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

Terry12138qy/dsh-vision

30Last commit Aug 14, 2026

dsh-vision DSH plugin

dsh-vision is a bundle plugin for DeepSeek Harness that enables models without native vision capabilities to understand images. It accepts local file paths or network URLs, sends them to a primary multi-modal model (DashScope qwen3.5-omni-plus) with automatic fallback to a backup model (Zhipu glm-4.6v-flash) on failure. The tool is registered as `vision` and can be called in Code Mode.

visionskills-workflows

How to install the dsh-vision DSH plugin

dsh plugin --profile web add github:Terry12138qy/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

  • Accepts both local image file paths and network URLs as input.
  • Automatically switches to a backup model when the primary model fails (quota exhausted, overdue, or rate-limited).
  • Integrates seamlessly with DeepSeek Harness Code Mode via `await tools.vision(...)`.
  • Supports cancellation via `exec.signal` as required by DSH tool contract.

Where the dsh-vision DSH plugin fits

  • Let a non-vision DSH model describe the contents of an image in Chinese.
  • Analyze screenshots or photos uploaded to the DSH chat interface.
  • Extract textual information from images (e.g., signs, documents) in a DSH workflow.
  • Use as a fallback vision tool when the base model lacks multi-modal support.

Who the dsh-vision DSH plugin is for

  • DeepSeek Harness users who need image understanding capabilities.
  • Developers building DSH plugins who want to integrate vision functionality.

dsh-vision DSH plugin limitations

  • Relies on external API services (DashScope and Zhipu) – requires valid API keys.
  • Only supports Chinese language output for image descriptions.
  • Fallback model may introduce additional latency on failure.
  • The plugin is in early development; no specific image format restrictions are documented.

dsh-vision DSH plugin: from the repository README

Quoted from the Terry12138qy/dsh-vision README, the upstream source of the dsh-vision DSH plugin. Copyright remains with the original authors.

> 给不具备原生识图能力的 [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) 模型装上「眼睛」。 把图片（本地路径或网络 URL）发送给多模态大模型，返回中文文字描述。主模型为阿里云百炼（DashScope）`qwen3.5-omni-plus`，调用失败（额度用尽 / 欠费 / 限流）时自动切换智谱 `glm-4.6v-flash` 备用模型。 ## 原始来源与适配说明 - **原始项目**：[asuojun/claude-vision-skill](https://github.com/asuojun/claude-vision-skill) —— 面向 Claude Code 的识图 skill，核心是 `vision.js`（读图 → base64 → 调 OpenAI 兼容接口 → 返回文字描述，主模型失败自动切备用模型）。 - **本项目**：专门为 **DeepSeek Harness** 做了适配，把原来的 Claude skill 重写为 DSH 的 **bundle 插件**，注册一个 `vision` 工具。识图主流程（读图 → base64 → 调 API → 返回文字、主/备模型自动切换）**原样保留**。 相较原版的主要改动： - 打包成 DSH bundle（`dsh.bundle.patch`），可通过 `dsh plugin` 安装； - 注册为 DSH 的 `vision` 工具（Code Mode 里 `await tools.vision(...)` 可直接调用）； - CJS → ESM；配置改为「环境变量 > cordis.patch.yml config > 默认值」（原 `.env` 语义不变）； - 增加 `exec.signal` 取消支持（DSH 工具契约要求）。 ## 安装 ```sh dsh plugin --profile web add github:Terry12138qy/dsh-vision ``` 或本地路径： ```sh dsh plugin --profile web add file:/绝对/路径/dsh-vision ``` 安装后**重启** `dsh web` 生效。发布到 GitHub 后，记得给仓库添加 [`dsh-plugin`](https://github.com/topics/dsh-plugin) 主题，方便被 DeepSeek Harness 生态发现。 ## 配置 密钥通过**环境变量**（推荐，避免密钥进仓库）或编辑 `cordis.patch.yml` 的 `config`： | 用途 | 环境变量 | config 字段 | | ---- |

Read the full READMERepository license: MIT

dsh-vision DSH plugin questions

How do I install dsh-vision?

Run `dsh plugin --profile web add github:Terry12138qy/dsh-vision` and restart `dsh web`. You can also use a local path: `dsh plugin --profile web add file:/absolute/path/dsh-vision`.

What API keys do I need to configure?

You need at least a DashScope API key for the primary model (set `DASHSCOPE_API_KEY` environment variable). Optionally, set a fallback API key for Zhipu (`FALLBACK_API_KEY`) to enable automatic failover. All configuration variables are listed in the README.

Can I use a different model as the primary vision model?

Yes. You can set the `VISION_MODEL` environment variable or `model` config field to any model supported by DashScope’s OpenAI-compatible endpoint. The default is `qwen3.5-omni-plus`.

What happens if the primary model fails?

The plugin automatically switches to the backup model (default: Zhipu glm-4.6v-flash) and retries the request. You need to configure `FALLBACK_API_KEY` and related variables for the fallback to work.

Does the plugin support image URLs from the internet?

Yes, you can pass a network URL directly. The plugin fetches the image, converts it to base64, and sends it to the API. Local file paths are also supported.