mirror of
https://github.com/zhenxun-org/zhenxun_bot.git
synced 2026-10-16 01:30:03 +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 +1,176 @@
|
||||
from abc import ABC, abstractmethod
|
||||
import asyncio
|
||||
import threading
|
||||
from typing import Any, Literal, Protocol, runtime_checkable
|
||||
|
||||
from zhenxun.services.ai.core.messages import EmbedBatch
|
||||
from zhenxun.services.ai.llm.api import embed as api_embed
|
||||
from zhenxun.services.ai.message_builder import MessageBuilder
|
||||
from zhenxun.services.log import logger
|
||||
|
||||
EmbedTaskType = Literal[
|
||||
"general", "query", "document", "similarity", "classification", "clustering"
|
||||
]
|
||||
|
||||
|
||||
@runtime_checkable
|
||||
class Embedder(Protocol):
|
||||
"""
|
||||
向量化引擎协议 (Callable Protocol)。
|
||||
任何实现了异步 __call__ 的对象或闭包函数均可作为 Embedder。
|
||||
"""
|
||||
|
||||
async def __call__(
|
||||
self, input_batch: Any, task: EmbedTaskType = "general", **kwargs
|
||||
) -> list[list[float]]:
|
||||
"""
|
||||
将文本、多模态或预构建的 EmbedBatch 转换为向量列表。
|
||||
"""
|
||||
...
|
||||
|
||||
|
||||
class DefaultEmbedder(Embedder):
|
||||
"""系统默认的向量化引擎，调用大模型底座 API"""
|
||||
|
||||
def __init__(self, model_name: str | None = None, config: Any = None):
|
||||
self.model_name = model_name
|
||||
self.config = config
|
||||
|
||||
async def __call__(
|
||||
self, input_batch: Any, task: EmbedTaskType = "general", **kwargs
|
||||
) -> list[list[float]]:
|
||||
if not input_batch:
|
||||
return []
|
||||
try:
|
||||
res = await api_embed(
|
||||
input_batch, model=self.model_name, task=task, config=self.config
|
||||
)
|
||||
return res.embeddings
|
||||
except Exception as e:
|
||||
logger.error(f"DefaultEmbedder 向量化失败: {e}", e=e)
|
||||
return []
|
||||
|
||||
|
||||
class BaseLocalEmbedder(Embedder, ABC):
|
||||
"""本地向量化引擎基类，统一处理多模态降级与同步推理由协程包裹逻辑。"""
|
||||
|
||||
def __init__(self, model_name: str):
|
||||
self.model_name = model_name
|
||||
self._model_lock = threading.Lock()
|
||||
|
||||
@abstractmethod
|
||||
def _encode_texts(self, texts: list[str]) -> list[list[float]]:
|
||||
"""子类只需实现此同步的批量文本向量化方法即可。"""
|
||||
pass
|
||||
|
||||
async def __call__(
|
||||
self, input_batch: Any, task: EmbedTaskType = "general", **kwargs
|
||||
) -> list[list[float]]:
|
||||
if not input_batch:
|
||||
return []
|
||||
|
||||
if isinstance(input_batch, EmbedBatch):
|
||||
batch = input_batch
|
||||
else:
|
||||
batch = await MessageBuilder.normalize_to_embed_batch(input_batch)
|
||||
|
||||
texts = batch.to_text_only(f"本地模型 {self.model_name}")
|
||||
|
||||
if not texts:
|
||||
return []
|
||||
|
||||
def _sync_embed():
|
||||
return self._encode_texts(texts)
|
||||
|
||||
return await asyncio.to_thread(_sync_embed)
|
||||
|
||||
|
||||
class FastEmbedder(BaseLocalEmbedder):
|
||||
"""
|
||||
基于 FastEmbed 的轻量级本地向量化引擎。
|
||||
零 PyTorch 依赖，CPU 推理极快。
|
||||
"""
|
||||
|
||||
def __init__(self, model_name: str | None = None):
|
||||
super().__init__(model_name or "BAAI/bge-small-zh-v1.5")
|
||||
self.model = None
|
||||
|
||||
import importlib.util
|
||||
|
||||
if importlib.util.find_spec("fastembed") is None:
|
||||
raise ImportError(
|
||||
"⚠️ 使用 FastEmbed 需要额外依赖，请在终端执行: pip install fastembed"
|
||||
)
|
||||
|
||||
def _ensure_model_loaded(self):
|
||||
"""线程安全的懒加载机制"""
|
||||
if self.model is None:
|
||||
with self._model_lock:
|
||||
if self.model is None:
|
||||
try:
|
||||
from fastembed import TextEmbedding
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"⚠️ 使用 FastEmbed 需要额外依赖，"
|
||||
"请在终端执行: pip install fastembed"
|
||||
)
|
||||
logger.info(
|
||||
f"正在后台加载 FastEmbed 本地模型: {self.model_name} ... "
|
||||
"(首次加载可能需要极长时间下载)"
|
||||
)
|
||||
self.model = TextEmbedding(model_name=self.model_name)
|
||||
logger.info(f"FastEmbed 模型 {self.model_name} 加载完毕！")
|
||||
|
||||
def _encode_texts(self, texts: list[str]) -> list[list[float]]:
|
||||
self._ensure_model_loaded()
|
||||
assert self.model is not None
|
||||
return [vec.tolist() for vec in self.model.embed(texts)]
|
||||
|
||||
|
||||
class SentenceTransformerEmbedder(BaseLocalEmbedder):
|
||||
"""
|
||||
基于 Sentence-Transformers 的本地向量化引擎。
|
||||
支持 GPU 加速，适合重度用户。
|
||||
"""
|
||||
|
||||
def __init__(self, model_name: str | None = None):
|
||||
super().__init__(model_name or "BAAI/bge-small-zh-v1.5")
|
||||
self.model = None
|
||||
|
||||
import importlib.util
|
||||
|
||||
if importlib.util.find_spec("sentence_transformers") is None:
|
||||
raise ImportError(
|
||||
"⚠️ 使用 SentenceTransformers 需要额外依赖，"
|
||||
"请在终端执行: pip install sentence-transformers"
|
||||
)
|
||||
|
||||
def _ensure_model_loaded(self):
|
||||
"""线程安全的懒加载机制"""
|
||||
if self.model is None:
|
||||
with self._model_lock:
|
||||
if self.model is None:
|
||||
try:
|
||||
from sentence_transformers import (
|
||||
SentenceTransformer,
|
||||
)
|
||||
except ImportError:
|
||||
raise ImportError(
|
||||
"⚠️ 使用 SentenceTransformers 需要额外依赖，"
|
||||
"请在终端执行: pip install sentence-transformers"
|
||||
)
|
||||
logger.info(
|
||||
"正在后台加载 SentenceTransformer "
|
||||
f"本地模型: {self.model_name} ... "
|
||||
"(首次加载可能需要极长时间下载)"
|
||||
)
|
||||
self.model = SentenceTransformer(self.model_name)
|
||||
logger.info(
|
||||
f"SentenceTransformer 模型 {self.model_name} 加载完毕！"
|
||||
)
|
||||
|
||||
def _encode_texts(self, texts: list[str]) -> list[list[float]]:
|
||||
self._ensure_model_loaded()
|
||||
assert self.model is not None
|
||||
embeddings = self.model.encode(texts)
|
||||
return embeddings.tolist()
|
||||
Reference in New Issue
Block a user