给 DeepSeek Harness 纯文本模型装上原生视觉（Windows）：粘贴即看图——预注入描述，模型首轮就看见，不用选模型、不用调工具；see_image 精查；自定义视觉后端（任意 OpenAI 兼容模型）+ 四后端容灾；换主模型视觉自动跟随。| Give text-only DeepSeek Harness models native-feeling vision on Windows: paste and the model just sees it — pre-injected descriptions, see_image tool, custom backends, 4-backend failover.
plugin windows vision dsh ai-agents multimodal cordis llm image-description image-paste deepseek-harness text-only-model see-image
-
Updated
Aug 15, 2026 - JavaScript