_Book_Claude-Code

For the open-source project 'EcoTrack' (a carbon footprint calculator), I used Claude-Code to review the Python codebase. The review revealed three critical issues: 1) A memory leak in the carbon emission calculation module where the Pandas DataFrame wasn't being properly cleared after processing large datasets, causing memory usage to spike from 200MB to 2GB during batch operations. The AI suggested implementing a context manager to handle DataFrame cleanup. 2) Inconsistent error handling across the API endpoints—some used custom exceptions while others relied on generic HTTP errors. The recommendation was to standardize on FastAPI's built-in exception handlers. 3) The Dockerfile lacked multi-stage builds, resulting in a 1.2GB final image. The AI proposed a three-stage build process that reduced the image size to 350MB while maintaining all dependencies. Additionally, it identified a security vulnerability in the JWT token generation where the secret key was hardcoded in the environment variables. The fix involved using AWS Secrets Manager for key rotation. The output included a refactored code snippet for the memory leak fix and a PR-ready commit message summarizing all changes. The entire process took 12 minutes compared to the manual review's estimated 3 hours.