← Back to list

dsh-tool-visual-primitives

DeepSeek Harness Other Low risk

Analyzes conversation images through mode-specific prompts (caption, UI, document, grounding, topology, etc.) and injects structured evidence with coordinate primitives (boxes, points, refs) as text, with session-level caching for reuse across replay and compaction.

通过模式化提示分析对话图片(描述、UI、文档、定位、拓扑等),将含坐标原语(框、点、引用)的结构化证据以文本注入上下文,会话级缓存跨回放与压缩复用。

How to install

DeepSeek Harness dsh plugin add dsh-tool-visual-primitives

About

dsh-tool-visual-primitives 为 DeepSeek Harness(DSH) 的纯文本模型补充视觉能力。插件把图片交给外部视觉模型分析,再把带有空间定位信息的**纯文本视觉证据**交回原对话模型;因此,被增强的模型不需要原生图片输入能力。 核心思路参考 DeepSeek 的 Thinking with Visual Primitives:以归一化坐标和可引用对象,将图像理解转化为后续推理可使用、可检查的证据。 已发布至 npm。推荐使用下方的 DSH 官方一键安装命令;GitHub 源码挂载方式保留给开发与本地调试。 1.3.0 更新 兼容新版 DSH 宿主(0.1.1-rc.8 及以上)的适配器接口。 会话辅助调用(标题生成、上下文压缩)不再触发重复的视觉分析,降低配额消耗与偶发失败。 相同图片与提示词的并发分析自动合并为一次上游请求。 新增“诊断日志”设置项(默认关闭),可输出插件运行日志便于排查。 连接测试改进:使用真实 token 预算,失败时错误信息包含上游返回摘要。 安装命令更新为 dsh 直令形式(需已全局安装 DSH)。 1.2.0 更新 重构设置页面的信息层级与连接流程:已完成配置时默认折叠,并在展开后提供完整编辑与“保存并测试连接”。 改进视觉模型选择、模型目录读取、手动模型 ID、连接状态与错误反馈。 优化多轮图片对话:无图片的后续轮次…

Recommendation signals

51 Tool quality · Based on stars, downloads, maintenance, security and docs
User interest · Adjusted by in-site views, install copies and download clicks
51 Overall
0views
0unique visitors
0install copies
0download clicks
0outbound clicks

Meta

License
MIT
Language
JavaScript
GitHub stars
1
mo. downloads
Last push
2026-08-22
Created
2026-08-15

Links

Basic safety check

Findings
None
Sources
curated:awesome-dsh-plugin.com, curated:awesome-dsh-plugin/awesome-dsh-plugin
Topics
deepseek-harness, deepseek-harness-plugin, dsh, dsh-plugin, dsh-plugins, multimodal, plugin, vision, visual-primitives

Related plugins

DeepSeek HarnessClaude CodeCodex New
Score74

modlens

liustack/modlens

Vision bridge for text-only models: paste an image, get structured JSON evidence (OCR, layout, semantics).

☆ 3.7K ↓ 20.0K Other
DeepSeek Harness New
Score70

dsh-vision-router

ysr666/dsh-vision-router

Free vision for text-only agents: built-in keyless vision chain plus pixel tools (Q&A, grounding, crop, pixel diff, colors, OCR, SVG trace, cutout, screenshots); paste an image to use it.

☆ 1.0K ↓ 39.6K Other
DeepSeek Harness New
Score69

dsh-vision-toolkit

Anionex/dsh-vision-toolkit

Vision for text-only models: paste an image and the model switches to a Vision Toolkit variant for image Q&A, multi-image comparison, long-screenshot OCR, screenshot-to-UI reproduction, element grounding, and pixel diff. No API key by default — images are processed by the author-hosted free service, 100 per machine per day; configurable to your own provider.

☆ 837 ↓ 1.8K Other
DeepSeek Harness Featured
Score68

dsh-deepseek-vision

siegfly/dsh-deepseek-vision

A vision-language gateway provider route: pasted images are described by a configurable VL model (Qwen-VL by default) before the DeepSeek wire.

☆ 8 ↓ 3.4K Other