Rockset Expands AI Capabilities of Its Real-Time Database

Real-time database vendor Rockset is enhancing its namesake database with improved vector search and scalability. The company, founded by creators of the open-source RocksDB key-value store, has raised a total of $105 million in funding to support its development. The new update introduces vector search to the platform, expanding Rockset’s capabilities in the generative AI space. JetBlue, among other early adopters, has already provided positive feedback on Rockset’s vector search feature.

Vector Search and Real-Time Updates

With the general availability (GA) release of vector search, Rockset enables users to build similarity indexes using the approximate nearest neighbor (ANN) approach. Venkat Venkataramani, co-founder and CEO of Rockset, explains, “You can do that at a massive scale while also having real-time updates on your vector embeddings and your metadata.” This real-time ability to update vector embeddings and metadata sets Rockset apart from other vector databases in the market.

Rockset offers a unique approach to real-time data indexing and query performance through its compute-compute separation model. This model separates the compute resources used for building indexes from those used for queries, allowing the database to update indexes in real-time with single-digit millisecond latencies. Venkataramani emphasizes the advantage of this approach, stating, “With almost all other vector databases you can’t update in real time; you have to rebuild your index periodically.”

Use Cases and Future of Vector Databases

Vector search plays a vital role in powering large language models (LLMs). While other specialized vector databases exist, Rockset’s focus on real-time updates sets it apart. Venkataramani outlines how Rockset combines the ANN and the more precise K Nearest Neighbor (KNN) approaches based on the query and data. The query optimizer makes the decision on whether to use ANN or KNN, resulting in fast and accurate search results.

Despite the recent advancements in generative AI by OpenAI, Venkataramani remains confident in the continued need for vector databases for larger enterprise applications and complex datasets. He points out that organizations with security and compliance needs cannot rely solely on third-party companies for chatbot development. Additionally, he identifies various use cases, including similarity search at scale, where vector databases remain essential. Venkataramani believes that the use cases for vector databases are evolving rather than disappearing.

To summarize, Rockset’s latest update introduces vector search capabilities to its real-time database, offering users the ability to build similarity indexes. The focus on real-time updates sets Rockset apart from other vector databases in the market. Despite the advancements in generative AI, Venkataramani believes that vector databases will continue to play a crucial role in powering larger enterprise applications and complex datasets.