开发者在 Hacker News [1] 上介绍了一款名为 Arcaide 的新工具，旨在解决传统调用图在面对大型代码库时存在的架构上下文丢失和规模膨胀问题。该工具通过构建“多级别”调用图，将控制流分析从函数级扩展至类、包等高层单元 [1]。Arcaide 利用大语言模型（LLM）进行语义分析，以检测遥测数据并识别外部依赖关系 [1]。最终生成的复合图表整合了代码的结构信息与行为信息，包含包级依赖、类组成及用户交互细节，帮助开发者在代码生成时代更好地导航和理解大型项目 [1]。
Author of the new tool Arcaide has introduced a solution designed to address limitations in traditional call graphs, specifically regarding lost architectural context and scale expansion [1]. The application constructs "multi-level" call graphs that extend control flow analysis from the function level up to higher-order units such as classes and packages [1]. By leveraging large language models (LLMs) for semantic analysis, Arcaide detects telemetry data and identifies external dependencies within codebases [1]. This process generates composite charts containing both structural and behavioral information regarding package-level dependencies, class composition, and user interactions [1]. The tool aims to assist developers in navigating and understanding large-scale repositories more effectively during the era of AI-assisted code generation [1].