dsh-plugin-vision
Run the following command in DeepSeek Harness:
dsh plugin install MoneShadow/dsh-plugin-vision
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install MoneShadow/dsh-plugin-vision in your terminal to install; the plugin source is at https://github.com/MoneShadow/dsh-plugin-vision .
About this plugin
DeepSeek's flash and pro models are text-only. When an Agent needs to look at a screenshot, a design mock, or a table photo mid-conversation, it simply cannot see it. dsh-plugin-vision registers a vision_describe tool on the engine side so the main model can hand a local path, file:// URI, or http(s) URL to any OpenAI-compatible vision API and get back a structured text description.
A built-in answer cache keyed on image hash, endpoint, model, and prompt (with TTL and LRU cap) ensures that re-describing the same image hits the cache and costs no extra vision API call. Errors are never cached and the cache is fully transparent to the main model. Configuration lives in the vision section of settings.yaml and is hot-watched by the engine via chokidar, so changes take effect immediately without a restart. The API key is stored with secret role and is write-only in the official UI.
This plugin is for developers building Agent workflows on DeepSeek Harness who occasionally need to inspect images. It works across the official Web UI, DSH Desktop, and headless mode because the tool is registered at the engine layer. Even without the desktop app you can pass a file path or URL straight to vision_describe, and the core capability remains independent of any UI layer.
Screenshots
Use Cases
- Your Agent needs to inspect a local screenshot or design mock mid-conversation
- A text-only main model delegates image understanding to an external vision API
- Repeated lookups of the same image hit the cache and save vision API quota
Best For
- Developers building Agent workflows on DeepSeek Harness who occasionally need image support
- Teams running text-only DeepSeek flash/pro models without native vision
- Users who want to plug in any OpenAI-compatible vision service without vendor lock-in
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.