dsh-vision-recognizer
Run the following command in DeepSeek Harness:
dsh plugin install kaixinbaba/dsh-vision-recognizer
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install kaixinbaba/dsh-vision-recognizer in your terminal to install the plugin; the source is at https://github.com/kaixinbaba/dsh-vision-recognizer .
About this plugin
DeepSeek Harness conversations cannot natively accept image attachments — there is simply no entry point to paste a screenshot or a chart and ask the model to analyze it. This plugin registers an adaptive vision route on the existing conversation pipeline so you can attach images to any DeepSeek conversation.
The core mechanism is adaptive routing: if the selected conversation model declares native image input, image blocks are passed through unchanged; if it is text-only or of unknown capability, a vision engine you configure transcribes the image into text. Fifteen-plus providers are supported out of the box — OpenAI, Anthropic Claude, Google Gemini, OpenRouter, Ollama (local), Alibaba DashScope, Zhipu GLM, Moonshot Kimi, and more. All settings take effect immediately from the Settings → Plugins → Vision panel with no restart required. A zero-config local Ollama path keeps image data entirely on your machine.
Ideal for users who rely on DeepSeek for everyday conversation and occasionally need to analyze screenshots, product images, or documents; and for developers who prefer local inference, value data privacy, or need to switch among multiple domestic or international vision providers.
Use Cases
- Paste screenshots or charts into a DeepSeek conversation for analysis
- Infer images locally via Ollama without data leaving your machine
- Switch between multiple domestic and international vision APIs
Best For
- Daily DeepSeek users who occasionally need image analysis
- Developers who prefer local inference and value data privacy
- Technical users who need flexibility across multiple vision providers
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.