MongoDB reveals new generative AI, vector search tools

New AI capabilities

Like most tech vendors, MongoDB has made generative AI a focal point of its product development since OpenAI's launch of ChatGPT in November 2022 significantly advanced generative AI and large language model functionality.

In June, the vendor unveiled its first generative AI capabilities, including an extension of its partnership with Google Cloud that will enable developers to use the tech giant's generative AI and LLM capabilities as they create applications in MongoDB Atlas.

Now MongoDB is introducing specific generative AI tools that let users interact with data using natural language rather than code. They include the following:

• Natural language query in MongoDB Compass that lets users generate queries and infuse data assets in applications.
• Natural language visualization in MongoDB Atlas Charts so developers can create, share and embed visualizations.
• An AI-powered chatbot in MongoDB Documentation that provides users with tutorials, code samples and reference libraries as they build applications with MongoDB.

The chatbot is now generally available, while the NLP capabilities in Compass and Atlas Charts are in preview.

The features are similar to those being developed by other data management and analytics vendors, according to Stephen Catanzano, an analyst at TechTarget's Enterprise Strategy Group.

But they are nevertheless important given that they lessen the burden on data workers. In addition, whether one vendor's NLP capabilities are stronger than another's won't be known until all the tools are generally available, he noted.

"MongoDB, like other vendors, is integrating generative AI into their products to reduce the manual and repetitive tasks," Catanzano said. "These functions are similar to what others are doing. But there may be more nuances once we can see the full implementation on how they are doing this versus others."

Beyond NLP capabilities, MongoDB unveiled new AI-powered capabilities in MongoDB Relational Migrator that make it faster and easier for organizations to migrate data from relational and other database types to MongoDB. The vendor first made Relational Migrator generally available in June and added capabilities now in preview automatically convert SQL queries to MongoDB Query API syntax.

Meanwhile, improvements to MongoDB Atlas Vector Search -- a tool also still in preview -- are designed to reduce the frequency of AI hallucinations that plague generative AI and LLM outputs.

The most significant [new feature] is vector search. Most database companies like MongoDB are adding this to support generative AI workloads. Vector search has a lot of use cases and is a must-have going forward for databases to play in the generative AI space.

They include a dedicated data aggregation stage to filter results, accelerated indexing that shows metadata and other information that reveal data's lineage and whether it can be trusted, and faster and easier access to streaming data to enable real-time analysis from generative AI models.

Because vector search can reduce AI hallucinations, MongoDB's move to add functionality to Atlas Vector Search is critical, according to Catanzano.

He noted that vectors enable LLMs to identify similarity among data. Through identifying similar data, generative AI can learn from itself and return more accurate results.

"The most significant [new feature] is vector search," Catanzano said. "Most database companies like MongoDB are adding this to support generative AI workloads. Vector search has a lot of use cases and is a must-have going forward for databases to play in the generative AI space."

Beyond the new capabilities in Atlas Vector Search, MongoDB unveiled an integration between Atlas Vector Search and data streaming specialist Confluent that enables developers to access streaming data for use in generative AI models.

Andrew Davidson, MongoDB's senior vice president of products, said that the new capabilities in combination -- which the vendor terms its intelligent developer experience -- are designed to enable developers to build faster within MongoDB's operational data layer.

Meanwhile, coming just three months after the vendor's initial foray into generative AI, they represent progress from theory to practice, he said.

"In June, we talked about how we were looking at [generative AI]," Davidson said. "Now we're launching a bunch of things. Back in June, we knew we wanted to do a lot of things, but our plans weren't specific. We knew we could do a bunch of things to modernize, and now we're bringing it to life."

Living on the edge

Beyond new generative AI capabilities -- and those such as vector search that power generative AI -- MongoDB also unveiled Atlas for the Edge.

Atlas for the Edge, now in preview, is a set of capabilities aimed at enabling users to deploy MongoDB applications where data is created, processed and stored rather than just the vendor's database. In addition, the tool synchronizes an organization's edge data with the rest of its data so the edge data is not isolated.

Davidson noted that data can be created and stored anywhere from an IoT sensor or point-of-sale system to a major cloud provider's platform. Rather than force users to import that data into MongoDB to train models and inform other applications -- including generative AI -- the vendor is aiming to simplify development by bringing MongoDB to the data.

According to MongoDB, Atlas for the Edge enables customers to do the following:

• Deploy MongoDB using varied infrastructures from on-premises servers to remote locations in warehouses or hospitals that are generally disconnected from other data sources so that data is connected and available for real-time analysis.
• Run applications in locations with intermittent connectivity to prevent data loss when those locations are offline.
• Integrate with generative AI and machine learning tools to enable development and deployment of generative AI capabilities directly in devices.
• Store and process both real-time as well as batch data from IoT devices so it can be synchronized with other data in edge locations and used for predictive maintenance and anomaly detection.
• Secure edge applications to ensure data privacy and regulatory compliance.

"We know that there's more compute at the edge … where the business is happening," Davidson said. "Now we have the ability to have an edge server that synchronizes itself back up to Atlas."

Catanzano likewise noted that the key aspect of Atlas for the Edge is that it provides a unified interface to enable development where business happens.

Previously, customers could build their own tools to deploy MongoDB in varied environments and then synchronize the data created in those varied environments. Now, MongoDB is doing that for them.

"They have expanded Atlas to the Edge with a single user interface that connects all edge devices, which was not easily done before this improvement," Catanzano said. "It's an important update and improvement for large scale deployments such as kiosks and IoT devices."

Looking ahead, Davidson said MongoDB's roadmap will center on adding performance and scalability while at the same time making its tools easier to use.

Beyond generative AI, time series analysis and stream processing are specific areas of focus. But at the core of the vendor's product development planning is simplification.

Catanzano, meanwhile, said MongoDB's focus on generative AI and streaming data is appropriate.

"They are moving quickly on generative AI, streaming data and important use cases," he said.

Eric Avidon is a senior news writer for TechTarget Editorial and a journalist with more than 25 years of experience. He covers analytics and data management.