dsh-model-capabilities

The built-in Models page in DSH handles basic model management but deliberately leaves out two fields that matter a great deal for inference experience: input modalities (text / image) and per-model reasoning effort levels (reasoningEfforts). Changing them today means hand-editing settings.yaml line by line or hopping between every provider models array and its modelOverrides. dsh-model-capabilities pulls both into a dedicated Settings section so you manage everything in one place.

Open the Model Input & Reasoning panel to see every configured provider and its models, including modelOverrides. Toggle text or image modality per model, fill wire values for reasoning levels from minimal through max, or set a level to off so no reasoning parameter is sent. Writes go straight through the llm-pi-ai settings namespace, so the running adapter re-registers routes instantly, the model picker reflects new reasoning levels immediately, and the change persists to settings.yaml with no restart needed.

If you only need a quick input-type picker while adding a model, an inline field is the lighter choice. But if you also want reasoning wiring and a single overview of every configured model and override, this plugin was built for exactly that gap. Ideal for developers running multiple models on DSH Web or Desktop who care about inference depth and multimodal input.