dsh-tool-eyes
Run the following command in DeepSeek Harness:
dsh plugin install go-farther-and-farther/dsh-tool-eyes
Paste the following prompt into your AI chat to install this plugin:
Run the command 'dsh plugin install go-farther-and-farther/dsh-tool-eyes' to install the plugin from the source code at https://github.com/go-farther-and-farther/dsh-tool-eyes.
About this plugin
DeepSeek's chat models are inherently text-only, which limits their ability to process visual information. dsh-tool-eyes is designed for DeepSeek Harness to solve the problem of text agents lacking "visual perception." It equips your AI with "eyes" through two powerful local tools: a "screen" tool that converts screenshots or image files into text descriptions via a local OpenAI-compatible vision endpoint (like Ollama or llama.cpp), and an "ocr" tool that extracts text from the screen using Windows' built-in engine with millisecond speed, requiring zero models, zero GPU, and zero cloud dependencies.
The core advantage of this plugin lies in "privacy by default" and "low cost." It eliminates the need to upload images to the cloud, analyzing data securely on your machine by pointing to a local vision model interface. Furthermore, simple screen descriptions can be handled by local 0.8B–4B models, while OCR is virtually free. This makes it an excellent, lightweight solution for Windows users who want to let their text AI read screen content, process documents, or assist with development without the overhead of cloud APIs.
Use Cases
- Read error messages or pop-ups from the screen.
- Describe local image files or the current screen content.
- Perform vision analysis in a fully local environment without the cloud.
Best For
- Users needing visual capabilities for DeepSeek.
- Developers prioritizing privacy and keeping data local.
- Workers deploying LLMs on the Windows platform.
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.