dsh-vision-fallback：在 text-only 主模型里保留聊天图片
dsh-vision-fallback 是 DSH 的一款 MIT 许可插件，由 1HelloMan1 维护。它旨在解决主模型仅支持文本（text-only）但聊天框包含图片的落差，通过插件化方式在请求层自动将图片交给固定的视觉模型处理，转换为隐藏文本上下文返回给主模型，同时保留 UI 中的原图。该插件兼容 DSH 0.1.2-alpha.3 至 0.1.2-alpha.5 版本，要求 Node 环境不低于 22.13.0。 其核心功能包括：自动处理 MCP/ACP 持久化附件及 PTC 嵌套图片并保持顺序；支持通过 Web 设置页或 settings.yaml 配置视觉模型、Base URL
Read Moredeepseek-vl-support：为 DeepSeek 文本模型接入视觉的 DSH 插件
`deepseek-vl-support` 是一个 MIT 许可的插件，旨在解决 DeepSeek 文本模型无法直接读取图片的痛点。该插件通过拦截图片读取请求，将其发送至选定的 OpenAI 兼容视觉服务（如 OpenRouter、SiliconFlow、DashScope 等），并将返回的文字描述以固定格式回传给模型，从而实现图片理解。 该工具支持多种 Agent 环境，包括 Claude Code、Codex、DeepSeek Harness、Cursor、VS Code 等，并提供一次性安装向导、按 Agent 安装命令及 `doctor` 健康检查功能。用户只需具备 Node.js 1
Read Moredsh-free-vision: Enabling the DSH Plain Text Model to Read Screenshots, Errors, and UIs
dsh-free-vision is a free vision plugin developed specifically for DeepSeek Harness (DSH), designed to endow pure text models with image understanding capabilities. The plugin implements in-process startup through the built-in visual engine luma-mcp, eliminating the need for additional MCP service configuration, and addresses needs such as screenshot analysis, OCR, UI interpretation, and document recognition. Its core advantages include zero configuration, support for multiple free vision model providers such as Qwen, Volcengine, and SiliconFlow, and the ability to customize API addresses to adapt to different network environments. The plugin provides a single universal tool, `image_understand`, supports various task modes (e.g., OCR, debugging, description), and automatically optimizes the processing of large images. Users can quickly install it via the `dsh` command, restart the system, and then utilize the settings.
Read Moredsh-vision: Adding Image Understanding to DeepSeek Harness's Plain-Text DeepSeek
dsh-vision is a DeepSeek Harness plugin developed by 54xee, designed to add image understanding capabilities to the pure text model DeepSeek. It wraps adapters to convert images into text placeholders and backfills the recognition results for model processing. The plugin defaults to using the Doubao Web channel, requiring only a browser login, with no API key needed and zero cost. It also supports multiple access methods, including Reverse Gravity IDE, Gemini API, Cockpit reverse proxy, aicode direct connection, and any IDE CLI. Core features include visual evidence memory, content hash caching, automatic tier upgrades, and four task modes. Installation is simple, configuration is flexible, and it is suitable for developers who wish to add visual capabilities to DSH at a low cost or those who already have relevant IDE/API quotas. Please note that you need to check...
Read More