LUNAR – AI-Powered Technical Debt Guardian

LUNAR – AI-Powered Technical Debt Guardian is an intelligent developer-maintenance solution designed to make technical debt visible, understandable, and actionable. As software projects grow, technical debt can accumulate through duplicate code, complex functions, unused code, security issues, TODO/FIXME items, insufficient tests, and dependency-related concerns. These problems are often discovered only when developers modify, debug, or extend existing software, increasing maintenance effort and the risk of regressions. LUNAR addresses this problem through an end-to-end workflow: Scan → Detect → Understand → Prioritize → Fix → Verify. It analyzes a repository using automated code-analysis tools and custom rules, converts technical findings into understandable issues, and provides developers with context about why each issue matters. To help developers focus their effort, LUNAR uses a transparent, project-specific prioritization approach based on factors such as impact, risk, frequency, and estimated fixing effort. Rather than presenting developers with an overwhelming list of warnings, the system helps organize findings into actionable maintenance work. LUNAR also supports the improvement stage by providing fix guidance and connecting changes with automated testing. Developers can compare the state of the code before and after improvements and use test results as evidence that changes have not introduced new problems. The prototype is built using open-source technologies including Python, FastAPI, Ruff, Radon, Bandit, pytest, and a web-based frontend. An intentionally imperfect demonstration repository is included to show how LUNAR identifies realistic technical-debt issues. The goal of LUNAR is not simply to detect bad code, but to create a practical maintenance workflow where developers can understand problems, decide what deserves attention, make improvements, and verify the results. LUNAR demonstrates how AI-assisted development workflows can help transform technical debt from a hidden, reactive problem into a visible and structured engineering process.