Show HN: Crespo – Tree-sitter AST blueprints instead of raw code for LLMs (github.com)

Crespo, a new tool leveraging Tree-sitter's Abstract Syntax Tree (AST) parsing, aims to enhance the way developers utilize Large Language Models (LLMs) by simplifying code comprehension. Instead of feeding LLMs entire codebases that often exceed context limits, Crespo extracts essential structural components—such as imports, classes, and functions—into a compact XML blueprint. This allows AI models to grasp the underlying architecture of projects without the clutter of raw code, significantly reducing token usage by up to 96% in some cases. This innovation is significant for the AI/ML community as it addresses a common frustration: LLMs struggle with large code files, often losing track of the project's overall structure. By using Crespo, developers can ensure that models retain a clear understanding of the codebase, enhancing tasks such as architectural reviews and onboarding. Additionally, the tool redacts sensitive information before outputting the blueprint, ensuring secure interactions with LLMs. With plans for further enhancements, including support for more programming languages and additional output modes, Crespo represents a substantial step toward improving AI readability of complex code structures.