Upstash 向量存储¶

我们将探讨如何使用 LlamaIndex 与 Upstash Vector 进行交互！

In [ ]:

! pip install -q llama-index upstash-vector

! pip install -q llama-index upstash-vector

In [ ]:

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader
from llama_index.core.vector_stores import UpstashVectorStore
from llama_index.core import StorageContext
import textwrap
import openai

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader from llama_index.core.vector_stores import UpstashVectorStore from llama_index.core import StorageContext import textwrap import openai

In [ ]:

# Setup the OpenAI API
openai.api_key = "sk-..."

# Setup the OpenAI API openai.api_key = "sk-..."

In [ ]:

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

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

--2024-02-03 20:04:25--  https://raw.githubusercontent.com/run-llama/llama_index/main/docs/docs/examples/data/paul_graham/paul_graham_essay.txt
Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...
Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 75042 (73K) [text/plain]
Saving to: ‘data/paul_graham/paul_graham_essay.txt’

data/paul_graham/pa 100%[===================>]  73.28K  --.-KB/s    in 0.01s   

2024-02-03 20:04:25 (5.96 MB/s) - ‘data/paul_graham/paul_graham_essay.txt’ saved [75042/75042]

现在，我们可以使用 LlamaIndex 的 SimpleDirectoryReader 加载文档

In [ ]:

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

print("# Documents:", len(documents))

documents = SimpleDirectoryReader("./data/paul_graham/").load_data() print("# Documents:", len(documents))

# Documents: 1

要在 Upstash 上创建索引，请访问 https://console.upstash.com/vector，创建一个维度为 1536 且距离度量方式为 Cosine 的索引。复制下方的 URL 和令牌

In [ ]:

vector_store = UpstashVectorStore(url="https://...", token="...")

storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents, storage_context=storage_context
)

vector_store = UpstashVectorStore(url="https://...", token="...") storage_context = StorageContext.from_defaults(vector_store=vector_store) index = VectorStoreIndex.from_documents( documents, storage_context=storage_context )

现在我们已成功创建索引，并将论文中的向量数据填充其中！数据需要短暂时间建立索引，之后即可进行查询。

In [ ]:

query_engine = index.as_query_engine()
res1 = query_engine.query("What did the author learn?")
print(textwrap.fill(str(res1), 100))

print("\n")

res2 = query_engine.query("What is the author's opinion on startups?")
print(textwrap.fill(str(res2), 100))

query_engine = index.as_query_engine() res1 = query_engine.query("What did the author learn?") print(textwrap.fill(str(res1), 100)) print("\n") res2 = query_engine.query("What is the author's opinion on startups?") print(textwrap.fill(str(res2), 100))

The author learned that the study of philosophy in college did not live up to their expectations.
They found that other fields took up most of the space of ideas, leaving little room for what they
perceived as the ultimate truths that philosophy was supposed to explore. As a result, they decided
to switch to studying AI.


The author's opinion on startups is that they are in need of help and support, especially in the
beginning stages. The author believes that founders of startups are often helpless and face various
challenges, such as getting incorporated and understanding the intricacies of running a company. The
author's investment firm, Y Combinator, aims to provide seed funding and comprehensive support to
startups, offering them the guidance and resources they need to succeed.

元数据过滤¶

您可以在 VectorStoreQuery 中传递 MetadataFilters 参数，用于筛选从 Upstash 向量存储返回的节点。

In [ ]:

import os

from llama_index.vector_stores.upstash import UpstashVectorStore
from llama_index.core.vector_stores.types import (
    MetadataFilter,
    MetadataFilters,
    FilterOperator,
)

vector_store = UpstashVectorStore(
    url=os.environ.get("UPSTASH_VECTOR_URL") or "",
    token=os.environ.get("UPSTASH_VECTOR_TOKEN") or "",
)

index = VectorStoreIndex.from_vector_store(vector_store=vector_store)

filters = MetadataFilters(
    filters=[
        MetadataFilter(
            key="author", value="Marie Curie", operator=FilterOperator.EQ
        )
    ],
)

retriever = index.as_retriever(filters=filters)

retriever.retrieve("What is inception about?")

import os from llama_index.vector_stores.upstash import UpstashVectorStore from llama_index.core.vector_stores.types import ( MetadataFilter, MetadataFilters, FilterOperator, ) vector_store = UpstashVectorStore( url=os.environ.get("UPSTASH_VECTOR_URL") or "", token=os.environ.get("UPSTASH_VECTOR_TOKEN") or "", ) index = VectorStoreIndex.from_vector_store(vector_store=vector_store) filters = MetadataFilters( filters=[ MetadataFilter( key="author", value="Marie Curie", operator=FilterOperator.EQ ) ], ) retriever = index.as_retriever(filters=filters) retriever.retrieve("What is inception about?")

我们也可以将多个 MetadataFilters 通过 AND 或 OR 条件组合使用

In [ ]:

from llama_index.core.vector_stores import FilterOperator, FilterCondition

filters = MetadataFilters(
    filters=[
        MetadataFilter(
            key="theme",
            value=["Fiction", "Horror"],
            operator=FilterOperator.IN,
        ),
        MetadataFilter(key="year", value=1997, operator=FilterOperator.GT),
    ],
    condition=FilterCondition.AND,
)

retriever = index.as_retriever(filters=filters)
retriever.retrieve("Harry Potter?")

from llama_index.core.vector_stores import FilterOperator, FilterCondition filters = MetadataFilters( filters=[ MetadataFilter( key="theme", value=["Fiction", "Horror"], operator=FilterOperator.IN, ), MetadataFilter(key="year", value=1997, operator=FilterOperator.GT), ], condition=FilterCondition.AND, ) retriever = index.as_retriever(filters=filters) retriever.retrieve("Harry Potter?")