dsh-vision
Run the following command in DeepSeek Harness:
dsh plugin install yepyeel/dsh-vision
Paste the following prompt into your AI chat to install this plugin:
To install the plugin in DeepSeek Harness, run the command dsh plugin install yepyeel/dsh-vision; the source code is available at https://github.com/yepyeel/dsh-vision.
About this plugin
When working with DeepSeek Harness, text-only models like DeepSeek V4 Flash or Pro often fail to process image attachments, leaving valuable visual information unutilized. The dsh-vision plugin addresses this by automatically describing images and replacing them with text before sending requests to the model, effectively giving text models a "vision" capability. This process is transparent to users: original images remain in session logs and the UI, while only the model call is modified.
The plugin's core strengths lie in its flexible vision model management. It offers auto mode, which tries available vision models in sequence, or a specified mode for a chosen model. If no vision models are available, it falls back to system OCR (leveraging macOS Vision, Windows OCR, or Tesseract) to extract text from images. To optimize performance, descriptions are cached per image, preventing redundant processing across conversation turns.
dsh-vision is ideal for developers and users of DeepSeek Harness who need to handle image content without switching to a dedicated vision model. Whether for ad-hoc image processing or building multimodal workflows, this plugin integrates seamlessly, enabling the use of image data with existing text models.
Use Cases
- When text models need to process image attachments
- Automatically describing images for model understanding in conversations
- Using OCR to extract text information from images
Best For
- Developers using DeepSeek Harness
- Researchers needing to expand text model image processing capabilities
- Content creators handling multimodal content
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.