dsh-omniroute-models

DSH's built-in model page offers no search or filtering, and an OmniRoute gateway can expose hundreds of models without returning modality metadata—picking the right few is an eyeball exercise. dsh-omniroute-models pulls OmniRoute's /v1/models into DSH's settings panel, mapping input modalities, context window, and max output tokens, and gives you multi-dimensional filtering by vendor namespace, DSH route, and text/image modality alongside instant search. Hit Save Selected and the whitelist writes into llm-pi-ai config immediately; the next request picks it up with no restart, and any sibling fields you hand-edited in settings.yaml stay untouched. Only chat-capable models appear in the list—embedding, video-generation, and image-generation entries are filtered out automatically. The plugin also bridges DSH's web_search tool to OmniRoute's aggregate search backend (Tavily, Brave, Exa, and more), requiring no extra API key on the gateway side. If you run DSH behind an OmniRoute gateway juggling a large upstream model roster, or you want vision models and web search managed from a single settings page, this plugin was built for exactly that.