Hiding the LLM Cost Line on AI Stack Cards
Technical honesty is one of the most critical values in software engineering documentation. When designing the Engineering section of this portfolio (`/engineering/ai`) to showcase how AI coding assistants navigate our codebase, an early UI component mockup for our AI stack cards included a standard metric line: "Estimated LLM API Cost: $0.002 / run".
While that metric looked slick and modern in early design drafts, it suffered from a fundamental flaw: it was factually untrue. Our repository's AI context pipeline does not make external LLM API calls during developer builds. Keeping that cost metric in production would have been misleading and contradicted the core architectural philosophy of our project.
Our codebase context system is powered by Graphify, an open-source tool that constructs knowledge graphs of repositories using offline Abstract Syntax Tree (AST) parsing via Tree-sitter. When a developer or CI runner executes `npm run ai:setup` or `npm run ai:doctor`, Graphify parses code files into an interconnected graph stored locally in `graphify-out/graph.json`.