dsh-image-gen

If you're building your own AI agent workflows on DeepSeek Harness, you're probably used to having the agent call various tools to get things done. But image generation and editing still feel like an external chore — you switch windows, copy images, and paste results back into the conversation. dsh-image-gen is built to remove that friction: it embeds text-to-image, image-to-image, and multi-turn iterative editing directly into the chat flow, so you can get images with a simple sentence and keep refining the same picture in context, just like you would with ChatGPT.\n\nAt its core, this plugin makes "conversation equal operation." The agent automatically calls generate_image and edit_image tools, returns images inline in the chat, and remembers uploaded images, previously generated pictures, and last-round edit results within the same session — enabling true multi-round iteration. It also ships with a dedicated gallery panel that aggregates all historical images with search, preview, copy, download, and delete actions. On the backend, it supports Google Gemini, OpenAI Images / Compatible, ByteDance Seedream, and Alibaba Cloud DashScope Qwen Image, all configurable directly in DSH settings with your own API keys, models, and endpoints — a genuine BYOK approach.\n\nIf you are a developer or content creator who heavily relies on DeepSeek Harness and wants both text and image generation under one private deployment, or if your team needs to integrate multiple image model providers into a single agent environment instead of maintaining separate tools, this plugin is for you. It also suits users who are comfortable with OpenAI-compatible APIs but want to experiment with Gemini, Seedream, or Qwen image models. After installation there is no extra learning curve — just describe what you want in the chat.