mirror of
https://github.com/zhenxun-org/zhenxun_bot.git
synced 2026-10-08 13:20:01 +08:00
♻️ 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>
This commit is contained in:
Rumio
co-authored by
webjoin111
pre-commit-ci[bot]
webjoin111
pre-commit-ci[bot]
parent
922d092650
commit
52f7dbdedf
@@ -4,6 +4,7 @@ import threading
|
||||
from typing import Any, Literal, Protocol, runtime_checkable
|
||||
|
||||
from zhenxun.services.ai.core.messages import EmbedBatch
|
||||
from zhenxun.services.ai.core.options import LLMEmbeddingConfig
|
||||
from zhenxun.services.ai.llm.api import embed as api_embed
|
||||
from zhenxun.services.ai.message_builder import MessageBuilder
|
||||
from zhenxun.services.ai.utils.logger import log_rag as logger
|
||||
@@ -16,7 +17,7 @@ EmbedTaskType = Literal[
|
||||
@runtime_checkable
|
||||
class Embedder(Protocol):
|
||||
"""
|
||||
向量化引擎协议 (Callable Protocol)。
|
||||
向量化引擎协议。
|
||||
任何实现了异步 __call__ 的对象或闭包函数均可作为 Embedder。
|
||||
"""
|
||||
|
||||
@@ -32,7 +33,9 @@ class Embedder(Protocol):
|
||||
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):
|
||||
def __init__(
|
||||
self, model_name: str | None = None, config: LLMEmbeddingConfig | None = None
|
||||
):
|
||||
self.model_name = model_name
|
||||
self.config = config
|
||||
|
||||
@@ -57,10 +60,28 @@ class BaseLocalEmbedder(Embedder, ABC):
|
||||
def __init__(self, model_name: str):
|
||||
self.model_name = model_name
|
||||
self._model_lock = threading.Lock()
|
||||
self._model: Any | None = None
|
||||
|
||||
def _ensure_model_loaded(self) -> None:
|
||||
"""线程安全的懒加载机制"""
|
||||
if self._model is None:
|
||||
with self._model_lock:
|
||||
if self._model is None:
|
||||
logger.info(
|
||||
f"正在后台加载本地向量模型: {self.model_name} ... "
|
||||
"(首次加载可能需要较长时间下载)"
|
||||
)
|
||||
self._model = self._load_model_impl()
|
||||
logger.info(f"本地向量模型 {self.model_name} 加载完毕！")
|
||||
|
||||
@abstractmethod
|
||||
def _load_model_impl(self) -> Any:
|
||||
"""子类实现：执行具体的依赖导入与模型实例化，并返回模型对象。"""
|
||||
pass
|
||||
|
||||
@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。"""
|
||||
"""子类实现：执行同步的批量文本向量化方法。"""
|
||||
pass
|
||||
|
||||
async def __call__(
|
||||
@@ -93,7 +114,6 @@ class FastEmbedder(BaseLocalEmbedder):
|
||||
|
||||
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
|
||||
|
||||
@@ -102,29 +122,19 @@ class FastEmbedder(BaseLocalEmbedder):
|
||||
"⚠️ 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 _load_model_impl(self) -> Any:
|
||||
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"
|
||||
)
|
||||
return TextEmbedding(model_name=self.model_name)
|
||||
|
||||
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)]
|
||||
assert self._model is not None
|
||||
return [vec.tolist() for vec in self._model.embed(texts)]
|
||||
|
||||
|
||||
class SentenceTransformerEmbedder(BaseLocalEmbedder):
|
||||
@@ -135,7 +145,6 @@ class SentenceTransformerEmbedder(BaseLocalEmbedder):
|
||||
|
||||
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
|
||||
|
||||
@@ -145,32 +154,18 @@ class SentenceTransformerEmbedder(BaseLocalEmbedder):
|
||||
"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 _load_model_impl(self) -> Any:
|
||||
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"
|
||||
)
|
||||
return SentenceTransformer(self.model_name)
|
||||
|
||||
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)
|
||||
assert self._model is not None
|
||||
embeddings = self._model.encode(texts)
|
||||
return embeddings.tolist()
|
||||
|
||||
Reference in New Issue
Block a user