Run 70B+ LLMs on Apple Silicon by using SSD as extended memory — intelligent layer streaming and caching for Mac
sovereign embedded vector database, single-file .nest container with content-addressable citations, reproducible builds, and offline-first model verification.
