如果你在 Colab 中打开这个 Notebook，可能需要先安装 LlamaIndex 🦙。

%pip install llama-index-vector-stores-weaviate

!pip install llama-index

创建 Weaviate 客户端

import os
import openai

os.environ["OPENAI_API_KEY"] = ""
openai.api_key = os.environ["OPENAI_API_KEY"]

import logging
import sys

logging.basicConfig(stream=sys.stdout, level=logging.INFO)
logging.getLogger().addHandler(logging.StreamHandler(stream=sys.stdout))

import weaviate

# cloud
cluster_url = ""
api_key = ""

client = weaviate.connect_to_wcs(
    cluster_url=cluster_url,
    auth_credentials=weaviate.auth.AuthApiKey(api_key),
)

# local
# client = connect_to_local()

加载文档，构建 VectorStoreIndex

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.vector_stores.weaviate import WeaviateVectorStore
from IPython.display import Markdown, display

下载数据

!mkdir -p 'data/paul_graham/'
!wget 'https://raw.githubusercontent.com/run-llama/llama_index/main/docs/examples/data/paul_graham/paul_graham_essay.txt' -O 'data/paul_graham/paul_graham_essay.txt'

# load documents
documents = SimpleDirectoryReader("./data/paul_graham").load_data()

from llama_index.core import StorageContext

# If you want to load the index later, be sure to give it a name!
vector_store = WeaviateVectorStore(
    weaviate_client=client, index_name="LlamaIndex"
)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)

# NOTE: you may also choose to define a index_name manually.
# index_name = "test_prefix"
# vector_store = WeaviateVectorStore(weaviate_client=client, index_name=index_name)

使用自定义批处理配置

LlamaIndex 默认使用 Weaviate 的动态批处理，并针对大多数常见场景进行了优化。但在低延迟环境中，这可能会使服务器过载或达到现有 GRPC 消息大小上限。为了获得更多控制和更好的摄取流程，可以考虑使用固定大小的批次来调整批大小。

下面介绍如何微调 WeaviateVectorStore 并定义自定义批次：

from weaviate.classes.config import ConsistencyLevel

custom_batch = client.batch.fixed_size(
    batch_size=123,
    concurrent_requests=3,
    consistency_level=ConsistencyLevel.ALL,
)
vector_store_fixed = WeaviateVectorStore(
    weaviate_client=client,
    index_name="LlamaIndex",
    # we pass our custom batch as a client_kwargs
    client_kwargs={"custom_batch": custom_batch},
)

查询索引

# set Logging to DEBUG for more detailed outputs
query_engine = index.as_query_engine()
response = query_engine.query("What did the author do growing up?")

display(Markdown(f"<b>{response}</b>"))

加载索引

这里，我们使用与创建初始索引时相同的索引名称。这样可以避免自动生成索引名，并让我们轻松地重新连接到它。

cluster_url = ""
api_key = ""

client = weaviate.connect_to_wcs(
    cluster_url=cluster_url,
    auth_credentials=weaviate.auth.AuthApiKey(api_key),
)

# local
# client = weaviate.connect_to_local()

vector_store = WeaviateVectorStore(
    weaviate_client=client, index_name="LlamaIndex"
)

loaded_index = VectorStoreIndex.from_vector_store(vector_store)

# set Logging to DEBUG for more detailed outputs
query_engine = loaded_index.as_query_engine()
response = query_engine.query("What happened at interleaf?")
display(Markdown(f"<b>{response}</b>"))

元数据过滤

让我们插入一个虚拟文档，并尝试通过过滤只返回该文档。

from llama_index.core import Document

doc = Document.example()
print(doc.metadata)
print("-----")
print(doc.text[:100])

loaded_index.insert(doc)

from llama_index.core.vector_stores import ExactMatchFilter, MetadataFilters

filters = MetadataFilters(
    filters=[ExactMatchFilter(key="filename", value="README.md")]
)
query_engine = loaded_index.as_query_engine(filters=filters)
response = query_engine.query("What is the name of the file?")
display(Markdown(f"<b>{response}</b>"))

完全删除索引

你可以使用 delete_index 函数删除由向量存储创建的索引

vector_store.delete_index()

vector_store.delete_index()  # calling the function again does nothing

连接终止

你必须确保客户端连接已关闭：

client.close()