Evidence snapshot reviewed Sep 10, 2026GitHub checked Aug 21, 2026
Evidence-verifiedPlugin BundleSearch, Vision & DataWeb Profile

DSH Image Gen

An image-creation bundle for DeepSeek Harness with cloud providers and local ComfyUI workflows.

At a glance

What it does

An image-creation bundle for DeepSeek Harness with cloud providers and local ComfyUI workflows.

Use cases
Search, Vision & DataUIDataIntegrations
Works with
Deepseek HarnessGoogle GeminiOpenai ImagesSeedreamAliyun DashscopeComfyui
Compatibility

Web Profile
Not declared in supplied evidence

Trust & status

Evidence-verified
Checked Sep 6, 2026, 1:49 PM UTC

Code-evidenced contributions

What it adds to DSH

Model ToolsImage generation and editing

Adds chat-based image generation, image editing, reference-image workflows, and provider selection to DeepSeek Harness.

Mechanism evidence
Web UIImage Studio and gallery

Provides batch creation, multi-model comparison, a prompt inspiration library, and gallery management in the web client.

Mechanism evidence

Before you choose it

Generate or edit images in DeepSeek Harness chat, use a Studio for batch creation and cloud-model comparisons, browse a 500+ prompt library, and manage saved images in a gallery. It supports Google Gemini, OpenAI Images or compatible endpoints, Seedream, DashScope, and configured local ComfyUI workflows.

Best for

DeepSeek Harness Web users who want BYOK image generation, editing, or comparison workflows inside their existing chat environment.

Common tasks

  • Create an image from a chat prompt or revise an existing image with natural-language instructions.
  • Generate several candidates in Studio using reference images and adjustable model settings.
  • Compare configured cloud models with the same prompt and references.
  • Run named local ComfyUI API-format workflows from chat.

Permissions and data

Uses provider credentials and may send generation inputs externally; it can also retain image-related data locally.

Permissions
  • Adds a web-client plugin layer to a DeepSeek Harness Web profile.
  • Uses API keys saved through DSH Credentials for configured cloud providers.
  • Can connect to the ComfyUI URL that you configure.
Data handling
  • Sends the prompt and reference images for a generation request to the selected cloud provider.
  • When workspace saving is enabled, chat results are written to the workspace; Studio saves only results the user selects.
  • Gallery metadata and favorites are stored in browser local storage; inspiration images are loaded on demand and cached locally.
External services
  • Google Gemini
  • OpenAI Images or compatible endpoints
  • ByteDance Seedream
  • Aliyun DashScope
  • A user-configured ComfyUI server
Credentials
  • A user-supplied API key is required for a selected cloud provider.
  • Local ComfyUI requires a DSH-host-reachable URL and an imported API-format workflow JSON.

Limitations

  • Cloud-provider terms, availability, pricing, and output behavior are outside this bundle's supplied evidence.
  • Studio and multi-model comparison support cloud providers only; local ComfyUI is not supported in those views.
  • ComfyUI editing is documented as single-image editing.
  • Deleting local image files may be irreversible.

What DSHub checked

  • Pinned Git source, bundle manifest, and patch structure were verified.
  • The manifest declares Node.js ^22.19.0 or >=24.0.0 and a DeepSeek Harness Web client integration.
  • The supplied documentation describes supported providers, BYOK configuration, storage behavior, and installation steps.
  • The supplied license is MIT.

What DSHub did not check

  • Installation, startup, provider authentication, image generation, and rendered output were not executed for this record.
  • No DeepSeek Harness version range is declared in the supplied evidence.
  • The npm package contents were not audited, although its registry identity was verified.

Pinned install

Install DSH Image Gen

This plugin bundle does not have a DSH Plugin install action. Use its source documentation for the delivery method.

