简体中文 · English

🎨 dsh-image-gen

Native AI image creation suite for DeepSeek Harness

AI 创作画布 · 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 · 订阅免 Key

快速开始 · 核心能力 · Provider 支持 · 常见问题

<sub>左：在画布中绘制草稿、摆放参考图，并通过对话生成成图。 · 右：基于已有结果继续对话修改，在画布中持续迭代创作。

左：在对话中直接描述并生成图片。 · 右：工作台批量创作、灵感库与图库等更多功能。</sub>

Deliver a complete AI image creation workflow to DeepSeek Harness.

dsh-image-gen is far more than simple conversational image generation — it gives DSH a full image creation capability set spanning conversational generation & iterative refinement, the AI creation canvas, Studio batch creation, multi-model side-by-side comparison, the Prompt inspiration library, and local ComfyUI.

Supports mainstream cloud image models and local private workflows. Works with BYOK (Bring Your Own Key) or subscription accounts; results can be stored per workspace.

Already have a ChatGPT, Grok, or Google subscription? Log in directly to generate images — no extra API Key purchase needed.

Supported: Gemini · OpenAI / Compatible · Seedream · DashScope · Grok Imagine · GLM-Image · Local ComfyUI

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

Version update notice: Major changes in this release — please update to the latest version.

One plugin, covering the complete AI image creation workflow

Quick Start

1. Install the Plugin

Environment requirements: DeepSeek Harness stable release, Node.js ^22.19.0 or >= 24.0.0.

Run the following in the root directory of your DeepSeek Harness project:

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

💬 Geek tip: You can also send this sentence directly to the Agent in your DSH chat:
帮我安装生图插件，在终端执行：pnpm dsh plugin --profile web add dsh-image-gen@latest

# 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

2. Configure a Provider

After restarting DSH, go to:

Settings → Plugins → Plugin Configuration → Image Generation

Select a Provider, enter your own API Key, and adjust the model, Endpoint / Base URL, and workspace save options as needed. After filling in the Key, you can click "Test Connection" to verify availability, or click "Fetch Models" to retrieve all supported image-generation models for that vendor with one click — no need to look up documentation manually. When using ComfyUI, please provide a service address accessible from the DSH Host and import the API Format Workflow JSON.

Already have a ChatGPT, Grok, or Google subscription? No API Key needed: expand the corresponding subscription Provider row, click "Log in" and complete authorization in the browser, then start text-to-image and image-to-image generation directly.

3. Start Creating

Describe the image you want in the chat box:

画一张雨夜霓虹街头的赛博朋克猫咪，电影感光线，16:9。

You can also upload reference images for style reconstruction or partial editing:

保持角色与构图不变，给猫咪戴上一副黑色墨镜。

_{左：Provider 配置 · 右：在 DSH 对话中直接生图、图生图与连续编辑。}

When you need finer parameter control, click the Gallery entry at the top of the conversation to open Gallery / Studio / Inspiration / Favorites.

Core Capabilities

💬 Chat-based generation, editing & version switching

• Natural language: text-to-image, image-to-image, multi-reference, style transfer.
• Edit the Prompt and regenerate; switch versions in the same card.

_{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。}

✏️ From sketch to final image: AI creation canvas

Express ideas on an infinite canvas; conversation turns sketches into images.

• Draw sketches, add references, and organize ideas freely on the canvas.
• Chat with the AI to turn sketches into complete works quickly.
• Iterate on existing results with continuous editing, modification, and new directions, preserving the creative process.

_{左：在画布中绘制草稿、摆放参考图，并通过对话生成成图。 · 右：基于已有结果继续对话修改，在画布中持续迭代创作。}

🎛️ Studio batch creation

• Multiple reference images supported; generate multiple candidates at once.
• Control Provider, model, aspect ratio, and sharpness freely — keep only what you like.

_{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。}

⚖️ Multi-model side-by-side comparison

Call multiple models concurrently with the same Prompts and references; compare and save on one canvas.

_{在同一画布中比较不同模型结果，再批量保存满意的图片。}

✨ 500+ Prompt inspiration examples

• Browse and filter 500+ Prompt examples, with support for favorites, copying, and one-click import into Studio.
• Images are cached locally; browsing and learning do not consume Tokens or generation quota.

_{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。}

🖼️ Gallery, favorites, and batch management

• Unified management of images saved from chat and Studio, isolated by workspace.
• Supports search, filter, favorite, download, continued editing, regeneration, and batch management.

_{图库支持多维度筛选、收藏、批量管理与工作区数据隔离。}

🧩 Local ComfyUI multi-workflow

Leverage local GPU compute to integrate private image generation seamlessly into Agent conversations.

• Import and manage multiple named workflows with preset Prompts and common placeholders.
• The Agent selects workflows by name to complete text-to-image and image-to-image in conversations.

ComfyUI is not yet integrated with Studio and multi-model comparison.

_{为不同用途维护独立工作流，并通过名称让 Agent 精确选择。}

Provider Support

Studio and multi-model comparison currently only support cloud Providers (including subscription channels); multi-model comparison calls the models already configured for each Provider in Settings. Zhipu GLM-Image's upstream itself does not support image-to-image; xAI's image-to-image uses the OpenAI-compatible protocol (multipart), and some gateways may need to wait for future adaptation. Subscription channels work via account login (no API Key needed), supporting conversational text-to-image / image-to-image, Studio batch generation, and multi-model comparison; image-to-image uses each subscription channel's editing API, with channel-default parameters (up to 5 reference images).

Data & Privacy

• BYOK: API Keys are stored via the DSH Credentials service; the settings page does not display the Key in plaintext.
• Subscription login: Subscription account tokens are saved in DSH Credentials after your authorized login, fully isolated from API Keys; the browser side does not touch the tokens.
• Cloud requests: Prompts and the reference images used in the current session are sent to the selected Provider; please comply with the corresponding terms of service.
• Local ComfyUI: Requests are sent to the user-configured ComfyUI address.
• Workspace files: When workspace saving is enabled, conversational results are written to disk; Studio only saves candidate images selected by the user.
• Gallery & favorites: Gallery metadata and favorite status are stored in the current browser's local storage.
• Inspiration cache: Example metadata ships with the plugin; images are loaded on demand and cached locally, and can be cleared at any time.

FAQ

dsh --profile web --dump-config

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

Local Development

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

We welcome bug reports via Issues, or read CONTRIBUTING.md and then submit a Pull Request.

License

This project is open-sourced under the MIT License.