Promptware Engineering for Prompt-Enabled Systems

This paper is a research article that introduces the concept of promptware engineering, a methodology aimed at improving the development of prompt-enabled systems that utilize Large Language Models (LLMs). It outlines the unique challenges associated with prompt development, which often relies on ad hoc and trial-and-error methods, leading to what is termed the promptware crisis. The authors propose a systematic framework that adapts established Software Engineering (SE) principles to the specific needs of prompt development. This framework encompasses various stages including requirements engineering, design, implementation, testing, debugging, evolution, deployment, and monitoring. By integrating these core SE activities, the paper aims to provide a structured approach to prompt development, moving beyond experimental practices to ensure quality and manage complexity. The authors also discuss key challenges in the field and highlight potential research directions to advance the discipline of promptware engineering, emphasizing its significance in the evolving landscape of software applications that incorporate LLMs.