The OffloadingConnector is a KV cache connector that offloads KV cache blocks from GPU memory to another medium (e.g., CPU memory or disk) and loads them back when needed. This enables prefix caching across a larger capacity than GPU memory alone.