MongoDB Atlas
This notebook covers how to MongoDB Atlas vector search in LangChain, using the langchain-mongodb
package.
MongoDB Atlas is a fully-managed cloud database available in AWS, Azure, and GCP. It supports native Vector Search, full text search (BM25), and hybrid search on your MongoDB document data.
MongoDB Atlas Vector Search allows to store your embeddings in MongoDB documents, create a vector search index, and perform KNN search with an approximate nearest neighbor algorithm (
Hierarchical Navigable Small Worlds
). It uses the $vectorSearch MQL Stage.
Setup
*An Atlas cluster running MongoDB version 6.0.11, 7.0.2, or later (including RCs).
To use MongoDB Atlas, you must first deploy a cluster. We have a Forever-Free tier of cluster on a cloud of your choice available. To get started head over to Atlas here: quick start.
You'll need to install langchain-mongodb
and pymongo
to use this integration.
pip install -qU langchain-mongodb pymongo
Credentials
For this notebook you will need to find your MongoDB cluster URI.
For information on finding your cluster URI read through this guide.
import getpass
MONGODB_ATLAS_CLUSTER_URI = getpass.getpass("MongoDB Atlas Cluster URI:")
If you want to get best in-class automated tracing of your model calls you can also set your LangSmith API key by uncommenting below:
# os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")
# os.environ["LANGSMITH_TRACING"] = "true"
Initialization
- OpenAI
- HuggingFace
- Fake Embedding
pip install -qU langchain-openai
import getpass
os.environ["OPENAI_API_KEY"] = getpass.getpass()
from langchain_openai import OpenAIEmbeddings
embeddings = OpenAIEmbeddings(model="text-embedding-3-large")
pip install -qU langchain-huggingface
from langchain_huggingface import HuggingFaceEmbeddings
embeddings = HuggingFaceEmbeddings(model="sentence-transformers/all-mpnet-base-v2")
pip install -qU langchain-core
from langchain_core.embeddings import FakeEmbeddings
embeddings = FakeEmbeddings(size=4096)
from langchain_mongodb import MongoDBAtlasVectorSearch
from pymongo import MongoClient
# initialize MongoDB python client
client = MongoClient(MONGODB_ATLAS_CLUSTER_URI)
DB_NAME = "langchain_test_db"
COLLECTION_NAME = "langchain_test_vectorstores"
ATLAS_VECTOR_SEARCH_INDEX_NAME = "langchain-test-index-vectorstores"
MONGODB_COLLECTION = client[DB_NAME][COLLECTION_NAME]
vector_store = MongoDBAtlasVectorSearch(
collection=MONGODB_COLLECTION,
embedding=embeddings,
index_name=ATLAS_VECTOR_SEARCH_INDEX_NAME,
relevance_score_fn="cosine",
)
# Create vector search index on the collection
# Since we are using the default OpenAI embedding model (ada-v2) we need to specify the dimensions as 1536
vector_store.create_vector_search_index(dimensions=1536)
[OPTIONAL] Alternative to the vector_store.create_vector_search_index
command above, you can also create the vector search index using the Atlas UI with the following index definition:
{
"fields":[
{
"type": "vector",
"path": "embedding",
"numDimensions": 1536,
"similarity": "cosine"
}
]
}
Manage vector store
Once you have created your vector store, we can interact with it by adding and deleting different items.
Add items to vector store
We can add items to our vector store by using the add_documents
function.
from uuid import uuid4
from langchain_core.documents import Document
document_1 = Document(
page_content="I had chocalate chip pancakes and scrambled eggs for breakfast this morning.",
metadata={"source": "tweet"},
)
document_2 = Document(
page_content="The weather forecast for tomorrow is cloudy and overcast, with a high of 62 degrees.",
metadata={"source": "news"},
)
document_3 = Document(
page_content="Building an exciting new project with LangChain - come check it out!",
metadata={"source": "tweet"},
)
document_4 = Document(
page_content="Robbers broke into the city bank and stole $1 million in cash.",
metadata={"source": "news"},
)
document_5 = Document(
page_content="Wow! That was an amazing movie. I can't wait to see it again.",
metadata={"source": "tweet"},
)
document_6 = Document(
page_content="Is the new iPhone worth the price? Read this review to find out.",
metadata={"source": "website"},
)
document_7 = Document(
page_content="The top 10 soccer players in the world right now.",
metadata={"source": "website"},
)
document_8 = Document(
page_content="LangGraph is the best framework for building stateful, agentic applications!",
metadata={"source": "tweet"},
)
document_9 = Document(
page_content="The stock market is down 500 points today due to fears of a recession.",
metadata={"source": "news"},
)
document_10 = Document(
page_content="I have a bad feeling I am going to get deleted :(",
metadata={"source": "tweet"},
)
documents = [
document_1,
document_2,
document_3,
document_4,
document_5,
document_6,
document_7,
document_8,
document_9,
document_10,
]
uuids = [str(uuid4()) for _ in range(len(documents))]
vector_store.add_documents(documents=documents, ids=uuids)
['03ad81e8-32a0-46f0-b7d8-f5b977a6b52a',
'8396a68d-f4a3-4176-a581-a1a8c303eea4',
'e7d95150-67f6-499f-b611-84367c50fa60',
'8c31b84e-2636-48b6-8b99-9fccb47f7051',
'aa02e8a2-a811-446a-9785-8cea0faba7a9',
'19bd72ff-9766-4c3b-b1fd-195c732c562b',
'642d6f2f-3e34-4efa-a1ed-c4ba4ef0da8d',
'7614bb54-4eb5-4b3b-990c-00e35cb31f99',
'69e18c67-bf1b-43e5-8a6e-64fb3f240e52',
'30d599a7-4a1a-47a9-bbf8-6ed393e2e33c']
Delete items from vector store
vector_store.delete(ids=[uuids[-1]])
True
Query vector store
Once your vector store has been created and the relevant documents have been added you will most likely wish to query it during the running of your chain or agent.
