TokenSpace is a database schema for storing and navigating meaning—not just documents. It combines four layers: content (docs/chunks + embeddings), token (words/forms with learned senses and real text instances), cognition (conversations, reflections, memories), and a lattice that links everything together with typed edges, multi-scale clusters (“cells”), neighbor caches, and lightweight dynamics (activations, decay/reinforcement). Configurable constants (e.g., the golden ratio Φ) and optional spiral/toroidal coordinates let us overlay simple geometry on top of vector space. The aim is to mimic a living lattice: ideas organize themselves locally, form larger patterns over time, and strengthen or fade based on use—so the system can move from static storage to an evolving memory.