NagasakiSoyo-ui/dsh-llm-deepseek-vision

“Vision-augmented DeepSeek adapter: a vision-capable model describes image input, then a text-only DeepSeek model reasons over the description.”