dsh-image-gen: A DSH plugin that generates images from plain chat in DeepSeek Harness
hiocobueno653-pixel/dsh-image-gen
A DSH plugin that brings image generation into DeepSeek Harness: just talk and get an image, with automatic prompt refinement and a persistent turn-tail image card.
dsh-image-gen is a DSH plugin that embeds image generation directly into the DeepSeek Harness chat flow. It targets two annoyances: the model writing long polishing text around a generation request, and finished images being folded away inside the process group. Prompts are optimized inside the tool, while images land in a card at the end of each turn and are rebuilt from the session log after a refresh or on another device. No provider is bundled; you supply the endpoint, model, and key, with multiple image models switchable, under the MIT license.
How to Install
dsh plugin --profile web add github:hiocobueno653-pixel/dsh-image-gen- Category
- Tools & Capabilities
- Platform
- DSH Plugin
- Author
- hiocobueno653-pixel
- Distribution
- Plugin
dsh-image-gen Key Features
dsh-image-gen Repo Summary
What Does It Do?
It is a DSH plugin for DeepSeek Harness that brings image generation into the chat flow: you simply say "draw a picture of..." and it calls a swappable image API in the background, with no slash command required. It solves two problems — chatty generation prompts and images that get folded away — by optimizing the prompt inside the tool so the model does not write long polishing text, and by rendering the finished image in a card at the end of each turn. The plugin bundles no provider, endpoint, or key; you supply your own, and it is maintained by hiocobueno653-pixel under the MIT license.
Core Features
- Command-free generation: just describe what you want; the model extracts size tier and aspect ratio from your words, defaulting to 2K and 1:1.
- GridReveal animation: the tool card plays a grid-reveal animation while generating, then swaps it for the finished image.
- Turn-tail image card: images render below the closing text and outside the process group, so they are never collapsed and are rebuilt from the session log after a refresh or on another device.
- Provider-agnostic: uses the image generation endpoint by default, can switch to an OpenAI-compatible shape, and supports multiple configured image models with a switchable active one.
- Layered key lookup: keys are resolved from the page entry, the entry's environment variable, then the plugin-wide environment variable, and are read only from environment variables or local config rather than shipped with the repo.
How to Use This Plugin?
After enabling it in DSH, fill in the two required fields — the image API base URL and the model name — either in the plugin configuration or via Settings · Image Models by clicking "+ Add Image Model" and completing the form, including the key. Then simply say something like "draw a picture: an orange cat on a windowsill in afternoon sunlight" or "make me a product poster, a glass cube in a white studio, 16:9", and the plugin generates the image and shows it in the card at the end of that turn. If your service is OpenAI-compatible, select "OpenAI compatible" as the interface type.
This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: hiocobueno653-pixel/dsh-image-gen. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.
DSH Plugin FAQ
