TigerGraph Inc. is upgrading its graph database with a hybrid search capability designed to power artificial intelligence applications.

The Redwood City, California-based startup debuted the feature today alongside a new free edition of the database.

A graph is a data structure that holds not only business records but also information about how those records are connected to one another. For example, it can point out if two purchase logs were produced by the same e-commerce website. The ability to track such data relationships is necessary for many analytics projects.

TigerGraph provides a popular graph database of the same name. The platform can store graphs with up to trillions of edges, data points that show how different pieces of information are connected to one another. The company counts Microsoft Corp., JPMorgan Chase and other major enterprises among its customers.

TigerGraph’s new hybrid search capability combines its existing graph-based tools for finding connections between data points with a vector search capability. According to the company, the enhancements will enable AI applications powered by its database to retrieve information for users more reliably. 

Vectors are data structures that can track relationships between different snippets of information, much like a graph. However, they often store different kinds of relationships. Whereas a graph might highlight that two business documents belong to the same department, a vector stores semantic similarities such as the fact the documents discuss the same topic. 

According to TigerGraph, combining graph and vector search allows AI models to retrieve information more reliably than would be possible using only one or the other. That helps improve the quality of prompt responses. 

One of the tasks that TigerGraph promises to ease with the new feature is knowledge base navigation. A company could use hybrid search to build an AI chatbot that helps workers find internal documents. Such a chatbot can use graphs to find all the files created by a user’s business unit, then leverage vectors to find the file most relevant to the user’s search query. 

The feature also lends itself to other tasks. According to TigerGraph, the ability to find patterns in interconnected datasets makes it easier to generate shopping recommendations and identify supply chain inefficiencies. The company says that its database can complete some queries five times faster than competing platforms. 

Developers can access the hybrid search feature via GSQL, the custom query language that TigerGraph ships with its database. The syntax is similar to SQL, which makes it relatively simple to learn. There’s also a library that enables software teams to access data using the Python programming language. 

TigerGraph introduced its new hybrid search capability today alongside a free version of its database called TigerGraph DB Community Edition. The offering supports environments with up to 16 central processing units. Free users can store up to 200 gigabytes of graph data and 100 gigabytes’ worth of vectors. 

Image: Unsplash