MongoDB expands AI search tools to free databases

MongoDB users now have access to search capabilities in free and self-managed deployments, making it easier to build AI applications on all versions of its noSQL database.

The full-text search and vector search tools are now available in public preview on its Enterprise Server and free Community Edition after previously being limited to its Atlas cloud platform.

Head of Core Products Benjamin Cefalo said the expansion gave “developers unparalleled flexibility to build in the environment of their choice” in local, on-premise, and self-managed MongoDB products.

AI applications need search tools

Full-text search is an embedded capability letting users search a NoSQL database and retrieve a broad range of topical, phrasing and citation results related to their query, beyond traditional exact phrase results.

Launched on Atlas in June 2023, vector search goes beyond this again, using machine learning techniques to surface “relevant and context-aware responses” from a search query.

See also: Dropbox's vector search: Challenges and model choices...

As a result, it can be used to let AI agents quickly collect data from a database, or to support RAG workflows by providing more context for LLM responses.

Microsoft offered vector search in it's Azure Cosmos DB for NoSQL earlier this year for the explicit purpose of “making it more versatile for… AI applications.”

Testing purposes only

MongoDB advised that the public preview is currently intended for “testing, evaluation, and feedback purposes only”. Set up information is available through its blog.

However, it added there are no functional limitations in the preview, meaning “$search, $searchMeta, and $vectorSearch are all supported with functional parity to what is available in Atlas.”

The search tools were tested on Community Edition with LLM software frameworks LangChain and LlamaIndex, which already offer native integrations with the database platforms.

Following the preview, the tools are expected to go generally available for on-premise Community Edition deployments, and will be part of a new paid subscription tier for Enterprise Server.