dsh-vision
Run the following command in DeepSeek Harness:
dsh plugin install kaaaahn/dsh-vision
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install kaaaaahn/dsh-vision in your terminal to install; the source is at https://github.com/kaaaaahn/dsh-vision. Restart DSH after installation to activate.
About this plugin
Pasting a screenshot directly into the DSH chat and having the AI read and analyze it on the spot is the core problem dsh-vision solves. Getting an AI to look at images typically means sending them to a cloud API, letting sensitive content leave your machine — this plugin keeps the entire vision pipeline local on your Mac, so images never leave your device.
It works on two levels. The lower layer calls the macOS Vision framework for OCR, returning both the recognized text and pixel-level coordinates so you can point to exact UI elements. The upper layer runs a Qwen3-VL model via Ollama to interpret layout, element relationships, and anomalous regions at a semantic level. Setup is truly zero-config: after installation it auto-detects your RAM, selects the right model size (2B / 4B / 8B), and installs Ollama if it is missing. While the larger model downloads in the background (1.8 to 5.7 GB), OCR is already fully usable, so you can start working immediately.
If you are on macOS and regularly need an AI to read screenshots, extract UI copy, or diagnose error dialogs — while keeping every pixel on your local machine and skipping cloud API fees — dsh-vision is built for you.
Use Cases
- Paste a screenshot with error text and let the AI extract words with pixel-level UI coordinates
- Analyze interface layout and anomalous regions entirely on-device
- Have the AI read a local screenshot file and produce a semantic description during development
Best For
- macOS developers who need AI help reading screenshots and UI copy
- Privacy-conscious users who want images to never leave their machine
- Users who prefer zero-cost, cloud-free visual analysis
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.