向量存储索引使用示例¶

本指南将展示如何搭配不同的向量存储实现来使用向量存储索引。

从如何通过默认内存向量存储和基础查询配置快速入门（仅需几行代码），到使用自定义托管向量存储及元数据过滤器等高级设置。

构建向量存储与索引¶

默认配置

默认情况下，VectorStoreIndex 会使用内存中的 SimpleVectorStore，该存储会作为默认存储上下文的一部分进行初始化。

In [ ]:

from llama_index import VectorStoreIndex, SimpleDirectoryReader

# Load documents and build index
documents = SimpleDirectoryReader(
    "../../examples/data/paul_graham"
).load_data()
index = VectorStoreIndex.from_documents(documents)

from llama_index import VectorStoreIndex, SimpleDirectoryReader # Load documents and build index documents = SimpleDirectoryReader( "../../examples/data/paul_graham" ).load_data() index = VectorStoreIndex.from_documents(documents)

自定义向量存储

您可以通过以下方式使用自定义向量存储（本例中为 PineconeVectorStore）：

In [ ]:

import pinecone
from llama_index import VectorStoreIndex, SimpleDirectoryReader, StorageContext
from llama_index.vector_stores import PineconeVectorStore

# init pinecone
pinecone.init(api_key="<api_key>", environment="<environment>")
pinecone.create_index(
    "quickstart", dimension=1536, metric="euclidean", pod_type="p1"
)

# construct vector store and customize storage context
storage_context = StorageContext.from_defaults(
    vector_store=PineconeVectorStore(pinecone.Index("quickstart"))
)

# Load documents and build index
documents = SimpleDirectoryReader(
    "../../examples/data/paul_graham"
).load_data()
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)

import pinecone from llama_index import VectorStoreIndex, SimpleDirectoryReader, StorageContext from llama_index.vector_stores import PineconeVectorStore # init pinecone pinecone.init(api_key="", environment="") pinecone.create_index( "quickstart", dimension=1536, metric="euclidean", pod_type="p1" ) # construct vector store and customize storage context storage_context = StorageContext.from_defaults( vector_store=PineconeVectorStore(pinecone.Index("quickstart")) ) # Load documents and build index documents = SimpleDirectoryReader( "../../examples/data/paul_graham" ).load_data() index = VectorStoreIndex.from_documents( documents, storage_context=storage_context )

如需查看更多初始化不同向量存储的示例， 请参阅向量存储集成。

连接外部向量数据库（使用现有嵌入向量）¶

若您已计算好嵌入向量并将其存入外部向量数据库（如 Pinecone、Chroma），可通过以下方式与 LlamaIndex 配合使用：

In [ ]:

vector_store = PineconeVectorStore(pinecone.Index("quickstart"))
index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

vector_store = PineconeVectorStore(pinecone.Index("quickstart")) index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

查询¶

默认设置

您可以通过获取默认查询引擎开始进行查询：

In [ ]:

query_engine = index.as_query_engine()
response = query_engine.query("What did the author do growing up?")

query_engine = index.as_query_engine() response = query_engine.query("What did the author do growing up?")

配置标准查询设置

要配置查询设置，您可以在构建查询引擎时直接将其作为关键字参数传递：

In [ ]:

from llama_index.vector_stores.types import ExactMatchFilter, MetadataFilters

query_engine = index.as_query_engine(
    similarity_top_k=3,
    vector_store_query_mode="default",
    filters=MetadataFilters(
        filters=[
            ExactMatchFilter(key="name", value="paul graham"),
        ]
    ),
    alpha=None,
    doc_ids=None,
)
response = query_engine.query("what did the author do growing up?")

from llama_index.vector_stores.types import ExactMatchFilter, MetadataFilters query_engine = index.as_query_engine( similarity_top_k=3, vector_store_query_mode="default", filters=MetadataFilters( filters=[ ExactMatchFilter(key="name", value="paul graham"), ] ), alpha=None, doc_ids=None, ) response = query_engine.query("what did the author do growing up?")

