MongoDB today revealed it is making available a public preview of the community and enterprise editions of its core document database platform that now have integrated search and vector search capabilities.
Announced at a MongoDB.local NYC event, MongoDB is also now previewing a Queryable Encryption capability by introducing support for prefix, suffix, and substring queries. Available in version 8.2 of the MongoDB database, Queryable Encryption enables IT teams to encrypt and store sensitive application data while still being able to launch expressive queries without first decrypting data, which MongoDB claims is an industry-first capability that will be made available at no additional cost.
Ben Cefalo, senior vice president and head of core products at MongoDB, said as more organizations look to operationalize artificial intelligence (AI) the need for search and vector search capabilities has significantly expanded. Many of those AI applications are being built on top of the MongoDB database because of its inherent ability to model data using the JavaScript Object Notation (JSON) format, he added.
Previously, the search and vector search capabilities were only available via the fully managed MongoDB Atlas cloud platform. IT organizations looking to build AI applications using either MongoDB Community Edition and MongoDB Enterprise Server would have needed to deploy external search engines or vector databases, which creates additional operational overhead, noted Cefalo
IT teams now will be able to more broadly take advantage of capabilities such as full-text, semantic retrieval, and hybrid search to create more accurate and context-aware AI applications using, for example, retrieval-augmented generation (RAG) techniques, he said.
The move to add search and vector search capabilities comes at a time when more organizations are moving to build and deploy AI agents. MongoDB can provide the long-term memory store that these AI agents require to autonomously perform tasks, noted Cefalo. That capability is especially critical for AI agents that will need to be able to reason in real time, he added.
A recent Futurum Group report projects AI agents will drive up to $6 trillion in economic value by 2028. Many of those organizations have already deployed MongoDB databases. In fact, the company claims more than 75% of the Fortune 100 are MongoDB customers.
It’s not clear to what degree IT teams are looking for databases that have integrated search and vector capabilities versus deploying an external platform that might be managed by the team building AI applications and databases. In general, there is an argument to be made for centralizing database platforms to contain costs, but there may be instances where the performance requirements of an AI application may necessitate deployment of, for example, a separate vector database. In many cases, IT organizations will find themselves managing a mix of database platforms that, to varying degrees, are being used by AI applications and agents as their primary data store.
Regardless of how database platforms in the age of AI are deployed and managed, the one thing that is certain is that it’s unlikely there will be any one size fits all solution any time soon.
