Interactive KV cache memory calculator for LLMs — supports MLA, GQA, hybrid attention, sliding window, and linear attention architectures. Estimate GPU memory for serving any model at any context length.