DSH plugin: text-only models (e.g. DeepSeek-V4) automatically see images via a vision model. Official surface-replace, cache-friendly, human transcript untouched. 纯文本模型自动识图桥
Automatic model routing for DeepSeek Harness: three difficulty tiers (hard/normal/easy) plus vision routing, picking models you already configured under Settings ? Models. ???? + ?????????