Skip to main content

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 and full text search (BM25) 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 clusters available. To get started head over to Atlas here: quick start.

You'll need to install langchain-mongodb and pymongo to use this integration, and we will also download langchain-openai for this notebook in order to use it for our embeddings.

pip install -qU langchain-mongodb pymongo langchain-openai

[notice] A new release of pip is available: 24.0 -> 24.1.2
[notice] To update, run: pip install --upgrade pip
Note: you may need to restart the kernel to use updated packages.

Credentials​

For this notebook you will need to set an OPENAI_API_KEY and also find your MongoDB cluster URI.

For information on finding your cluster URI read through this guide.

import getpass
import os

MONGODB_ATLAS_CLUSTER_URI = getpass.getpass("MongoDB Atlas Cluster URI:")
if 'OPENAI_API_KEY' not in os.environ:
    os.environ['OPENAI_API_KEY'] = getpass.getpass("OpenAI API Key:")

Instantiation​

from pymongo import MongoClient
from langchain_openai import OpenAIEmbeddings
from langchain_mongodb.vectorstores 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]
embedding_function = OpenAIEmbeddings()

vector_store = MongoDBAtlasVectorSearch(
  collection=MONGODB_COLLECTION,
  embedding=embedding_function,
  index_name=ATLAS_VECTOR_SEARCH_INDEX_NAME,
  relevance_score_fn="cosine",
)

Manage vector store​

Add items to vector store​

from langchain_core.documents import Document

document_1 = Document(
    page_content="foo",
    metadata={"source": "https://example.com"}
)

document_2 = Document(
    page_content="bar",
    metadata={"source": "https://example.com"}
)

document_3 = Document(
    page_content="baz",
    metadata={"source": "https://example.com"}
)

documents = [document_1, document_2, document_3]

vector_store.add_documents(documents=documents,ids=["1","2","3"])
API Reference:Document
['1', '2', '3']

Delete items from vector store​

vector_store.delete(ids=["3"])
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​

Performing a simple similarity search can be done as follows:

results = vector_store.similarity_search(query="foo",k=1)
for doc in results:
    print(f"* {doc.page_content} [{doc.metadata}]")
* foo [{'_id': '1', 'source': 'https://example.com'}]

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.

{
  "fields":[
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1536,
      "similarity": "cosine"
    },
    {
      "type": "filter",
      "path": "source"
    }
  ]
}

You can also update the index programmatically using the MongoDBAtlasVectorSearch.create_index method.

vectorstore.create_index(
  dimensions=1536,
  filters=[{"type":"filter", "path":"source"}],
  update=True
)

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 Notes

  • More documentation can be found at LangChain-MongoDB 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://api.python.langchain.com/en/latest/mongodb_api_reference.html

Was this page helpful?


You can also leave detailed feedback on GitHub.