Local AI voice assistant stack for Home Assistant (GPU-accelerated) with persistent memory, follow-up conversation, and Ollama model recommendations - settings designed for low VRAM systems.