Visit the source project

Maintainer source

Project README

View at commit d6aacb3
Maintainer-authored contentCaptured from README.md on Sep 6, 2026. The text and repository-relative media are fixed to commit d6aacb3109b6 with content hash 66a0bc2c5b4c; provider-hosted badges may update independently. README commands are upstream documentation; the DSHub copy action above is the verified, version-pinned install.
<div align="center"><img src="docs/assets/readme/hero-zh.webp" alt="dsh-image-gen 中文功能概览" width="100%" /><br /><p><strong>简体中文</strong> · <a href="README.en.md">English</a></p>

🎨 dsh-image-gen

DeepSeek Harness 的原生 AI 图像创作套件

<p><b>对话生图与编辑 · Studio 批量创作 · 多模型对比 · 500+ Prompt 灵感 · 图库管理 · 本地 ComfyUI</b></p><p> <a href="https://www.npmjs.com/package/dsh-image-gen"><img src="https://img.shields.io/npm/v/dsh-image-gen?style=flat-square&color=4f6ef7" alt="npm version" /></a> <a href="https://www.npmjs.com/package/dsh-image-gen"><img src="https://img.shields.io/npm/dm/dsh-image-gen?style=flat-square&color=10b981" alt="npm downloads" /></a> <a href="https://github.com/shanliuling/dsh-image-gen/actions/workflows/ci.yml"><img src="https://github.com/shanliuling/dsh-image-gen/actions/workflows/ci.yml/badge.svg" alt="CI" /></a> <a href="https://github.com/shanliuling/dsh-image-gen/stargazers"><img src="https://img.shields.io/github/stars/shanliuling/dsh-image-gen?style=flat-square" alt="GitHub stars" /></a> <a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-f5c542?style=flat-square" alt="License: MIT" /></a> <a href="https://linux.do/"><img src="https://img.shields.io/badge/LINUX%20DO-社区友链-555?style=flat-square" alt="LINUX DO" /></a> </p><p> <a href="#快速开始">快速开始</a> · <a href="#核心能力">核心能力</a> · <a href="#provider-支持情况">Provider 支持</a> · <a href="#常见问题">常见问题</a> </p><br /></div>

为 DeepSeek Harness 带来完整的 AI 图像创作工作流。

dsh-image-gen 不只是简单的对话生图,而是为 DSH 补齐了从自然语言连续修图Studio 批量创作多模型横向对比,到 500+ Prompt 灵感库本地 ComfyUI 的全流程能力。支持 Google Gemini、OpenAI Images / Compatible、ByteDance Seedream、Aliyun DashScope 及本地私有化工作流,采用 BYOK(自带 Key)模式,生成结果支持按工作区隔离存储。

pnpm dsh plugin --profile web add dsh-image-gen@latest

版本更新提示: 本次版本变化较大,老用户请更新至最新版本。

<img src="docs/assets/readme/workflow-overview.webp" alt="dsh-image-gen 完整 AI 图像创作工作流" width="100%" />

一个插件,四种创作方式

入口 最适合 你可以做什么
💬 对话 让 Agent 理解自然语言需求 文生图、图生图、多图参考、连续编辑、原位重新生成
🎛️ 工作台 精确控制创作参数 批量生成、多张参考图、高级画布
灵感 找构图、风格与 Prompt 500+ 案例、搜索筛选、收藏、复制、一键带入工作台
🖼️ 图库 整理与复用生成结果 搜索、筛选、收藏、重新生成、下载、批量管理、工作区隔离

快速开始

1. 安装插件

环境要求:DeepSeek Harness 稳定版本,Node.js ^22.19.0>= 24.0.0

在你的 DeepSeek Harness 项目根目录下运行:

pnpm dsh plugin --profile web add dsh-image-gen@latest

💬 极客提示:你也可以直接把这句话发送给 DSH 对话中的 Agent:<br /> 帮我安装生图插件,在终端执行:pnpm dsh plugin --profile web add dsh-image-gen@latest

<details> <summary><strong>其他安装方式(全局 / GitHub 直装 / 本地调试)</strong></summary>
# 若已将 dsh 安装为系统全局命令:
dsh plugin --profile web add dsh-image-gen@latest

# 从 GitHub 仓库直接安装最新代码:
pnpm dsh plugin --profile web add git+https://github.com/shanliuling/dsh-image-gen.git

# 本地克隆源码开发安装:
git clone https://github.com/shanliuling/dsh-image-gen.git
pnpm dsh plugin --profile web add ./dsh-image-gen
</details>

2. 配置 Provider

重启 DSH 后进入:

设置 → 插件 → 插件配置 → 图像生成

选择 Provider,填写自己的 API Key,并按需调整模型、Endpoint / Base URL 与工作区保存选项。使用 ComfyUI 时,请填写 DSH Host 可访问的服务地址,并导入 API Format Workflow JSON

3. 开始创作

在聊天框中直接描述你想要的图片:

画一张雨夜霓虹街头的赛博朋克猫咪,电影感光线,16:9。

也可以直接上传参考图,让 Agent 进行风格重构或局部编辑:

