dsh-model-management

Once you wire up multiple OpenAI-compatible endpoints in DSH, the composer model picker and /model command quickly fill with models from every provider, yet DSH ships no built-in way to hide them in bulk or switch the default model from a single panel. dsh-model-management adds a Model Management section to DSH Settings that groups every registered model by provider, giving you one place to manage all of them.

Three capabilities cover the everyday pain points. Visibility control: hide a single model or close an entire provider and they disappear from both the model picker and the /model command; restore them anytime, and the model currently selected in your active session stays visible so you never lose routing mid-conversation. Default routing: pick a default provider, model, and reasoning effort once, and subsequent calls target that model automatically. Search follow-through: when the default model belongs to a configured openai-responses provider, the plugin registers a model-management-openai-responses web-search provider, so web search tracks your current default model and endpoint with zero extra setup.

The plugin is built for developers running multiple OpenAI-compatible endpoints side by side in a DSH Web Profile—DeepSeek, OpenAI, a self-hosted vLLM, or any mix. If you want the editor to show only the models you actually need and want web search to follow your default model without manual reconfiguration, dsh-model-management consolidates those scattered actions into one intuitive settings page. The project is still in early access, so verify model-picker compatibility after each DSH upgrade.