Favio8/dsh-plugin-deepeye
41Last commit Aug 14, 2026
dsh-plugin-deepeye DSH plugin
DeepEye Vision integrates directly into ctx.tools without MCP overhead. It supports multiple backends including OpenAI, Google Gemini, and custom OpenAI-compatible endpoints. It also handles clipboard screenshot analysis and pasted image translation in pure-text model sessions.
How to install the dsh-plugin-deepeye DSH plugin
dsh plugin --profile web add dsh-plugin-deepeyeCopying does not run this command. Review the repository and version before installing the dsh-plugin-deepeye DSH plugin.
dsh-plugin-deepeye DSH plugin data source
dsh-plugin-deepeye DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-plugin-deepeye DSH plugin can do
- Multi-backend support: OpenAI, Gemini, and custom endpoints
- Smart image preprocessing: auto-resize and JPEG conversion to save tokens
- Result caching with LRU cache to reduce API calls
- Paste compatibility: automatically translate pasted images into text for pure-text models
- Five vision tools: describe, OCR, visual QA, UI layout analysis, clipboard analysis
Where the dsh-plugin-deepeye DSH plugin fits
- Describe the content of an image in detail
- Extract text from images (OCR)
- Answer questions about an image
- Analyze UI layout structure from a screenshot
- Quickly analyze clipboard screenshots without saving
Who the dsh-plugin-deepeye DSH plugin is for
- Users of DeepSeek Harness who want to add vision capabilities to their chatbots
- Developers integrating vision features into DSH-based applications
dsh-plugin-deepeye DSH plugin limitations
- Requires external vision API services (OpenAI, Gemini, etc.) and API keys
- Paste compatibility only works for pure-text models and requires a compatible vision backend
- Image processing may incur additional API costs and latency
- Limited to still images; no video support
dsh-plugin-deepeye DSH plugin: from the repository README
Quoted from the Favio8/dsh-plugin-deepeye README, the upstream source of the dsh-plugin-deepeye DSH plugin. Copyright remains with the original authors.
为 [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) 提供视觉能力的原生 Cordis 插件。 让纯文本模型获得"眼睛"：图片描述、OCR 文字提取、视觉问答、UI 布局分析、剪贴板截图分析，以及**粘贴图片自动翻译**（在纯文本模型会话里直接粘贴图片也能发）。 ## 特性 - **原生集成** — 直接注册到 `ctx.tools`，无 MCP 中间层开销 - **多后端** — 支持 OpenAI (GPT-4o)、Google Gemini、自定义 OpenAI-compatible 端点 - **智能预处理** — 自动缩放过大图片、转换 JPEG 以节省 token - **结果缓存** — LRU 缓存减少重复 API 调用 - **System Prompt** — 自动注入提示段落，让模型知道何时使用视觉能力 - **粘贴图片兼容** — 纯文本模型（如 DeepSeek）会话中直接粘贴图片，自动翻译成文字后交给模型（见下文 pasteCompat） ## 安装 本插件是一个标准 dsh **bundle**（声明了 `dsh.bundle.patch`），通过 dsh 的 profile 插件机制安装： ```bash # 发布后（npm 安装） dsh plugin --profile web add dsh-plugin-deepeye # 本地开发时（从插件源码目录的上一级执行） dsh plugin --profile web add ./dsh-plugin-deepeye ``` 安装时 dsh 会： 1. 首次使用自动初始化 profile（含 `@deepseek-ai/dsh-base`） 2. 用 pnpm 把本包链接进 profile 目录 3. 因本包声明了 `dsh.bundle`，自动追加到 `dsh.profile.bundles` 层叠 验证与启动： ```bash dsh --profile web --dump-config # 确认出现 dsh-plugin-deepeye 层 dsh web # web 是 --profile web 的别名 ``` ## 配置 ### API Key 解析优先级 插件支持多层 API Key 解析，用户可以选择最方便的方式： 1. **cordis.yml 显式配置** — `config.apiKey: !!js process.env.XXX` 2. **自动环境变量回退** — 根据 `provider` 自动匹配对应环境变量 3. **通用变量** — `DEEPEYE_API_KEY`（适
Read the full READMEThe repository declares no license. Check with the authors before using it.
dsh-plugin-deepeye DSH plugin questions
How do I install DeepEye Vision for DSH?
Install via DSH profile plugin mechanism: `dsh plugin --profile web add dsh-plugin-deepeye`. This will automatically link the package and add it to the profile bundles. After installation, start with `dsh web`.
What API keys do I need and how to configure them?
You need an API key from a supported provider. You can set it in cordis.yml under config.apiKey, or use environment variables. For OpenAI, set OPENAI_API_KEY; for Gemini, set GEMINI_API_KEY; for custom endpoints, set DEEPEYE_API_KEY. The plugin automatically falls back to environment variables if apiKey is not specified.
Does it support free vision models?
Yes, you can use the free model from Zhipu AI (glm-4v-flash) via the custom provider. Configure provider: custom, baseUrl: https://open.bigmodel.cn/api/paas/v4, model: glm-4v-flash. You need a Zhipu API key set as DEEPEYE_API_KEY.
What is paste compatibility and how does it work?
Paste compatibility allows pasted images in pure-text model sessions to be automatically translated into text descriptions. The plugin modifies the model info to bypass image support checks, then translates the image via a vision backend and replays the request. It has three modes: off, auto (default, only for models without image support), and force.
What tools does the plugin provide?
It provides five vision tools: vision_describe (detailed image description), vision_ocr (text extraction), vision_ask (answer questions about image), vision_layout (UI layout analysis), and vision_clipboard (analyze clipboard screenshot). All tools require an image_source parameter except clipboard which reads from clipboard.