Independent open-source plugins for DeepSeek Harness.

Hyper DSH Plugins develops independent, open-source plugins for DeepSeek Harness. Each project has a defined purpose and can be installed independently.

Use the multimodal model you choose to understand images in DeepSeek Harness.

Add screenshots, photos, charts, or interface images to a conversation. Open Eyes sends them to your configured vision model and returns the analysis to the same conversation. The main model can continue from there, even when it does not support images itself.

• Create multiple vision configurations and switch the active one from DSH settings.
• Connect OpenAI Responses, OpenAI Chat Completions, or Anthropic Messages endpoints.
• Set the analysis depth, focus areas, and optional guidance for tasks such as OCR, chart reading, interface review, and detailed inspection.

View repository · npm

Add files to a conversation and let the Agent work with their contents.

Select, drop, or paste multiple files into DSH Web. Open File stores them in the active session workspace, where the Agent can inspect their structure, read content, run OCR, or render document pages as needed.

• Attach any file type. Open File can read or process text, PDF, DOCX, PPTX, XLSX, ZIP archives, and common images, and retains metadata for other files.
• Run local OCR in English and Simplified Chinese.
• Preserve source references and reading positions for follow-up questions and verification.

View repository · npm

The projects are created and maintained by Hyp6666. Hyper DSH Plugins has no affiliation with or official endorsement from DeepSeek.

Hyper DSH Plugins 开发面向 DeepSeek Harness 的独立开源插件。每个项目都有明确用途，可按需单独安装。

在 DeepSeek Harness 中使用自己选择的多模态模型理解图片。

把截图、照片、图表或界面图片加入对话。Open Eyes 会将图片交给已配置的视觉模型，再把分析结果送回原对话。即使主模型本身不支持图片，也可以继续依据结果作答。

• 在 DSH 设置中创建多套视觉模型配置，并随时切换当前方案。
• 接入 OpenAI Responses、OpenAI Chat Completions 或 Anthropic Messages 接口。
• 可设置分析深度、关注重点与补充说明，适合 OCR、图表读取、界面检查和细节分析。

查看项目 · npm

把文件加入对话，让 Agent 直接读取和处理其中的内容。

可通过文件选择器、全页面拖放或剪贴板一次添加多个文件。Open File 会将文件保存到当前会话工作区，Agent 可按需检查结构、读取内容、执行 OCR 或渲染文档页面。

• 支持任意格式附件。Open File 可读取或处理文本、PDF、DOCX、PPTX、XLSX、ZIP 与常见图片，并保留其他文件的元数据。
• 支持英文与简体中文本地 OCR。
• 保留来源引用和读取位置，方便继续追问与核对。

查看项目 · npm

这些项目由 Hyp6666 创建并维护。Hyper DSH Plugins 与 DeepSeek 无隶属关系，也未获得官方背书。