CortexAST

소개

CortexAST serves as the AI-native backend for code intelligence, designed to extract essential signals from codebases while drastically reducing noise for large language model (LLM) agents. Unlike traditional tools built for human interaction, CortexAST leverages Tree-sitter and Rust to provide LLMs with a deterministic, AST-level understanding of code, significantly cutting token usage by up to 90%. It facilitates persistent cross-session memory by integrating with CortexSync and offers a robust, multi-tier rule engine for managing AI coding standards across global, team, and project levels.

주요 기능

• Micro-Bootloader for Dynamic Rule Retrieval: Eliminates large, static instruction files by allowing AI agents to dynamically fetch documentation, history, and rules, leading to drastic token savings and higher accuracy.
• Nuclear Skeletonisation: Optimizes token usage by collapsing function bodies to signatures, stripping imports, and flattening indentation, focusing on architectural logic.
• JIT Hybrid Vector Search: Provides on-demand, incremental, and efficient semantic code search using a pure Rust `model2vec-rs` implementation.
• Bulletproof Safety & Output Control: Includes features like null-byte detection, file caps, minified-bundle guards, and `max_chars` truncation to ensure stable and secure operation, preventing editor-side output spill.
• 3-Tier Rule Engine: Manages enterprise-grade AI coding rules across global, team/cluster, and project-specific levels for consistent standards and overrides.
• 7 GitHub stars

사용 사례

• Efficient Codebase Exploration for AI Agents: Rapidly generate bird's-eye architectural maps and token-budget-aware slices of repositories, providing deep insights without overwhelming context windows.
• Performing Safe and Accurate Code Refactoring: Utilize AST time machine snapshots (`cortex_chronos`) for structural diffs, ensuring crystal-clear semantic changes without whitespace noise, critical for AI-driven refactors.
• Preventing Cross-Service Propagation Errors: Automatically generate strict checklists for changes to shared types, interfaces, or API contracts across multiple languages/domains (e.g., Proto → Rust → TS), ensuring complete updates.