RefactorGPT Code and dataset files

This repository contains the source code, benchmark datasets, and experimental results for the study titled "RefactorGPT: A ChatGPT-Based Multi-Agent Framework for Automated Code Refactoring". RefactorGPT is a modular and extensible system that utilizes multiple task-oriented agents powered by large language models to automatically analyze, refactor, refine, and validate Python code. The framework enables step-by-step interaction with LLMs through structured prompts, ensuring functional correctness, improved modularity, and enhanced code readability. This open-access release includes: Python scripts for all four agents (AnalyzerAgent, RefactorAgent, RefineAgent, FixerAgent) Prompt templates used for each stage of the refactoring pipeline Benchmark dataset with original and refactored Python code samples Evaluation scripts and metrics (e.g., execution time, memory usage, LOC, function count, token count) Sample output files and human evaluation annotations (if applicable) The project supports reproducible research and encourages future exploration in the domain of LLM-based software engineering. All materials are released under an open license and can be reused with attribution.