MongoDB for AI: A Complete Guide to Vector Search and RAG Implementation

Will you stand by as the future unfolds, or will you seize the opportunity to create it?

Be part of Gen AI Launch Pad 2025 and take control.

Introduction

In the age of artificial intelligence (AI) and large-scale data applications, having a flexible, high-performance database is essential. MongoDB, a NoSQL database, provides an ideal solution with its document-oriented storage, powerful querying capabilities, and seamless integration with AI frameworks. In this guide, we will walk through how MongoDB can be leveraged for AI applications, particularly for Retrieval-Augmented Generation (RAG). By the end of this post, you'll learn how to:

Set up MongoDB and establish a connection
Store and retrieve structured and unstructured data
Perform CRUD (Create, Read, Update, Delete) operations
Use MongoDB's vector search for AI-driven applications
Implement RAG with MongoDB and OpenAI embeddings

Setting Up MongoDB

Before diving into AI-powered applications, let’s install and set up MongoDB.

Installing Required Packages

To interact with MongoDB, we need to install the required Python libraries:

!pip install pymongo langchain tiktoken faiss-cpu pypdf

These libraries help us connect to MongoDB, process text data, generate embeddings, and perform similarity searches.

Setting Up API Keys

To access MongoDB Atlas and OpenAI APIs, we configure the API keys securely:

import os
from google.colab import userdata

os.environ["OPENAI_API_KEY"] = userdata.get('OPENAI_API_KEY')
MONGODB_URI = userdata.get('MONGODB_URI')

This ensures that sensitive credentials are not hardcoded into the script.

Connecting to MongoDB

Once the environment is set up, we establish a connection with MongoDB Atlas.

from pymongo import MongoClient

connection_string = MONGODB_URI
client = MongoClient(connection_string)
client.admin.command('ping')
print("✅ Connected to MongoDB Atlas successfully!")

Expected Output:

✅ Connected to MongoDB Atlas successfully!

This confirms that we have successfully connected to our database.

Working with Databases and Collections

MongoDB stores data in databases and collections. Let's create a database and a collection.

db = client["my_database"]
collection = db["users"]

Now, we can insert and retrieve data.

CRUD Operations in MongoDB

Insert a Single Document

user = {"name": "John Doe", "age": 30, "city": "New York"}
collection.insert_one(user)
print("✅ One document inserted!")

Expected Output:

✅ One document inserted!

Insert Multiple Documents

users = [
    {"name": "Alice", "age": 25, "city": "Los Angeles"},
    {"name": "Bob", "age": 28, "city": "Chicago"},
]
collection.insert_many(users)
print("✅ Multiple documents inserted!")

Expected Output:

✅ Multiple documents inserted!

Retrieve Data

Find a single document:

user = collection.find_one({"name": "John Doe"})
print("👤 Found user:", user)

Expected Output:

👤 Found user: {'_id': ObjectId('xxx'), 'name': 'John Doe', 'age': 30, 'city': 'New York'}

Find all documents:

for user in collection.find():
    print(user)

Find documents with a condition:

for user in collection.find({"city": "New York"}):
    print(user)

Update a Document

collection.update_one({"name": "John Doe"}, {"$set": {"age": 31}})
print("✅ User updated!")

Delete a Document

collection.delete_one({"name": "Alice"})
print("✅ One document deleted!")

Drop a Collection

collection.drop()
print("🗑️ Collection deleted!")

Implementing Retrieval-Augmented Generation (RAG) with MongoDB

Retrieval-Augmented Generation (RAG) enhances AI models by retrieving relevant data before generating a response. MongoDB’s vector search capabilities make this possible.

Generate and Store Embeddings

from langchain.embeddings.openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings()

def get_embedding(text: str):
    return embeddings.embed_query(text)

Load and Split Text Data

from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

pdf_url = "https://investors.mongodb.com/node/12236/pdf"
loader = PyPDFLoader(pdf_url)
data = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=400, chunk_overlap=20)
documents = text_splitter.split_documents(data)

Expected Output:

Loaded and split 88 documents from PDF.

