With our 24.2 release, CockroachDB adds support for the VECTOR data type, along with a set of pgvector compatible functions for doing interesting things like computing similarity between vectors. This article provides a brief overview of some related concepts and introduces a semantic search application to illustrate this new capability in action. These new functionalities demonstrate CockroachDB’s expanding support for AI-driven applications, such as Large Language Models (LLMs).

Michael Goddard

October 23, 2024

We’ve been planning a visit to Orange County, California, where I grew up, over the upcoming holidays. My favorite Mexican restaurant is there, in Placentia: Q-Tortas! I have frequent cravings for their carnitas burrito, so a visit there is obligatory (I promise this blog is about search).

Michael Goddard

December 12, 2022

Full text indexing and search is such a common feature of applications these days. Users expect to be able to find a restaurant, a product, a movie review, or any number of other things quickly by entering a few search terms into a user interface. Many of these apps are built using a relational database as the data store, so the apps aren’t generally dedicated to search, but incorporate that as an added feature.

Michael Goddard

July 27, 2022

Last week, while running a workload consisting of 200 different queries, we noticed right away that a CPU imbalance was causing a performance issue. Looking at the first graph, below, you can see right away that one of the three CockroachDB nodes was operating at near 100% CPU. Not ideal.

Michael Goddard

May 23, 2022

A recent tweet inspired me to address the need for fuzzy matching by combining some existing capabilities of CockroachDB. Note the key features mentioned in the tweet: - similar but not equal sporting events names: a common pattern. Users tend to mis-type data in input fields, and data isn’t always correct. Nevertheless, we’d like to return the closest match. - I’d rather use this in-built feature than pay for a whole ES cluster with added maintenance overhead to boot: This is the second time I’ve heard this sentiment in the past couple of months. ES is a full-featured search engine and delivers a great experience but, for this purpose, would be overkill and would require additional time and expense to deploy and operate.

Michael Goddard

April 18, 2022

Imagine you’re driving a rental car in Rome and the satnav (or GPS) on your phone stops working. This happened to me two years ago when I was commuting by car each day from an Airbnb in Trastevere to an office on Via Amsterdam.

Michael Goddard

March 9, 2022

In this post, I’ll skim the surface of a very common pattern in application development: full text indexing and search. I’ll start with a bit of motivation, what prompted me to explore this using CockroachDB. Next, I’ll introduce the initial pass at a solution, followed by a deeper explanation of how that was done, and I’ll then improve on that result by adding a “score”. Finally, I’ll discuss the limitations of this simplistic approach, within the context of information retrieval, ending with my answer to “So, why’d you do it?” Let’s get started.

Michael Goddard

August 27, 2020