mirror of
https://github.com/RYDE-WORK/Langchain-Chatchat.git
synced 2026-01-31 19:33:26 +08:00
d0846f88cc
- 重构 api.py：
- 按模块划分为不同的 router
- 添加 openai 兼容的转发接口，项目默认使用该接口以实现模型负载均衡
- 添加 /tools 接口，可以获取/调用编写的 agent tools
- 移除所有 EmbeddingFuncAdapter，统一改用 get_Embeddings
- 待办：
- /chat/chat 接口改为 openai 兼容
- 添加 /chat/kb_chat 接口，openai 兼容
- 改变 ntlk/knowledge_base/logs 等数据目录位置
99 lines
3.5 KiB
Python
99 lines
3.5 KiB
Python
from typing import List, Dict, Optional
|
|
from langchain.embeddings.base import Embeddings
|
|
from langchain.schema import Document
|
|
from langchain.vectorstores import Zilliz
|
|
from configs import kbs_config
|
|
from server.knowledge_base.kb_service.base import KBService, SupportedVSType, \
|
|
score_threshold_process
|
|
from server.knowledge_base.utils import KnowledgeFile
|
|
from server.utils import get_Embeddings
|
|
|
|
|
|
class ZillizKBService(KBService):
|
|
zilliz: Zilliz
|
|
|
|
@staticmethod
|
|
def get_collection(zilliz_name):
|
|
from pymilvus import Collection
|
|
return Collection(zilliz_name)
|
|
|
|
def get_doc_by_ids(self, ids: List[str]) -> List[Document]:
|
|
result = []
|
|
if self.zilliz.col:
|
|
# ids = [int(id) for id in ids] # for zilliz if needed #pr 2725
|
|
data_list = self.zilliz.col.query(expr=f'pk in {ids}', output_fields=["*"])
|
|
for data in data_list:
|
|
text = data.pop("text")
|
|
result.append(Document(page_content=text, metadata=data))
|
|
return result
|
|
|
|
def del_doc_by_ids(self, ids: List[str]) -> bool:
|
|
self.zilliz.col.delete(expr=f'pk in {ids}')
|
|
|
|
@staticmethod
|
|
def search(zilliz_name, content, limit=3):
|
|
search_params = {
|
|
"metric_type": "IP",
|
|
"params": {},
|
|
}
|
|
c = ZillizKBService.get_collection(zilliz_name)
|
|
return c.search(content, "embeddings", search_params, limit=limit, output_fields=["content"])
|
|
|
|
def do_create_kb(self):
|
|
pass
|
|
|
|
def vs_type(self) -> str:
|
|
return SupportedVSType.ZILLIZ
|
|
|
|
def _load_zilliz(self):
|
|
zilliz_args = kbs_config.get("zilliz")
|
|
self.zilliz = Zilliz(embedding_function=get_Embeddings(self.embed_model),
|
|
collection_name=self.kb_name, connection_args=zilliz_args)
|
|
|
|
def do_init(self):
|
|
self._load_zilliz()
|
|
|
|
def do_drop_kb(self):
|
|
if self.zilliz.col:
|
|
self.zilliz.col.release()
|
|
self.zilliz.col.drop()
|
|
|
|
def do_search(self, query: str, top_k: int, score_threshold: float):
|
|
self._load_zilliz()
|
|
embed_func = get_Embeddings(self.embed_model)
|
|
embeddings = embed_func.embed_query(query)
|
|
docs = self.zilliz.similarity_search_with_score_by_vector(embeddings, top_k)
|
|
return score_threshold_process(score_threshold, top_k, docs)
|
|
|
|
def do_add_doc(self, docs: List[Document], **kwargs) -> List[Dict]:
|
|
for doc in docs:
|
|
for k, v in doc.metadata.items():
|
|
doc.metadata[k] = str(v)
|
|
for field in self.zilliz.fields:
|
|
doc.metadata.setdefault(field, "")
|
|
doc.metadata.pop(self.zilliz._text_field, None)
|
|
doc.metadata.pop(self.zilliz._vector_field, None)
|
|
|
|
ids = self.zilliz.add_documents(docs)
|
|
doc_infos = [{"id": id, "metadata": doc.metadata} for id, doc in zip(ids, docs)]
|
|
return doc_infos
|
|
|
|
def do_delete_doc(self, kb_file: KnowledgeFile, **kwargs):
|
|
if self.zilliz.col:
|
|
filepath = kb_file.filepath.replace('\\', '\\\\')
|
|
delete_list = [item.get("pk") for item in
|
|
self.zilliz.col.query(expr=f'source == "{filepath}"', output_fields=["pk"])]
|
|
self.zilliz.col.delete(expr=f'pk in {delete_list}')
|
|
|
|
def do_clear_vs(self):
|
|
if self.zilliz.col:
|
|
self.do_drop_kb()
|
|
self.do_init()
|
|
|
|
|
|
if __name__ == '__main__':
|
|
from server.db.base import Base, engine
|
|
|
|
Base.metadata.create_all(bind=engine)
|
|
zillizService = ZillizKBService("test")
|