Azure Cosmos DB mongo vcore 集成

使用 LangChain Python 集成 Azure Cosmos DB mongo vcore 向量存储。

This notebook shows you how to leverage this integrated vector database to store documents in collections, create indicies and perform vector search queries using approximate nearest neighbor algorithms such as COS (cosine distance), L2 (Euclidean distance), and IP (inner product) to locate documents close to the query vectors.

Azure Cosmos DB is the database that powers OpenAI’s ChatGPT service. It offers single-digit millisecond response times, automatic and instant scalability, along with guaranteed speed at any scale.

Azure Cosmos DB for MongoDB vCore provides developers with a fully managed MongoDB-compatible database service for building modern applications with a familiar architecture. You can apply your MongoDB experience and continue to use your favorite MongoDB drivers, SDKs, and tools by pointing your application to the API for MongoDB vCore account’s connection string.

Sign Up for lifetime free access to get started today.

pip install -qU pymongo langchain-openai langchain-azure-ai

import os

CONNECTION_STRING = "YOUR_CONNECTION_STRING"
INDEX_NAME = "langchain-test-index"
NAMESPACE = "langchain_test_db.langchain_test_collection"
DB_NAME, COLLECTION_NAME = NAMESPACE.split(".")

We want to use AzureOpenAIEmbeddings so we need to set up our Azure OpenAI API Key alongside other environment variables.

# Set up the OpenAI environment variables

os.environ["AZURE_OPENAI_API_KEY"] = "YOUR_AZURE_OPENAI_API_KEY"
os.environ["AZURE_OPENAI_ENDPOINT"] = "YOUR_AZURE_OPENAI_ENDPOINT"
os.environ["AZURE_OPENAI_API_VERSION"] = "2023-05-15"
os.environ["OPENAI_EMBEDDINGS_MODEL_NAME"] = "text-embedding-ada-002"
os.environ["OPENAI_EMBEDDINGS_DEPLOYMENT"] = "text-embedding-ada-002"

Now, we need to load the documents into the collection, create the index and then run our queries against the index to retrieve matches.

Please refer to the documentation if you have questions about certain parameters

from langchain_community.document_loaders import TextLoader
from langchain_azure_ai.vectorstores.azure_cosmos_db_mongo_vcore import (
    AzureCosmosDBMongoVCoreVectorSearch,
    CosmosDBSimilarityType,
    CosmosDBVectorSearchType,
)
from langchain_openai import AzureOpenAIEmbeddings
from langchain_text_splitters import CharacterTextSplitter

SOURCE_FILE_NAME = "../../how_to/state_of_the_union.txt"

loader = TextLoader(SOURCE_FILE_NAME)
documents = loader.load()
text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
docs = text_splitter.split_documents(documents)

# OpenAI Settings
model_deployment = os.getenv(
    "OPENAI_EMBEDDINGS_DEPLOYMENT", "smart-agent-embedding-ada"
)
model_name = os.getenv("OPENAI_EMBEDDINGS_MODEL_NAME", "text-embedding-ada-002")


openai_embeddings: AzureOpenAIEmbeddings = AzureOpenAIEmbeddings(
    model=model_name, chunk_size=1
)

docs[0]

Document(metadata={'source': '../../how_to/state_of_the_union.txt'}, page_content='Madam Speaker, Madam Vice President, our First Lady and Second Gentleman. Members of Congress and the Cabinet. Justices of the Supreme Court. My fellow Americans.  \n\nLast year COVID-19 kept us apart. This year we are finally together again. \n\nTonight, we meet as Democrats Republicans and Independents. But most importantly as Americans. \n\nWith a duty to one another to the American people to the Constitution. \n\nAnd with an unwavering resolve that freedom will always triumph over tyranny. \n\nSix days ago, Russia’s Vladimir Putin sought to shake the foundations of the free world thinking he could make it bend to his menacing ways. But he badly miscalculated. \n\nHe thought he could roll into Ukraine and the world would roll over. Instead he met a wall of strength he never imagined. \n\nHe met the Ukrainian people. \n\nFrom President Zelenskyy to every Ukrainian, their fearlessness, their courage, their determination, inspires the world.')

