CoRApp: A Web-Based Prototype and Prompt Suite for the Chain-of-Repair Artifact-Aware Debugging Framework

This Zenodo archive contains the complete implementation of CoRApp, an interactive web-based prototype developed to operationalize the Chain-of-Repair framework for artifact-aware fault localization and automated program repair using Large Language Models (LLMs). The archive also includes the full set of structured prompt templates employed by the Localization, Reasoning, and Validation modules described in the accompanying research paper, “Chain-of-Repair: An Artifact-Aware Multi-Stage Reasoning Framework for Fault Localization and Program Repair with Large Language Models.” CoRApp enables researchers and practitioners to reproduce the end-to-end debugging workflow proposed in the study. Through a web-based interface, users can specify the target programming language, provide faulty source code, and optionally supply complementary software artifacts such as bug reports, stack traces, compilation feedback, or revision-history information when available. The system then executes the Chain-of-Repair workflow, which consists of three coordinated stages: (i) artifact-aware fault localization, where suspicious program locations are ranked using heterogeneous software artifacts; (ii) structured multi-stage reasoning, where potential root causes are inferred and candidate patches are generated; and (iii) validation and refinement, where candidate repairs are assessed using compilation feedback, static analysis signals, lightweight semantic constraints, and fault–patch causal alignment checks. The framework does not use executable test suites during fault localization, diagnosis, patch generation, or in-loop validation. Instead, benchmark-provided test suites are reserved exclusively for retrospective correctness assessment, consistent with the experimental methodology reported in the accompanying paper. All prompt templates included in this archive are provided in the same form used during the experimental evaluation, supporting transparency and reproducibility. The prototype reflects the localization, reasoning, validation, and refinement workflow reported in the study and allows users to inspect intermediate reasoning artifacts, generated patches, validation outcomes, and developer-oriented explanations. This artifact is intended for research, educational, and experimental use, particularly in software maintenance scenarios where executable tests are unavailable, incomplete, costly to execute, or intentionally excluded from the repair-generation workflow. By releasing both the prototype and the complete prompt suite, this archive supports independent replication, qualitative analysis, and future extensions of the Chain-of-Repair framework.