Architecture-first AI runtime that separates intelligence from language models. Jarvis Runtime treats LLMs as replaceable language engines, not system brains. It provides stable behavior, persistent memory, and long-term context continuity outside the model layer. Designed for developers building production-grade assistants, not prompt experiments.