跳到主要内容
Open In ColabOpen on GitHub

DashVector

DashVector 是一种全托管的向量数据库服务，支持高维稠密和稀疏向量，提供实时插入和带过滤的搜索。它旨在自动扩展并能适应不同的应用需求。

本 Notebook 展示了如何使用与 DashVector 向量数据库相关的功能。

要使用 DashVector，您必须拥有一个 API 密钥。此处是安装说明。

安装​

%pip install --upgrade --quiet  langchain-community dashvector dashscope

我们要使用 DashScopeEmbeddings，因此我们还需要获取 DashScope API 密钥。

import getpass
import os

if "DASHVECTOR_API_KEY" not in os.environ:
    os.environ["DASHVECTOR_API_KEY"] = getpass.getpass("DashVector API Key:")
if "DASHSCOPE_API_KEY" not in os.environ:
    os.environ["DASHSCOPE_API_KEY"] = getpass.getpass("DashScope API Key:")

示例​

from langchain_community.embeddings.dashscope import DashScopeEmbeddings
from langchain_community.vectorstores import DashVector
from langchain_text_splitters import CharacterTextSplitter
from langchain_community.document_loaders import TextLoader

loader = TextLoader("../../how_to/state_of_the_union.txt")
documents = loader.load()
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
docs = text_splitter.split_documents(documents)

embeddings = DashScopeEmbeddings()
API 参考：TextLoader

我们可以从文档创建 DashVector。

dashvector = DashVector.from_documents(docs, embeddings)

query = "What did the president say about Ketanji Brown Jackson"
docs = dashvector.similarity_search(query)
print(docs)

我们可以添加带有元数据和 ID 的文本，并使用元过滤器进行搜索。

texts = ["foo", "bar", "baz"]
metadatas = [{"key": i} for i in range(len(texts))]
ids = ["0", "1", "2"]

dashvector.add_texts(texts, metadatas=metadatas, ids=ids)

docs = dashvector.similarity_search("foo", filter="key = 2")
print(docs)
[Document(page_content='baz', metadata={'key': 2})]

操作带 partition 参数​

partition 参数默认为 default，如果传入一个不存在的 partition 参数，该 partition 将自动创建。

texts = ["foo", "bar", "baz"]
metadatas = [{"key": i} for i in range(len(texts))]
ids = ["0", "1", "2"]
partition = "langchain"

# add texts
dashvector.add_texts(texts, metadatas=metadatas, ids=ids, partition=partition)

# similarity search
query = "What did the president say about Ketanji Brown Jackson"
docs = dashvector.similarity_search(query, partition=partition)

# delete
dashvector.delete(ids=ids, partition=partition)