dsh-open-eyes

When using DeepSeek Harness, the main model may lack native image processing capabilities, leaving users unable to derive visual insights from attached images. dsh-open-eyes addresses this limitation by adding configurable vision support, enabling the delegation of image analysis to a multimodal model of the user's choice, such as OpenAI or Anthropic models. This allows visual tasks to be seamlessly integrated into conversations, with results returned as text for the main model to utilize, making images a part of the dialogue flow.

The plugin's core strengths lie in its flexibility and ease of use. Users can create multiple vision schemes in the settings, selecting different API endpoints and models, and switch the default scheme at any time. It also offers visual analysis preferences, such as focusing on text OCR, tables and charts, or interface layouts, to tailor the analysis process. All insights are returned as text, maintaining conversation continuity, while supporting local image files and controlled remote URLs.

dsh-open-eyes is ideal for a wide range of users, particularly those who rely on DeepSeek Harness but need image processing that their main model cannot provide. Developers testing image recognition features, researchers analyzing chart data, or everyday users seeking to extract information from photos will all find value in this plugin. It prioritizes privacy and security, with API keys managed through DSH credentials, local image processing by default, and remote URLs disabled unless enabled, ensuring data safety.