Skip to content
  • Home
  • Plugin manual
  • dsh-vision

oil-oil/dsh-vision

414Last commit Aug 14, 2026

dsh-vision DSH plugin

dsh-vision is a plugin for DeepSeek Harness that provides native image handling for vision-capable models and a bridge for text-only models. When the main model supports images, they are passed through directly; otherwise, a separate vision model analyzes the images and injects its output as context. It supports multiple cloud vision providers and local fallback via macOS Vision or Tesseract.

vision

How to install the dsh-vision DSH plugin

dsh plugin --profile web add github:oil-oil/dsh-vision

Copying does not run this command. Review the repository and version before installing the dsh-vision DSH plugin.

dsh-vision DSH plugin data source

dsh-vision DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-vision DSH plugin can do

  • Passes images natively to vision-capable models without preprocessing or OCR.
  • Bridges text-only models by forwarding images to a configured vision model and injecting its output as untrusted context.
  • Supports multiple image attachments simultaneously, enabling comparisons and combined evidence.
  • Configurable cloud vision providers: ZenMux, Alibaba Cloud Model Studio, TokenDance, OpenRouter.
  • Local fallback using macOS Vision OCR or Tesseract (Linux/Windows) when cloud is unavailable.
  • API keys stored securely via Harness credential service, write-only in the browser.
  • Vision output marked as untrusted observation data, preventing prompt injection from images.

Where the dsh-vision DSH plugin fits

  • Let a text-only model (e.g., deepseek-official) understand and describe images in a chat.
  • Compare multiple images by asking the model to analyze them together, e.g., find differences or similarities.
  • Use the plugin without changing the main model; the vision bridge activates automatically when needed.
  • Choose a preferred cloud vision provider or fall back to local OCR for privacy-sensitive scenarios.

Who the dsh-vision DSH plugin is for

  • DeepSeek Harness users who want to add image understanding to their workflows.
  • Users who prefer using a text-only model but occasionally need to process images.

dsh-vision DSH plugin limitations

  • Currently targets DeepSeek Harness 0.1.0-rc.6 exactly; compatibility with other versions is not guaranteed.
  • Local fallback is primarily OCR and does not provide full multimodal understanding (e.g., cannot interpret complex scenes like a vision model).

dsh-vision DSH plugin: from the repository README

Quoted from the oil-oil/dsh-vision README, the upstream source of the dsh-vision DSH plugin. Copyright remains with the original authors.

English | <a href="./README.zh.md">中文</a> </p> <p align="center"> <a href="https://github.com/oil-oil/dsh-vision/actions"><img alt="CI" src="https://img.shields.io/github/actions/workflow/status/oil-oil/dsh-vision/ci.yml?style=flat-square&label=CI"></a> <a href="./LICENSE"><img alt="MIT License" src="https://img.shields.io/badge/license-MIT-4D6BFE?style=flat-square"></a> <img alt="DeepSeek Harness" src="https://img.shields.io/badge/DeepSeek%20Harness-0.1.0--rc.6-4D6BFE?style=flat-square"> </p> `dsh-vision` is a plugin for DeepSeek Harness. Vision-capable models keep receiving images natively. When the selected main model is text-only, the plugin asks a separate vision model to observe the original images, then lets the original DeepSeek model produce the final answer. ## How it works | Main model | Image path | Final answer | | --- | --- | --- | | Supports images | Original images are sent directly, without preprocessing or OCR | Current model | | `deepseek-official` or another text-only model | A configured vision model observes the original images; its output is injected as untrusted attachment context | DeepSeek | | Cloud vision unavailable | Falls back to macOS Vision

Read the full READMERepository license: MIT

dsh-vision DSH plugin questions

How do I install dsh-vision?

Run the command `npx @deepseek-ai/dsh plugin --profile web add github:oil-oil/dsh-vision` in your terminal. Then restart DeepSeek Harness. The plugin will be automatically activated.

Does dsh-vision replace my main model?

No. The plugin does not replace the model selected in Harness. It only adds a vision bridge when the main model is text-only. For vision-capable models, images are passed through natively.

Which cloud vision providers are supported?

The plugin supports ZenMux, Alibaba Cloud Model Studio, TokenDance, and OpenRouter. You can configure the provider in the Vision Recognition card under Settings > Plugins > Plugin configuration.

Can I use the plugin without a cloud API key?

Yes, if you choose the "Automatic" option in Vision Recognition, the plugin will try image-capable models already configured in Harness, then fall back to local OCR (macOS Vision or Tesseract). However, local fallback is OCR only and not equivalent to full multimodal understanding.

Is my API key safe?

Yes. API keys are stored through Harness's official credential service. They are write-only in the browser: the plugin can report whether a key exists, but never reads it back into the page, chat, settings document, or session log.