请注意，元数据筛选是针对 Node.metadata 中指定的元数据进行的。

或者，如果您正在使用较低级别的组合式 API：

In [ ]:

from llama_index import get_response_synthesizer
from llama_index.indices.vector_store.retrievers import VectorIndexRetriever
from llama_index.query_engine.retriever_query_engine import (
    RetrieverQueryEngine,
)

# build retriever
retriever = VectorIndexRetriever(
    index=index,
    similarity_top_k=3,
    vector_store_query_mode="default",
    filters=[ExactMatchFilter(key="name", value="paul graham")],
    alpha=None,
    doc_ids=None,
)

# build query engine
query_engine = RetrieverQueryEngine(
    retriever=retriever, response_synthesizer=get_response_synthesizer()
)

# query
response = query_engine.query("what did the author do growing up?")

from llama_index import get_response_synthesizer from llama_index.indices.vector_store.retrievers import VectorIndexRetriever from llama_index.query_engine.retriever_query_engine import ( RetrieverQueryEngine, ) # build retriever retriever = VectorIndexRetriever( index=index, similarity_top_k=3, vector_store_query_mode="default", filters=[ExactMatchFilter(key="name", value="paul graham")], alpha=None, doc_ids=None, ) # build query engine query_engine = RetrieverQueryEngine( retriever=retriever, response_synthesizer=get_response_synthesizer() ) # query response = query_engine.query("what did the author do growing up?")

配置向量存储的特定关键字参数

您还可以通过传入 vector_store_kwargs 来自定义特定向量存储实现独有的关键字参数。

In [ ]:

query_engine = index.as_query_engine(
    similarity_top_k=3,
    # only works for pinecone
    vector_store_kwargs={
        "filter": {"name": "paul graham"},
    },
)
response = query_engine.query("what did the author do growing up?")

query_engine = index.as_query_engine( similarity_top_k=3, # only works for pinecone vector_store_kwargs={ "filter": {"name": "paul graham"}, }, ) response = query_engine.query("what did the author do growing up?")

使用自动检索器

你也可以利用大语言模型（LLM）自动为你决定查询设置！ 目前我们支持自动配置精确匹配的元数据过滤器和 top k 参数。

In [ ]:

from llama_index import get_response_synthesizer
from llama_index.indices.vector_store.retrievers import (
    VectorIndexAutoRetriever,
)
from llama_index.query_engine.retriever_query_engine import (
    RetrieverQueryEngine,
)
from llama_index.vector_stores.types import MetadataInfo, VectorStoreInfo


vector_store_info = VectorStoreInfo(
    content_info="brief biography of celebrities",
    metadata_info=[
        MetadataInfo(
            name="category",
            type="str",
            description="Category of the celebrity, one of [Sports, Entertainment, Business, Music]",
        ),
        MetadataInfo(
            name="country",
            type="str",
            description="Country of the celebrity, one of [United States, Barbados, Portugal]",
        ),
    ],
)

# build retriever
retriever = VectorIndexAutoRetriever(
    index, vector_store_info=vector_store_info
)

# build query engine
query_engine = RetrieverQueryEngine(
    retriever=retriever, response_synthesizer=get_response_synthesizer()
)

# query
response = query_engine.query(
    "Tell me about two celebrities from United States"
)

from llama_index import get_response_synthesizer from llama_index.indices.vector_store.retrievers import ( VectorIndexAutoRetriever, ) from llama_index.query_engine.retriever_query_engine import ( RetrieverQueryEngine, ) from llama_index.vector_stores.types import MetadataInfo, VectorStoreInfo vector_store_info = VectorStoreInfo( content_info="brief biography of celebrities", metadata_info=[ MetadataInfo( name="category", type="str", description="Category of the celebrity, one of [Sports, Entertainment, Business, Music]", ), MetadataInfo( name="country", type="str", description="Country of the celebrity, one of [United States, Barbados, Portugal]", ), ], ) # build retriever retriever = VectorIndexAutoRetriever( index, vector_store_info=vector_store_info ) # build query engine query_engine = RetrieverQueryEngine( retriever=retriever, response_synthesizer=get_response_synthesizer() ) # query response = query_engine.query( "Tell me about two celebrities from United States" )