VA-Blueprint presents a methodology and the resulting knowledge base for uncovering and structuring the fundamental building blocks of Visual Analytics (VA) systems. We systematically extracted and organized components, operations, and their dependencies from a corpus of existing VA system research papers. The core of the approach involves a multi-level hierarchical structure (High Blocks, Intermediate Blocks, Granular Blocks) formalized into a blueprint (represented as JSON) that details system composition and data/interaction flows. This knowledge base, encompassing 101 systems, was constructed via initial manual analysis followed by Large Language Model (LLM) automation for scalable extraction. The goal is to provide a structured, queryable repository that reveals common design patterns and architectures, thereby offering a practical foundation to support more structured, reproducible, and efficient VA system development, comparison, and understanding.