from pymongo import MongoClient

# INDEX_NAME = "izzy-test-index-2"
# NAMESPACE = "izzy_test_db.izzy_test_collection"
# DB_NAME, COLLECTION_NAME = NAMESPACE.split(".")

client: MongoClient = MongoClient(CONNECTION_STRING)
collection = client[DB_NAME][COLLECTION_NAME]

model_deployment = os.getenv(
    "OPENAI_EMBEDDINGS_DEPLOYMENT", "smart-agent-embedding-ada"
)
model_name = os.getenv("OPENAI_EMBEDDINGS_MODEL_NAME", "text-embedding-ada-002")

vectorstore = AzureCosmosDBMongoVCoreVectorSearch.from_documents(
    docs,
    openai_embeddings,
    collection=collection,
    index_name=INDEX_NAME,
)

# Read more about these variables in detail here. https://learn.microsoft.com/en-us/azure/cosmos-db/mongodb/vcore/vector-search
num_lists = 100
dimensions = 1536
similarity_algorithm = CosmosDBSimilarityType.COS
kind = CosmosDBVectorSearchType.VECTOR_IVF
m = 16
ef_construction = 64
ef_search = 40
score_threshold = 0.1

vectorstore.create_index(
    num_lists, dimensions, similarity_algorithm, kind, m, ef_construction
)

"""
# DiskANN vectorstore
maxDegree = 40
dimensions = 1536
similarity_algorithm = CosmosDBSimilarityType.COS
kind = CosmosDBVectorSearchType.VECTOR_DISKANN
lBuild = 20

vectorstore.create_index(
            dimensions=dimensions,
            similarity=similarity_algorithm,
            kind=kind ,
            max_degree=maxDegree,
            l_build=lBuild,
        )

# -----------------------------------------------------------

# HNSW vectorstore
dimensions = 1536
similarity_algorithm = CosmosDBSimilarityType.COS
kind = CosmosDBVectorSearchType.VECTOR_HNSW
m = 16
ef_construction = 64

vectorstore.create_index(
            dimensions=dimensions,
            similarity=similarity_algorithm,
            kind=kind ,
            m=m,
            ef_construction=ef_construction,
        )
"""

'\n# DiskANN vectorstore\nmaxDegree = 40\ndimensions = 1536\nsimilarity_algorithm = CosmosDBSimilarityType.COS\nkind = CosmosDBVectorSearchType.VECTOR_DISKANN\nlBuild = 20\n\nvectorstore.create_index(\n            dimensions=dimensions,\n            similarity=similarity_algorithm,\n            kind=kind ,\n            max_degree=maxDegree,\n            l_build=lBuild,\n        )\n\n# -----------------------------------------------------------\n\n# HNSW vectorstore\ndimensions = 1536\nsimilarity_algorithm = CosmosDBSimilarityType.COS\nkind = CosmosDBVectorSearchType.VECTOR_HNSW\nm = 16\nef_construction = 64\n\nvectorstore.create_index(\n            dimensions=dimensions,\n            similarity=similarity_algorithm,\n            kind=kind ,\n            m=m,\n            ef_construction=ef_construction,\n        )\n'

# perform a similarity search between the embedding of the query and the embeddings of the documents
query = "What did the president say about Ketanji Brown Jackson"
docs = vectorstore.similarity_search(query)

print(docs[0].page_content)

Tonight. I call on the Senate to: Pass the Freedom to Vote Act. Pass the John Lewis Voting Rights Act. And while you’re at it, pass the Disclose Act so Americans can know who is funding our elections.

