mirror of
https://github.com/RYDE-WORK/Langchain-Chatchat.git
synced 2026-01-19 13:23:16 +08:00
deed92169f
* 新功能： - 支持在线 Embeddings：zhipu-api, qwen-api, minimax-api, qianfan-api - API 增加 /other/embed_texts 接口 - init_database.py 增加 --embed-model 参数，可以指定使用的嵌入模型（本地或在线均可） 问题修复： - API 中 list_config_models 会删除 ONLINE_LLM_MODEL 中的敏感信息，导致第二轮API请求错误 开发者： - 优化 kb_service 中 Embeddings 操作： - 统一加载接口： server.utils.load_embeddings，利用全局缓存避免各处 Embeddings 传参 - 统一文本嵌入接口：server.embedding_api.[embed_texts, embed_documents]
97 lines
3.3 KiB
Python
97 lines
3.3 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, EmbeddingsFunAdapter, \
|
|
score_threshold_process
|
|
from server.knowledge_base.utils import KnowledgeFile
|
|
|
|
|
|
class ZillizKBService(KBService):
|
|
zilliz: Zilliz
|
|
|
|
@staticmethod
|
|
def get_collection(zilliz_name):
|
|
from pymilvus import Collection
|
|
return Collection(zilliz_name)
|
|
|
|
# def save_vector_store(self):
|
|
# if self.zilliz.col:
|
|
# self.zilliz.col.flush()
|
|
|
|
def get_doc_by_id(self, id: str) -> Optional[Document]:
|
|
if self.zilliz.col:
|
|
data_list = self.zilliz.col.query(expr=f'pk == {id}', output_fields=["*"])
|
|
if len(data_list) > 0:
|
|
data = data_list[0]
|
|
text = data.pop("text")
|
|
return Document(page_content=text, metadata=data)
|
|
|
|
@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=EmbeddingsFunAdapter(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()
|
|
return score_threshold_process(score_threshold, top_k, self.zilliz.similarity_search_with_score(query, top_k))
|
|
|
|
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")
|
|
|