保持角色与构图不变,给猫咪戴上一副黑色墨镜。
<br /><div align="center"> <img src="docs/assets/readme/provider-settings.webp" alt="DSH 插件配置界面" width="46%" /> <img src="docs/assets/readme/chat-example.webp" alt="对话生图与风格重构效果" width="46%" /> <br /> <sub>左:Provider 配置 · 右:在 DSH 对话中直接生图、图生图与连续编辑。</sub> </div><br />

需要更细的参数控制时,点击会话顶部的 画廊 入口,进入 图库 / 工作台 / 灵感 / 收藏


核心能力

💬 对话生图、编辑与版本切换

  • 用自然语言完成文生图、图生图、多图参考和风格迁移。
  • 直接修改原图 Prompt 重新生成,并在同一卡片中切换历史版本。
<br /><div align="center"> <img src="docs/assets/readme/regenerating.webp" alt="图片正在重新生成" width="46%" /> <img src="docs/assets/readme/revision-switcher.webp" alt="在同一图片卡片中切换生成版本" width="46%" /> <br /> <sub>修改 Prompt 后原位重新生成,并在同一张图片卡片中切换历史版本。</sub> </div><br />

🎛️ Studio 批量创作

  • 支持多张参考图,一次生成多张候选图。
  • 自由控制 Provider、模型、比例和清晰度,只保存满意的结果。
<br /><div align="center"> <img src="docs/assets/readme/studio-workbench.webp" alt="dsh-image-gen Studio 工作台" width="100%" /> <br /> <sub>在同一个 Studio 中完成参考图导入、参数控制、批量生成、结果筛选与保存。</sub> </div><br />

⚖️ 多模型横向对比

使用同一组 Prompt 和参考图并发调用多个模型,在一张画布中比较并保存结果。

<br /><div align="center"> <img src="docs/assets/readme/multi-model-compare.webp" alt="同一 Prompt 的多模型生成对比" width="100%" /> <br /> <sub>在同一画布中比较不同模型结果,再批量保存满意的图片。</sub> </div><br />

✨ 500+ Prompt 灵感案例

  • 浏览和筛选 500+ Prompt 案例,支持收藏、复制及一键带入 Studio。
  • 图片缓存在本地,浏览和学习不消耗 Token 或生成额度。
<br /><div align="center"> <img src="docs/assets/readme/inspiration-library.webp" alt="Prompt 灵感素材库" width="100%" /> <br /> <sub>先找灵感,再把 Prompt 带入工作台;全本地缓存,浏览或复制不消耗生成额度。</sub> </div><br />

🖼️ 图库、收藏与批量管理

  • 统一管理对话和 Studio 中保存的图片,并按工作区隔离。
  • 支持搜索、筛选、收藏、下载、继续编辑、重新生成和批量管理。
<br /><div align="center"> <img src="docs/assets/readme/gallery-management.webp" alt="图库筛选、收藏与批量管理" width="100%" /> <br /> <sub>图库支持多维度筛选、收藏、批量管理与工作区数据隔离。</sub> </div><br />

🧩 本地 ComfyUI 多工作流

灵活调用本地 GPU 算力,让私有化绘图无缝融入 Agent 对话。

  • 导入并管理多个命名工作流,支持预设 Prompt 和常用占位符。
  • Agent 可按名称选择工作流,在对话中完成文生图和图生图。

ComfyUI 暂未接入 Studio 和多模型对比。

<br /><div align="center"> <img src="docs/assets/readme/comfyui-workflows.webp" alt="ComfyUI 多工作流配置" width="58%" /> <br /> <sub>为不同用途维护独立工作流,并通过名称让 Agent 精确选择。</sub> </div><br />

Provider 支持情况

Provider 对话生图 对话编辑 Studio 多模型对比
Google Gemini ✅ 多图
OpenAI Images / Compatible ✅ 多图
ByteDance Seedream / 火山方舟 ✅ 多图
Aliyun DashScope / Qwen Image ✅ 多图
Local ComfyUI ✅ 单图