Query directly
Similarity search
Performing a simple similarity search can be done as follows:
results = vector_store.similarity_search(
"LangChain provides abstractions to make working with LLMs easy", k=2
)
for res in results:
print(f"* {res.page_content} [{res.metadata}]")
* Building an exciting new project with LangChain - come check it out! [{'_id': 'e7d95150-67f6-499f-b611-84367c50fa60', 'source': 'tweet'}]
* LangGraph is the best framework for building stateful, agentic applications! [{'_id': '7614bb54-4eb5-4b3b-990c-00e35cb31f99', 'source': 'tweet'}]
Similarity search with score
You can also search with score:
results = vector_store.similarity_search_with_score("Will it be hot tomorrow?", k=1)
for res, score in results:
print(f"* [SIM={score:3f}] {res.page_content} [{res.metadata}]")
* [SIM=0.784560] The weather forecast for tomorrow is cloudy and overcast, with a high of 62 degrees. [{'_id': '8396a68d-f4a3-4176-a581-a1a8c303eea4', 'source': 'news'}]
Pre-filtering with Similarity Search
Atlas Vector Search supports pre-filtering using MQL Operators for filtering. Below is an example index and query on the same data loaded above that allows you do metadata filtering on the "page" field. You can update your existing index with the filter defined and do pre-filtering with vector search.
To enable pre-filtering you need to update the index definition to include a filter field. In this example, we will use the source
field as the filter field.
This can be done programmatically using the MongoDBAtlasVectorSearch.create_vector_search_index
method.
vectorstore.create_vector_search_index(
dimensions=1536,
filters=[{"type":"filter", "path":"source"}],
update=True
)
Alternatively, you can also update the index using the Atlas UI with the following index definition:
{
"fields":[
{
"type": "vector",
"path": "embedding",
"numDimensions": 1536,
"similarity": "cosine"
},
{
"type": "filter",
"path": "source"
}
]
}
And then you can run a query with filter as follows:
results = vector_store.similarity_search(query="foo", k=1, pre_filter={"source": {"$eq": "https://example.com"}})
for doc in results:
print(f"* {doc.page_content} [{doc.metadata}]")
Other search methods
There are a variety of other search methods that are not covered in this notebook, such as MMR search or searching by vector. For a full list of the search abilities available for MongoDBAtlasVectorStore
check out the API reference.
Query by turning into retriever
You can also transform the vector store into a retriever for easier usage in your chains.
Here is how to transform your vector store into a retriever and then invoke the retreiever with a simple query and filter.
retriever = vector_store.as_retriever(
search_type="similarity_score_threshold",
search_kwargs={"k": 1, "score_threshold": 0.2},
)
retriever.invoke("Stealing from the bank is a crime")
[Document(metadata={'_id': '8c31b84e-2636-48b6-8b99-9fccb47f7051', 'source': 'news'}, page_content='Robbers broke into the city bank and stole $1 million in cash.')]
Usage for retrieval-augmented generation
For guides on how to use this vector store for retrieval-augmented generation (RAG), see the following sections:
- Tutorials: working with external knowledge
- How-to: Question and answer with RAG
- Retrieval conceptual docs
Other Notes
- More documentation can be found at MongoDB's LangChain Docs site
- This feature is Generally Available and ready for production deployments.
- The langchain version 0.0.305 (release notes) introduces the support for $vectorSearch MQL stage, which is available with MongoDB Atlas 6.0.11 and 7.0.2. Users utilizing earlier versions of MongoDB Atlas need to pin their LangChain version to <=0.0.304
API reference
For detailed documentation of all MongoDBAtlasVectorSearch
features and configurations head to the API reference: https://python.langchain.com/api_reference/mongodb/index.html
Related
- Vector store conceptual guide
- Vector store how-to guides