MongoDB Atlas

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.

MongoDBAtlasHybridSearchRetriever combines vector search and full-text search using Reciprocal Rank Fusion (RRF) to return the most relevant documents from an Atlas collection. For keyword retrieval without a vector index, MongoDBAtlasFullTextSearchRetriever uses Atlas Search’s Lucene (BM25) analyzer. Both retrievers subclass BaseRetriever and ship in the langchain-mongodb package. You can also cast a MongoDB Atlas vector store to a retriever with .as_retriever(). Use the dedicated retriever classes when you need hybrid or full-text-only search. This guide helps you get started with the MongoDB Atlas retriever. For detailed documentation of all MongoDBAtlasHybridSearchRetriever features and configurations, see the API reference.

Integration details

Setup

You need a running MongoDB Atlas cluster with Atlas Search enabled. Create a cluster in the Atlas UI if you do not already have one. MongoDBAtlasHybridSearchRetriever requires two search indexes on the target collection:

A vector search index for semantic recall
A full-text search index for keyword recall

See Create indexes below, or create the indexes manually in the Atlas UI. If you want automated tracing from individual queries, set your LangSmith API key:

import getpass
import os

os.environ["LANGSMITH_API_KEY"] = getpass.getpass("Enter your LangSmith API key: ")
os.environ["LANGSMITH_TRACING"] = "true"

Installation

Install langchain-mongodb. The following examples also use Voyage AI embeddings:

pip install -qU langchain-mongodb langchain-voyageai

uv add langchain-mongodb langchain-voyageai

Configure

Set your Atlas connection string and open a collection handle:

import getpass
import os

from pymongo import MongoClient

if not os.environ.get("MONGODB_ATLAS_URI"):
    os.environ["MONGODB_ATLAS_URI"] = getpass.getpass("Enter your MongoDB Atlas URI: ")

client = MongoClient(os.environ["MONGODB_ATLAS_URI"])
collection = client["<db_name>"]["<collection_name>"]

Create indexes

Create a vector search index and a full-text search index on the collection before you use hybrid search. Match dimensions to your embeddings model (voyage-4-lite uses 1024):

from langchain_mongodb import MongoDBAtlasVectorSearch
from langchain_mongodb.index import create_fulltext_search_index
from langchain_voyageai import VoyageAIEmbeddings  # swap in any embeddings class

embeddings = VoyageAIEmbeddings(model="voyage-4-lite")

vector_store = MongoDBAtlasVectorSearch(
    collection=collection,
    embedding=embeddings,
    index_name="vector_index",
)

# Create the vector search index (runs asynchronously on Atlas)
vector_store.create_vector_search_index(
    dimensions=1024,
    wait_until_complete=60,
)

# Create the full-text search index on the text field
create_fulltext_search_index(
    collection=collection,
    field="text",
    index_name="search_index",
    wait_until_complete=60,
)

For more index options, see the MongoDB Atlas vector store guide.

Instantiation

Hybrid search

MongoDBAtlasHybridSearchRetriever fuses vector and keyword results with RRF. Increasing vector_penalty or fulltext_penalty reduces the weight of that channel:

from langchain_mongodb.retrievers import MongoDBAtlasHybridSearchRetriever

retriever = MongoDBAtlasHybridSearchRetriever(
    vectorstore=vector_store,
    search_index_name="search_index",
    k=5,
    vector_penalty=60.0,
    fulltext_penalty=60.0,
)

Full-text search

MongoDBAtlasFullTextSearchRetriever runs BM25 keyword search and does not require a vector index:

from langchain_mongodb.retrievers import MongoDBAtlasFullTextSearchRetriever

ft_retriever = MongoDBAtlasFullTextSearchRetriever(
    collection=collection,
    search_index_name="search_index",
    search_field="text",
    k=5,
)

Usage

docs = retriever.invoke("What is MongoDB Atlas?")

Use within a chain

from langchain_anthropic import ChatAnthropic
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnablePassthrough

prompt = ChatPromptTemplate.from_template(
    "Answer the question based only on the following context:\n\n{context}\n\nQuestion: {question}"
)

llm = ChatAnthropic(model="claude-sonnet-4-6")

chain = (
    {"context": retriever, "question": RunnablePassthrough()}
    | prompt
    | llm
    | StrOutputParser()
)

chain.invoke("What is MongoDB Atlas?")

API reference

For detailed documentation of all MongoDBAtlasHybridSearchRetriever features and configurations, see the API reference. For full-text search, see MongoDBAtlasFullTextSearchRetriever.

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.

Edit this page on GitHub or file an issue.

Popular Providers

Integrations by component

Integration details

Setup

Installation

Configure

Create indexes

Instantiation

Hybrid search

Full-text search

Usage

Use within a chain

API reference

​Integration details

​Setup

​Installation

​Configure

​Create indexes

​Instantiation

​Hybrid search

​Full-text search

​Usage

​Use within a chain

​API reference

Integration details

Setup

Installation

Configure

Create indexes

Instantiation

Hybrid search

Full-text search

Usage

Use within a chain

API reference