Vector Databases

Chapter 9. Vector Query Language

This chapter is an experiment. It is not a typical technology book chapter. The goal here is to present a prototype of an idea for discussion and consensus building in the community for a longer-term collaborative development project, ideally an official standards body effort.

Vector Query Language (VQL) is a SQL-inspired query language designed for vector databases. It combines the familiar syntax of SQL with powerful vector operations, making it intuitive for developers while providing advanced capabilities for machine learning and AI applications.

This language doesn’t currently exist as an implementation. This chapter therefore describes a hypothetical version, including its syntax and supported operations.

Being a first attempt at writing down such a language, there will be areas that may have been overlooked; certainly the scope is larger than what this chapter expresses. For instance, the equivalent of indexing and data definition language (DDL) are missing. The intent is to focus first on those aspects of VQL that are unique to vector spaces, leaving the routine aspects—such as creating collections and indexes, which follow naturally—for later discussion.

The larger intent is to inspire the community to come together and create a consensus implementation and then, in the future, have a standards body that works similarly to the SQL body, with representatives from vector database companies, tool builders, and middleware vendors. While ...