Studio 与多模型对比目前只支持云端 Provider;多模型对比调用的是各 Provider 在设置中已配置的模型。

<details> <summary><strong>当前默认模型与 Endpoint(均可修改)</strong></summary>
Provider 默认模型 默认 Endpoint / Base URL
Google Gemini gemini-3.1-flash-image https://generativelanguage.googleapis.com/v1beta/interactions
OpenAI Images gpt-image-2 https://api.openai.com/v1
OpenAI Compatible 自定义 自定义 Base URL
ByteDance Seedream doubao-seedream-5-0-260128 https://ark.cn-beijing.volces.com/api/v3
Aliyun DashScope qwen-image-3.0 https://dashscope.aliyuncs.com/api/v1
Local ComfyUI 用户导入的 API Workflow http://127.0.0.1:8188
</details>

数据与隐私

  • BYOK:API Key 通过 DSH Credentials 服务保存,设置页不会回显 Key 明文。
  • 云端请求:Prompt 与本次使用的参考图会发送给所选 Provider,请遵守对应服务条款。
  • 本地 ComfyUI:请求发送到用户配置的 ComfyUI 地址。
  • 工作区文件:开启工作区保存后,对话结果会落盘;Studio 仅保存用户选中的候选图。
  • 图库与收藏:图库元数据和收藏状态保存在当前浏览器本地存储中。
  • 灵感缓存:案例元数据随插件提供,图片按需加载并缓存在本机,可随时清理。

常见问题

<details> <summary><strong>安装后找不到“图像生成”设置怎么办?</strong></summary>

先完全重启当前 DSH Profile,再检查插件配置:

dsh --profile web --dump-config

如果输出中没有 dsh-image-gen,请重新执行安装命令。提交 Issue 时请附 DSH 版本、插件版本和错误日志,不要上传 API Key。

</details><details> <summary><strong>生成图片保存在哪里?</strong></summary>

开启“保存到工作区”后,对话生成结果默认保存在当前工作区的 dsh-image-gen/ 子目录,也可以在设置中修改。Studio 候选图先留在临时画布,只有用户选中的结果才会进入图库并保存。

</details><details> <summary><strong>为什么 ComfyUI 没有出现在 Studio 中?</strong></summary>

当前 Studio 与多模型对比只支持四类云端 Provider。ComfyUI 已支持在 Agent 对话中进行文生图、单图编辑以及多个命名工作流。

</details><details> <summary><strong>从图库删除会删除聊天记录吗?</strong></summary>

不会。删除图库记录不会修改原聊天消息。你可以额外选择是否清理工作区中的本地图片文件;文件删除通常无法恢复,请确认后操作。

</details><details> <summary><strong>如何升级?</strong></summary>
pnpm dsh plugin --profile web add dsh-image-gen@latest

升级后重启对应的 DSH Profile。

</details>

本地开发

git clone https://github.com/shanliuling/dsh-image-gen.git
cd dsh-image-gen

pnpm install
pnpm run typecheck
pnpm test
pnpm run build
pnpm run pack:check

欢迎通过 Issues 反馈问题,或阅读 CONTRIBUTING.md 后提交 Pull Request。

License

本项目基于 MIT License 开源。

<div align="center">

如果 dsh-image-gen 对你的工作流有所帮助,欢迎在 GitHub 点亮一颗 ⭐ Star 支持持续维护。

查看 Releases · 提交 Issue · 参与贡献

</div>

Operate deliberately

Install and manage

Prerequisites and target Profile

Target Web Profile

Delivery Dsh Bundle Git — shanliuling/dsh-image-gen#d6aacb3109b638559669944bcc15e28e33c708cd

Verify, update, and remove

Show lifecycle commands
Verify
dsh plugin --profile web list

Compatibility and access

DeepSeek Harness Web profile; Node.js 22.19+ or 24+ Not declared in supplied evidence

Review compatibility evidence

Risk facts

Credentials

Requires your own provider API key for cloud image generation.

Evidence
External Data Transfer

Prompts and selected reference images are sent to the chosen cloud provider; ComfyUI requests go to the configured URL.

Evidence
Local Storage

Selected results and gallery metadata may be stored in the workspace or browser local storage.

Evidence
Evidence and editorial reviewManifest, Bundle patch, distribution and freshness

Immutable evidence

Review status and source activity

AI reviewed

Use a non-sensitive test prompt first. Do not paste API keys into chat, logs, or issue reports.

AI reviewed Sep 10, 2026, 11:36 AM UTCGitHub facts last checked Sep 10, 2026, 11:36 AM UTC

No material source change has been recorded since this evidence baseline.

Next step

Follow the Plugin installation workflow

Subscribe to material changes for DSH Image Gen