dsh-model-info-hint

Juggling a dozen-plus models across multiple providers in DSH comes with a small, recurring annoyance: the model picker shows you a name, and to confirm what input modalities it accepts, how large its context window is, what the max output tokens are, which API endpoint it talks to, or which environment variable carries the auth key, you end up switching to a terminal, grepping through YAML, and switching back. That back-and-forth breaks your flow, especially when you are comparing two or three models side by side.

dsh-model-info-hint collapses that lookup into a single hover. Rest the cursor on any model entry in the selector—the model-button menu next to the input box, or the /model dialog—and a non-blocking tooltip appears right beside it: input types, context window, max output tokens, reasoning level, API protocol name, baseURL, and the authentication environment-variable name. The hint surfaces only fields that are explicitly declared in your settings document, so unconfigured entries are not padded in; apiKeyEnv shows the variable name, never the secret itself. The tooltip disappears the instant you move the cursor away, click, scroll, press any key, or after six seconds, whichever comes first, and pointer events pass through cleanly, so it never blocks your next action. Data is read through the built-in settings service of DSH, supporting both .yaml and .json formats, with $DSH_HOME overrides and hot-reload working out of the box.

This is aimed at DSH users who frequently switch between models for comparison or debugging and would prefer a glanceable configuration summary over yet another tab full of YAML. As long as your model routes live in a settings document (any registered namespace, not limited to llm-pi-ai), the plugin picks them up automatically with no extra configuration needed.