当前位置：首页 > news >正文

langchain Chroma 构建本地向量数据库

news 2025/8/27 10:26:39

langchain Chroma 构建本地向量数据库

# import
from langchain_community.document_loaders import TextLoader
from langchain_community.embeddings.sentence_transformer import (SentenceTransformerEmbeddings,
)
from langchain_community.embeddings import HuggingFaceEmbeddings 
from langchain_community.vectorstores import Chroma
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.document_loaders.word_document import Docx2txtLoaderimport glob
import os# 数据库路径
db_dir = "./db"
# 文档路径
source_directory = "./docs"
# 文件后缀
file_ext = '*.docx'# create the open-source embedding function
# embedding_function = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
# 使用中文嵌入层编码器
ebd_function = HuggingFaceEmbeddings(model_name="shibing624/text2vec-base-chinese")def add_files_to_db(filepath:str="",file_ext:str=""):docx_files = glob.glob(os.path.join(source_directory, file_ext))text_list=[]for file_name in docx_files:print(file_name)loader = Docx2txtLoader(file_name)documents = loader.load()text_list.extend(documents)# split it into chunkstext_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=50)docs = text_splitter.split_documents(text_list)# load it into Chromadb = Chroma.from_documents(docs, ebd_function, persist_directory=db_dir)# save db to diskdb.persist()def query_db(db:Chroma,query:str=""):# query itdocs = db.similarity_search(query)# print resultsprint(docs[0].page_content)print("-----------------------------------------")if __name__=="__main__":# 只需执行一次# add_files_to_db(source_directory,file_ext)db = Chroma(persist_directory=db_dir,embedding_function=ebd_function)query = "怎么治疗骨质疏松症?"query_db(db,query)query = "怎么治疗鼻炎?"query_db(db,query)db = Nonepass