dsh-vision
Run the following command in DeepSeek Harness:
dsh plugin install 54xkeee/dsh-vision
Paste the following prompt into your AI chat to install this plugin:
In DeepSeek Harness, run the command dsh plugin install 54xkeee/dsh-vision to install the dsh-vision plugin. Source: https://github.com/54xkeee/dsh-vision
About this plugin
Vision has long been a challenge for pure-text LLMs like DeepSeek: traditional vision APIs involve costs and keys, while simply pasting images into chat is rejected. The dsh-vision plugin fundamentally solves this, acting as a smart adapter that gives DeepSeek 'eyes'. Its core capability lies in enabling image understanding through a default zero-cost Doubao Web channel (requiring just a browser login), and it offers multiple structured vision modes ranging from general understanding and OCR to region focusing and multi-image comparison. More powerfully, it introduces 'vision evidence memory', allowing recognition results to be reused and queried across conversation turns, paired with content-hashing caches to avoid redundant computations and automatic depth escalation for complex scenes. This plugin is ideal for both developers and everyday users looking to expand DeepSeek's vision capabilities without added costs or complex API configurations, especially those in WSL or network-restricted environments seeking a reliable vision solution.
Use Cases
- Adding image processing capabilities to the DeepSeek model
- Needing OCR to recognize text in screenshots or documents
- Performing comparative analysis and finding differences between multiple images
Best For
- Developers wanting to extend DeepSeek's vision capabilities at zero cost
- Users in restricted network environments (like WSL) needing a stable vision solution
- Users frequently performing image understanding, OCR, or visual comparison tasks
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.