dsh-model-switcher

When juggling multiple LLM providers in the DeepSeek Harness Web GUI, switching models traditionally means opening settings, verifying each API key, and manually picking a target—a tedious loop that pulls you out of flow. dsh-model-switcher collapses that into a single click: a small toggle in the session header opens a two-level cascading panel where you pick a provider first, then a model, and the switch is done instantly.

Under the hood, the plugin replaces the native model selector in the input bar and shares the same directory-state controller as the /model command and the composer model slot, so a switch made from any entry point syncs everywhere in real time. The panel only lists providers whose API keys are actually configured, always shows the current provider and model with a blue highlight at the top, surfaces clear error messages when a directory fails to load or a provider is unreachable, and gracefully disables the entry point in sub-agent sessions where model switching is not supported.

If you routinely hop between DeepSeek, OpenAI, local Ollama, and other providers while staying in the Web GUI rather than dropping to a terminal, this plugin is built for you. It runs entirely on the browser side with zero build step—drop it in and it simply works.