This deposit contains the full Python code for the LLM-based approach described in “Leveraging Large Language Models to Classify and Inspect Defects in Reinforced Autoclaved Aerated Concrete (RAAC).” It includes six scripts: two parallel prompt modules for RAAC mention detection and definition extraction; a seven-question defect-extraction script; and a data-aggregation script that produces a unified defect database. All scripts are versioned for reproducibility and require Python 3.11+, the Anthropic Claude 3 Opus API, and standard data-analysis libraries.

A comprehensive README.md is included, detailing environment setup, dependency installation, API key configuration, and step-by-step execution instructions. The code is mirrored on GitHub for ongoing collaboration and version tracking. An interactive project overview and navigation interface is also provided via index.html on the project’s GitHub Pages site.