Hanzoly: A Survey of AI-Powered Software Handover and Codebase Understanding Using Local Language Models

HANZOLY is proposed as an AI-powered software handover and understanding platform. Its primary workflow scans an existing project locally, extracts code structure, organizes dependencies, uses a local LLM to explain technical findings, and produces a structured handover report. Change impact analysis, Git Diff, Pytest comparison, and optional security checks provide supporting maintenance information. The approach aims to organize knowledge transfer while keeping analysis in a local or private environment when configured accordingly. Future work should implement the modules, evaluate them on representative projects, measure extraction and impactanalysis quality, and document privacy and security boundaries.

[1]   Python Software Foundation, “ast — AbstractSyntax Trees,” Python Documentation. Available: https://docs.python.org/3/library/ast.html [2]   pytest Development Team, “pytest documentation.” Available: https://docs.pytest.org/ [3]   Git Project, “git-diff — Show changes betweencommits, commit and working tree, etc.” Available: https://git-scm.com/docs/git-diff [4]   FastAPI, “FastAPI Documentation.” Available: https://fastapi.tiangolo.com/ [5]   Ollama, “Ollama Documentation.” Available: https://docs.ollama.com/ Python Software Foundation, “json — JSONencoder and decoder,” Python Documentation. Available: https://docs.python.org/3/library/json.html