For mainstream consumer-grade GPUs with 8GB or 16GB of VRAM, the heavy-load tasks of a full ODS site configuration cannot run smoothly, but lightweight LLM conversations and basic image generation are still feasible, making deployment of core components only more practical