Store Embeddings in MongoDB

docs_to_insert = []
for doc in documents:
    embedding = get_embedding(doc.page_content)
    docs_to_insert.append({"text": doc.page_content, "embedding": embedding})
collection.insert_many(docs_to_insert)
print(f"Inserted {len(docs_to_insert)} documents into MongoDB")

Performing Vector Search in MongoDB

Create Vector Search Index

from pymongo.operations import SearchIndexModel

index_name = "vector_index"
search_index_model = SearchIndexModel(
    definition={"fields": [{"type": "vector", "numDimensions": 1536, "path": "embedding", "similarity": "cosine"}]},
    name=index_name,
    type="vectorSearch"
)
collection.create_search_index(model=search_index_model)
print(f"Created vector search index '{index_name}' on MongoDB collection.")

Run Vector Search Query

def get_query_results(query: str, limit: int = 5):
    query_embedding = get_embedding(query)
    pipeline = [{"$vectorSearch": {"index": index_name, "queryVector": query_embedding, "path": "embedding", "numCandidates": 10, "limit": limit}},
        {"$project": {"_id": 0, "text": 1}}]
    return list(collection.aggregate(pipeline))

Conclusion

In this guide, we explored MongoDB's capabilities in AI applications, from basic CRUD operations to advanced vector search for RAG. MongoDB's flexibility, scalability, and AI-friendly integrations make it an excellent choice for modern AI-driven systems.

Resources

---------------------------

Stay Updated:- Follow Build Fast with AI pages for all the latest AI updates and resources.

Experts predict 2025 will be the defining year for Gen AI Implementation. Want to be ahead of the curve?

Join Build Fast with AI’s Gen AI Launch Pad 2025 - your accelerated path to mastering AI tools and building revolutionary applications.

---------------------------

Resources and Community

Join our community of 12,000+ AI enthusiasts and learn to build powerful AI applications! Whether you're a beginner or an experienced developer, this tutorial will help you understand and implement AI agents in your projects.

Website: www.buildfastwithai.com
LinkedIn: linkedin.com/company/build-fast-with-ai/
Instagram: instagram.com/buildfastwithai/
Twitter: x.com/satvikps
Telegram: t.me/BuildFastWithAI

Will you stand by as the future unfolds, or will you seize the opportunity to create it?

Be part of Gen AI Launch Pad 2025 and take control.

Introduction

Set up MongoDB and establish a connection
Store and retrieve structured and unstructured data
Perform CRUD (Create, Read, Update, Delete) operations
Use MongoDB's vector search for AI-driven applications
Implement RAG with MongoDB and OpenAI embeddings

Setting Up MongoDB

Before diving into AI-powered applications, let’s install and set up MongoDB.

Installing Required Packages

To interact with MongoDB, we need to install the required Python libraries:

!pip install pymongo langchain tiktoken faiss-cpu pypdf

These libraries help us connect to MongoDB, process text data, generate embeddings, and perform similarity searches.

Setting Up API Keys

To access MongoDB Atlas and OpenAI APIs, we configure the API keys securely:

import os
from google.colab import userdata

os.environ["OPENAI_API_KEY"] = userdata.get('OPENAI_API_KEY')
MONGODB_URI = userdata.get('MONGODB_URI')

This ensures that sensitive credentials are not hardcoded into the script.

Connecting to MongoDB

Once the environment is set up, we establish a connection with MongoDB Atlas.

from pymongo import MongoClient

connection_string = MONGODB_URI
client = MongoClient(connection_string)
client.admin.command('ping')
print("✅ Connected to MongoDB Atlas successfully!")

Expected Output:

✅ Connected to MongoDB Atlas successfully!

This confirms that we have successfully connected to our database.

Working with Databases and Collections

MongoDB stores data in databases and collections. Let's create a database and a collection.

db = client["my_database"]
collection = db["users"]

Now, we can insert and retrieve data.

CRUD Operations in MongoDB

Insert a Single Document

user = {"name": "John Doe", "age": 30, "city": "New York"}
collection.insert_one(user)
print("✅ One document inserted!")

Expected Output:

✅ One document inserted!

Insert Multiple Documents

users = [
    {"name": "Alice", "age": 25, "city": "Los Angeles"},
    {"name": "Bob", "age": 28, "city": "Chicago"},
]
collection.insert_many(users)
print("✅ Multiple documents inserted!")

Expected Output:

✅ Multiple documents inserted!

Retrieve Data

Find a single document:

user = collection.find_one({"name": "John Doe"})
print("👤 Found user:", user)

Expected Output:

👤 Found user: {'_id': ObjectId('xxx'), 'name': 'John Doe', 'age': 30, 'city': 'New York'}

Find all documents:

for user in collection.find():
    print(user)

Find documents with a condition:

for user in collection.find({"city": "New York"}):
    print(user)

Update a Document

collection.update_one({"name": "John Doe"}, {"$set": {"age": 31}})
print("✅ User updated!")

