Show HN: Graphenium, persistent repo memory for AI coding assistants (github.com)

Graphenium, a new tool for AI coding assistants, enhances the way these systems navigate code repositories by transforming them into queryable graphs. This innovation allows AI assistants like Claude and Cursor to quickly answer structural questions—such as dependencies and relationships between components—without having to read multiple files, achieving response times of around 20 milliseconds. This capability is particularly significant for large and complex codebases, where traditional grep-and-trace methods are inefficient and time-consuming. The implications for the AI/ML community are profound. Graphenium addresses persistent issues such as repeated cold starts, context window pressure, and the lack of structural memory in AI navigation. By running a one-time analysis to create a persistent graph, Graphenium enables AI assistants to maintain knowledge across sessions, allowing for faster onboarding and more effective impact analysis during code changes. Its architecture leverages AST extraction across various programming languages and provides AI assistants with tools to query and analyze the code structure, ultimately enhancing the efficiency and effectiveness of AI-assisted coding and maintenance tasks.