Tonight, I’d like to honor someone who has dedicated his life to serve this country: Justice Stephen Breyer—an Army veteran, Constitutional scholar, and retiring Justice of the United States Supreme Court. Justice Breyer, thank you for your service.

One of the most serious constitutional responsibilities a President has is nominating someone to serve on the United States Supreme Court.

And I did that 4 days ago, when I nominated Circuit Court of Appeals Judge Ketanji Brown Jackson. One of our nation’s top legal minds, who will continue Justice Breyer’s legacy of excellence.

Once the documents have been loaded and the index has been created, you can now instantiate the vector store directly and run queries against the index

vectorstore = AzureCosmosDBMongoVCoreVectorSearch.from_connection_string(
    CONNECTION_STRING, NAMESPACE, openai_embeddings, index_name=INDEX_NAME
)

# perform a similarity search between a query and the ingested documents
query = "What did the president say about Ketanji Brown Jackson"
docs = vectorstore.similarity_search(query)

print(docs[0].page_content)

Tonight. I call on the Senate to: Pass the Freedom to Vote Act. Pass the John Lewis Voting Rights Act. And while you’re at it, pass the Disclose Act so Americans can know who is funding our elections.

Tonight, I’d like to honor someone who has dedicated his life to serve this country: Justice Stephen Breyer—an Army veteran, Constitutional scholar, and retiring Justice of the United States Supreme Court. Justice Breyer, thank you for your service.

One of the most serious constitutional responsibilities a President has is nominating someone to serve on the United States Supreme Court.

And I did that 4 days ago, when I nominated Circuit Court of Appeals Judge Ketanji Brown Jackson. One of our nation’s top legal minds, who will continue Justice Breyer’s legacy of excellence.

vectorstore = AzureCosmosDBMongoVCoreVectorSearch(
    collection, openai_embeddings, index_name=INDEX_NAME
)

# perform a similarity search between a query and the ingested documents
query = "What did the president say about Ketanji Brown Jackson"
docs = vectorstore.similarity_search(query)

print(docs[0].page_content)

Tonight. I call on the Senate to: Pass the Freedom to Vote Act. Pass the John Lewis Voting Rights Act. And while you’re at it, pass the Disclose Act so Americans can know who is funding our elections.

Tonight, I’d like to honor someone who has dedicated his life to serve this country: Justice Stephen Breyer—an Army veteran, Constitutional scholar, and retiring Justice of the United States Supreme Court. Justice Breyer, thank you for your service.

One of the most serious constitutional responsibilities a President has is nominating someone to serve on the United States Supreme Court.

And I did that 4 days ago, when I nominated Circuit Court of Appeals Judge Ketanji Brown Jackson. One of our nation’s top legal minds, who will continue Justice Breyer’s legacy of excellence.

Filtered vector search (Preview)

Azure Cosmos DB for MongoDB supports pre-filtering with $lt, $lte, $eq, $neq, $gte, $gt, $in, $nin, and $regex. To use this feature, enable "filtering vector search" in the “Preview Features” tab of your Azure Subscription. Learn more about filtered vector search preview features.

# Create a filter index
vectorstore.create_filter_index(
    property_to_filter="metadata.source", index_name="filter_index"
)

{'raw': {'defaultShard': {'numIndexesBefore': 3,
   'numIndexesAfter': 4,
   'createdCollectionAutomatically': False,
   'ok': 1}},
 'ok': 1}

query = "What did the president say about Ketanji Brown Jackson"
docs = vectorstore.similarity_search(
    query, pre_filter={"metadata.source": {"$ne": "filter content"}}
)

len(docs)

4

docs = vectorstore.similarity_search(
    query,
    pre_filter={"metadata.source": {"$ne": "../../how_to/state_of_the_union.txt"}},
)

len(docs)

0

---

连接这些文档到 Claude、VSCode 等工具，通过 MCP 获取实时答案。

在 GitHub 上编辑此页面或提交 issue。