Delete a Document

collection.delete_one({"name": "Alice"})
print("✅ One document deleted!")

Drop a Collection

collection.drop()
print("🗑️ Collection deleted!")

Implementing Retrieval-Augmented Generation (RAG) with MongoDB

Retrieval-Augmented Generation (RAG) enhances AI models by retrieving relevant data before generating a response. MongoDB’s vector search capabilities make this possible.

Generate and Store Embeddings

from langchain.embeddings.openai import OpenAIEmbeddings

embeddings = OpenAIEmbeddings()

def get_embedding(text: str):
    return embeddings.embed_query(text)

Load and Split Text Data

from langchain.document_loaders import PyPDFLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

pdf_url = "https://investors.mongodb.com/node/12236/pdf"
loader = PyPDFLoader(pdf_url)
data = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=400, chunk_overlap=20)
documents = text_splitter.split_documents(data)

Expected Output:

Loaded and split 88 documents from PDF.

Store Embeddings in MongoDB

docs_to_insert = []
for doc in documents:
    embedding = get_embedding(doc.page_content)
    docs_to_insert.append({"text": doc.page_content, "embedding": embedding})
collection.insert_many(docs_to_insert)
print(f"Inserted {len(docs_to_insert)} documents into MongoDB")

Performing Vector Search in MongoDB

Create Vector Search Index

from pymongo.operations import SearchIndexModel

index_name = "vector_index"
search_index_model = SearchIndexModel(
    definition={"fields": [{"type": "vector", "numDimensions": 1536, "path": "embedding", "similarity": "cosine"}]},
    name=index_name,
    type="vectorSearch"
)
collection.create_search_index(model=search_index_model)
print(f"Created vector search index '{index_name}' on MongoDB collection.")

Run Vector Search Query

def get_query_results(query: str, limit: int = 5):
    query_embedding = get_embedding(query)
    pipeline = [{"$vectorSearch": {"index": index_name, "queryVector": query_embedding, "path": "embedding", "numCandidates": 10, "limit": limit}},
        {"$project": {"_id": 0, "text": 1}}]
    return list(collection.aggregate(pipeline))

Conclusion

Resources

---------------------------

Stay Updated:- Follow Build Fast with AI pages for all the latest AI updates and resources.

Experts predict 2025 will be the defining year for Gen AI Implementation. Want to be ahead of the curve?

Join Build Fast with AI’s Gen AI Launch Pad 2025 - your accelerated path to mastering AI tools and building revolutionary applications.

---------------------------

Resources and Community

Website: www.buildfastwithai.com
LinkedIn: linkedin.com/company/build-fast-with-ai/
Instagram: instagram.com/buildfastwithai/
Twitter: x.com/satvikps
Telegram: t.me/BuildFastWithAI

BuildFast Bot

Introduction

Setting Up MongoDB

Installing Required Packages

Setting Up API Keys

Connecting to MongoDB

Expected Output:

Working with Databases and Collections

CRUD Operations in MongoDB

Insert a Single Document

Expected Output:

Insert Multiple Documents

Expected Output:

Retrieve Data

Expected Output:

Update a Document

Delete a Document

Drop a Collection

Implementing Retrieval-Augmented Generation (RAG) with MongoDB

Generate and Store Embeddings

Load and Split Text Data

Expected Output:

Store Embeddings in MongoDB

Performing Vector Search in MongoDB

Create Vector Search Index

Run Vector Search Query

Conclusion

Resources

Resources and Community

BuildFast Bot

Introduction

Setting Up MongoDB

Installing Required Packages

Setting Up API Keys

Connecting to MongoDB

Expected Output:

Working with Databases and Collections

CRUD Operations in MongoDB

Insert a Single Document

Expected Output:

Insert Multiple Documents

Expected Output:

Retrieve Data

Expected Output:

Update a Document

Delete a Document

Drop a Collection

Implementing Retrieval-Augmented Generation (RAG) with MongoDB

Generate and Store Embeddings

Load and Split Text Data

Expected Output:

Store Embeddings in MongoDB

Performing Vector Search in MongoDB

Create Vector Search Index

Run Vector Search Query

Conclusion

Resources

Resources and Community