Generating UI Code for Scientific Command Line Tools Using Large Language Models
Skogberg, Kristian (2025)
2025
Tietotekniikan DI-ohjelma - Master's Programme in Information Technology
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2025-05-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202505074943
https://urn.fi/URN:NBN:fi:tuni-202505074943
Tiivistelmä
This thesis explores the current capabilities and limitations of using artificial intelligence (AI) and large language models (LLMs) to generate user interface (UI) code. In the action research component of this thesis, a graphical user interface (GUI) was developed for VeRyPy, a scientific Python library for solving vehicle routing problems. The GUI code was generated using GitHub Copilot and OpenAI’s GPT-4o model.
In this action research, the VeRyPy GUI development process was carried out in five iterations, following cycles of planning, action, analysis, and conclusion. In the beginning of the research, the GUI requirements were gathered and structured into user stories, which were then mapped to an iteration plan. In the first iteration, a GUI design was generated using two AI tools: Vercel V0 and Galileo AI. In the subsequent iterations, the GUI features were generated in code according to the iteration plan. The GUI development workflow was documented in detail in the results chapter.
Although AI significantly accelerated especially the early stages of GUI development, it still has notable limitations, such as inability to manage large contexts, occasional unintended code modifications, and challenges in integrating the AI-generated code into existing codebases. Leveraging AI in software development is still a relatively manual process, as it requires writing numerous prompts, reviewing changes, and manual testing to achieve the best results.
Based on the results of the action research, an autonomous UI code generation process utilizing LLMs was proposed. In this process, AI would be used to generate tests prior to generating code for the required features. These tests would be executed whenever new code is applied to the codebase, with the context being updated in the background. Although this process could eliminate some of the manual work involved in AI-powered UI development, having clear requirements, continuous iteration, and validation remain essential aspects of software engineering.
In this action research, the VeRyPy GUI development process was carried out in five iterations, following cycles of planning, action, analysis, and conclusion. In the beginning of the research, the GUI requirements were gathered and structured into user stories, which were then mapped to an iteration plan. In the first iteration, a GUI design was generated using two AI tools: Vercel V0 and Galileo AI. In the subsequent iterations, the GUI features were generated in code according to the iteration plan. The GUI development workflow was documented in detail in the results chapter.
Although AI significantly accelerated especially the early stages of GUI development, it still has notable limitations, such as inability to manage large contexts, occasional unintended code modifications, and challenges in integrating the AI-generated code into existing codebases. Leveraging AI in software development is still a relatively manual process, as it requires writing numerous prompts, reviewing changes, and manual testing to achieve the best results.
Based on the results of the action research, an autonomous UI code generation process utilizing LLMs was proposed. In this process, AI would be used to generate tests prior to generating code for the required features. These tests would be executed whenever new code is applied to the codebase, with the context being updated in the background. Although this process could eliminate some of the manual work involved in AI-powered UI development, having clear requirements, continuous iteration, and validation remain essential aspects of software engineering.