Code Analysis with Drift: A New Approach

A developer has introduced Drift, an innovative code analysis tool that stands out for its approach based on semantic learning through AST parsing and Regex fallbacks. This tool aims to overcome the limitations of AI agents, often hampered by the narrow context window.

How Drift Works

Drift examines the codebase and extracts 15 different categories, identifying over 150 patterns. The collected information is stored and can be recalled via CLI or directly in the preferred IDE. The real strength of Drift lies in its ability to learn code conventions, thus avoiding the problems of forgetfulness and errors that plague AI agents.

Key Features

  • Call graph analysis: Drift doesn't just list existing functions, but also analyzes code reachability, identifying what data it can access and who can access a given field.
  • Impact analysis: The tool assesses the consequences of a code change, providing a risk score.
  • Security: Drift offers enterprise-grade security analysis, accessible via CLI or IDE.

Drift currently supports TypeScript, Python, Java, C#, PHP, and Go, using Tree sitter parsing and Regex fallbacks. Data is saved locally, and the user can approve, reject, or ignore specific components to refine agent training.