dsh-image-generation

The official image-generation tool in DeepSeek Harness registers an image_generate call but ships no settings page, so users cannot manage multiple providers or swap models without code-level edits. Meanwhile dsh-plugin-subscriptions is tied to ChatGPT and Grok subscription logins and offers no path for API Key + Base URL gateways. dsh-image-generation fills that gap by combining a multi-provider management UI with a single-model conversation tool, making gateway-based image generation as straightforward as configuring a chat model.

The workflow has three layers. Under Settings → Image generation you add as many providers as you need, each with a name, Base URL, API format (OpenAI Images, xAI Images, Gemini / Imagen, or Chat Completions), an API key, and a list of image models; keys are stored through the credentials service and never written to settings.yaml. Under Settings → Plugins → Plugin configuration → Image generation you select exactly one model as the active conversation generator and optionally toggle the tool, set a default size, and choose a quality level. Inside a chat the Agent simply calls image_generate whenever a picture is needed; the result is saved to the session workspace under generate/image/ and rendered inline in the message stream, clickable to enlarge. Optional parameters cover prompt, four size presets, three quality levels, and up to five reference images for edit-style workflows.

If your pipeline routes through OpenAI, xAI, Gemini, or a local-compatible gateway via API keys and you want the same central management experience you already enjoy for chat models, this plugin is built for you. Once installed, the standalone dsh-tool-image-generation or the subscription-based dsh-plugin-subscriptions are usually unnecessary.