessage, json.loads(line)
|
||
)
|
||
await read_prod.send(SessionMessage(message=msg))
|
||
except Exception as exc:
|
||
await read_prod.send(exc)
|
||
except Exception:
|
||
pass
|
||
finally:
|
||
await read_prod.aclose()
|
||
|
||
async def stream_writer():
|
||
try:
|
||
async for msg in write_cons:
|
||
data = (
|
||
model_dump_json(
|
||
msg.message, by_alias=True, exclude_none=True
|
||
).encode("utf-8")
|
||
+ b"\n"
|
||
)
|
||
await process_stream.write(data)
|
||
except Exception:
|
||
pass
|
||
|
||
async with create_task_group() as tg:
|
||
tg.start_soon(stream_reader)
|
||
tg.start_soon(stream_writer)
|
||
yield read_cons, write_prod
|
||
tg.cancel_scope.cancel()
|
||
|
||
|
||
SandboxRegistry.register_extension(UniversalMcpExtension)
|
||
|
||
__all__ = [
|
||
"UniversalMcpExtension",
|
||
]
|