The bot itaelf is running on a vps. The model generation part is based on sinking.ai because need ram and vram.