MongoDB added Atlas Vector Search and Atlas Stream Processing to its MongoDB Atlas platform along with other enhancements as it aims to be the top choice for data application developers.

The news, announced at its MongoDB.local NYC developer conference, highlights the race for enterprise developers looking to create modern applications that can readily incorporate generative AI capabilities at scale.

MongoDB's announcements come days after Databricks launched Lakehouse Apps to broaden its development platform ambitions. In addition, Snowflake will unveil updates at its Snowflake Summit next week. Snowflake CEO Frank Slootman last month promised "significant product announcements" at Snowflake Summit.

Dev Ittycheria, CEO of MongoDB, said during a keynote that developers spend most of their time working with data instead of creating software. Multiple clouds, endpoints and data stores have also made development more complicated. Streaming data technologies are also heterogenous. "AI is about building smarter and more intelligent applications," said Ittycheria. "There has been an explosion of AI companies running and building apps on MongoDB. We believe there are 1,500 companies building AI workloads on MongoDB today."

MongoDB Atlas Vector Search will bring generative AI capabilities to the Atlas platform by bringing forward highlight relevant information retrieval and personalization.

Doug Henschen, analyst at Constellation Research, put Atlas Vector Search in context:

"Vector Search isn't a generative AI capability on its own, it's an enabler for companies interested in developing their own generative AI capabilities. In announcing this feature, which is entering public preview, Mongo DB is joining a group of leading-edge data platform companies that have recently made, or are about to make, vector-search-related announcements."

In addition, MongoDB Atlas Stream Processing will surface high-velocity streams of complex data. Atlas Stream Processing, which is in private preview, will enable enterprises to leverage large language models (LLMs) and process streams of real-time data in one unified experience.

Henschen said Atlas Stream Processing is a key addition for MongoDB. He said:

"Atlas Stream Processing is the most important announcement at this week’s event, with Vector Search being the second most important announcement in my book. Low-latency workloads and requirements are only becoming more prevalent, so MongoDB really had to step up on this front if it is to live up to the company’s billing as a "developer data platform." Rival data platforms associated with analytics, such as Snowflake and Databricks, have already addressed real-time needs, so MongoDB is filling a competitive gap."

Vector Search and Stream Processing are likely to appeal to developers building AI-based applications. MongoDB said Beamable, Pureinsights, Anywhere Real Estate and Hootsuite are building next-gen applications with the new Atlas capabilities.

To round out the Atlas updates, MongoDB also added Atlas Search Nodes with dedicated resources for search workloads, efficiency improvements with MongoDB Time Series collections and new Atlas Data Federation for queries and isolating workloads on Microsoft Azure.

For MongoDB, the race to build enterprise-grade generative AI apps is an opportunity to grow its multi-cloud developer data platform. Although enterprises aren't scaling LLMs and generative AI applications yet, the interest is there.

To capitalize on the generative AI and LLM interest, MongoDB is looking to address the following with Atlas Vector Search:

  • Provide the flexibility to store and process different types of data. LLMs require data in the form of vectors to represent data types such as text, images and audio.
  • Store vectors so LLMs can use them without needing a specialized database that lacks integration with existing technology stacks.
  • Enable developers to deploy new workloads such as semantic search, text and image search and personalized product recommendations in one platform.
  • Provide developers with the ability to augment pre-trained generative AI models with their own data.
  • Integrate frameworks such as open source LangChain and LlamaIndex and use them so developers can access LLMs from partners.

With Atlas Stream Processing, MongoDB is looking to do the following:

  • Provide developers with real-time streaming data from IoT devices, browsing and inventory feeds to create real-time experience and optimize on the fly.
  • Leverage streaming data without specialized programming languages, APIs and drivers.
  • Give developers one interface to extract insights from streaming data across multiple data types, connectors and technologies.

While Atlas Vector Search and Atlas Stream Processing were the headliners, MongoDB had a series of other launches. Here's the breakdown.

  • MongoDB Atlas Search Nodes give developers dedicated resources so enterprises can scale search workloads independent of the database.
  • MongoDB Time Series collections provide options to modify data that has already been ingested. Time Series collections will also improve storage efficiency and query speeds.
  • The company is adding Microsoft Azure support to MongoDB Atlas Online Archive and Atlas Data Federation to go with Amazon Web Services. Support for Microsoft Azure Blog Storage means MongoDB customers can work with Azure and AWS datasets.
  • MongoDB launched MongoDB Relational Migrator, a tool that streamlines the process of migrating applications and legacy databases.
  • Google Cloud Vertex AI LLMs will integrate with MongoDB Atlas so developers can use Google Cloud foundational models across MongoDB Atlas Vector Search.
  • The company outlined MongoDB Atlas for Industries, which is a set of integrated tools for vertical use cases. The first industry targeted by MongoDB Atlas for Industries is financial services.
  • MongoDB also outlined additional programming language support for deploying MongoDB Atlas on AWS, the Kotlin Driver for MongoDB for server-side applications and more streamlined functionality for Kubernetes and Python.

Bottom line: MongoDB sent the message that developers can leverage the Atlas platform for generative AI capabilities. Henschen said:

"I think the 5% of companies that are innovators and the next 20% to 25% of companies that are fast followers will be the ones that are most interested in these features. It promises to make MongoDB stickier for developers at these companies as they now know they can turn to MongoDB, a tool they already know and love, for help in developing generative AI capabilities."