Future database search is temporal & semantic

Future database search is temporal & semantic

Databases evolve. If anyone thought that early versions of this data management fundamental (back somewhere in the mists of our progression from mainframes towards PCs in the sixties) would last forever, they would not believe the variety and diversity of options available today. At a time when the notion of multi-modal unstructured data search might have sounded like science fiction, this is part of the way we deal with the panoply of information analysis that we need to carry out today.

Bringing multi-modal search (the process of enabling search by text in traditional typed form, search via image, search by speech, semantic search inference of more than one thing in the context of another thing or object and more besides) among other core functionalities to its platform is KX, a firm known for its time-series vector database competencies and its wider approach to AI in relation to Large Language Models (LLMs) and more.

Real-time contextual AI

A snappy name if you’re fond of acronyms, the company’s latest development is the KDB.AI Cloud, described as a vector database for real-time contextual AI. While we might consider any AI that fails to be contextual to be something of a failure, KX presumably uses the term to explain that this free cloud-based version of the KDB.AI vector database has an acute ability to deliver analysis based on database object references that map to the real world.

It’s exactly that, the company says that KDB.AI Cloud enables developers to bring temporal (time-based) and semantic (meaning and intent-based) context and relevancy to AI-powered applications.

Engineered to handle high-speed time-series data and multi-mode query data processing, it allows users (for example) to query real-time financial market data using natural language search with semantic relevance. 

Temporal awareness 

But there’s temporal awareness here as well and that means KDB.AI Cloud answers questions based on ad-hoc time windows such as time markers that relate to when data creation happened, what kind of timeframes represent data modification recency, or when periodic comparisons have been made. This helps applications find and return more relevant data and allows for point-in-time and like-for-like comparisons.

KDB.AI Cloud works with popular LLMs and machine learning workflows and tools, including LangChain and ChatGPT, while native support for Python and RESTful APIs means developers can perform common operations like data ingestion, search and analytics using their preferred applications and languages.

“KDB.AI Cloud exemplifies our commitment to elevate the developer experience, setting them at the forefront of generative AI’s future, further enhanced by discriminative AI,” remarked Ashok Reddy, CEO of KX. “Our platform infuses time-awareness and situational understanding into vector database-driven AI processes, ensuring unmatched precision for generative AI applications. I eagerly anticipate the ground-breaking solutions that developers worldwide will forge with KDB.AI Cloud, reshaping industries and establishing fresh standards for innovation.”

Key use cases 

Key use cases here include the previously mentioned ability to use multi-modal unstructured data search. This also embodies the ability to use similarity search between objects in any data format – video, image, text, time series, or unstructured. This technology also offers automation for digital twins i.e. by providing situational awareness based on time, vectors can be applied to provide context into streams of IoT data from digital twins, to automate with confidence.

Also here we find pattern matching and anomaly detection. This database can spot anomalies in data sets to build data integrity and boost performance. There are also recommendation systems to help refine algorithms based on feedback loops for adaptive user experiences. Finally, sentiment analysis can be used to detect customer patterns and improve user experiences.

KDB.AI is used across multiple industries, including finance, energy, manufacturing and telecommunications. EnterpriseWeb, which offers an industrial-grade no-code automation platform for complex distributed systems, uses KDB.AI to integrate real-time location and cost-based analytics on advanced telecom use-cases for self-scaling, self-healing and self-optimizing networks.

“KX is renowned for its ability to process and analyse time series, historical and vector data at speed and scale,” said Dave Duggal, founder and CEO of EnterpriseWeb. “KDB.AI builds on these capabilities with sliding window search and the ability to augment LLMs. We’re excited to partner with KX to further the integration of KDB.AI into our platform.”

A whole new vector

KDB.AI Cloud is available now as a free to use, SaaS-based service from KX. For those just starting out in their use (or consideration of) vector databases, this offering could be of interest. We’re a long way on from the clunky era of the first databases. ThinkAutomation suggests that, “Charles Bachman designed the first database known as the Integrated Data Store, followed by the Information Management System developed by IBM.”

While reports trying to pin down the origin of database history disagree with each other and some feature apocryphal misinformation, we have certainly moved on to a time (pun not intended) of temporal contextual multi-modal data search in natural language – and that’s a whole new vector (pun intended) in and of itself.