dsh-tool-vision-read

In AI systems, text-only models are often limited by their inability to "see" images directly, making it difficult to process visual content. dsh-tool-vision-read is a lightweight plugin designed to solve this specific pain point. Its core capability lies in seamlessly delegating complex image description tasks to a dedicated vision model route. Supporting both "direct mode" and "subagent mode," it provides immediate text descriptions while also allowing for flexible, complex image analysis workflows—without manual forwarding or reliance on third-party CLI tools. This plugin is ideal for developers or teams who have a vision-capable model route but wish to keep their text models lightweight. It empowers text agents to "see" the world just like a human would, significantly expanding the boundaries of existing workflows with minimal configuration overhead.