Last released Nov 3, 2026
Unified LLM KV-Cache (-58.2% VRAM) & Vector DB / RAG Embedding (-57.3% RAM) Acceleration SDK for PyTorch & Hugging Face.