dsh-image-bridge

In DSH (DeepSeek Harness), pure text models traditionally cannot handle image inputs, limiting their application in visual tasks. The dsh-image-bridge plugin solves this by intelligently bridging images, automatically injecting image paths and processing instructions as text to let pure text models autonomously discover and invoke their visual skills (such as general QA, OCR, or grounding) to 'see' images. Its core capabilities include one-click sending, automatic adaptation to multimodal models, real-time sensing of model switches, and hands-off cleanup of temporary files—all transparent and zero-intervention for users. This is ideal for anyone using pure text models who needs to process image content, whether for visual analysis, document recognition, or multimodal experiments, seamlessly expanding model capabilities without relying on specific tools or complex configurations.