ChatQPT: Towards Conversing with Relational Query Engines

Abstract

Imagine a scenario where we could communicate with a relational query processor through messaging platforms like WeChat or WhatsApp , asking questions related to query processing and optimization. If realized, such a tool could prove invaluable in database education and administration, among others. In this demonstration, we present ChatQPT, a novel system that enables chat-based interactions with a relational query engine (PostgreSQL). It combines an LLM-powered interface with a set of specialized external tools tailored to understand various aspects of relational query processing, effectively addressing the shortcomings of both large language models and off-the-shelf RDBMSs in supporting accurate and effective conversations. Our preliminary evaluation through a user study demonstrates the promising capabilities of ChatQPT.