mirror of
https://github.com/multica-ai/multica.git
synced 2026-07-25 12:05:06 +02:00
46b91587ae
The agent profile hover card (AgentProfileCard) previously omitted the model. Add a Model meta row (raw id, matching the inspector) and, when the agent has a non-empty thinking_level override, a thinking-level chip next to it. The chip reuses the skills-chip visual and resolves the runtime-native token to its friendly catalog label (e.g. "xhigh" -> "Extra high") via the same RuntimeModel.thinking catalog the inspector's ThinkingPicker uses. The model catalog is only fetched when there is an override to resolve AND the runtime is online, so agents with no override (the common case, where no chip renders) never trigger daemon model discovery from this passive hover surface. Cold cache / offline degrades to the raw token, which is the accepted behavior. Adds the `profile_card.model_label` i18n key across en/zh-Hans/ja/ko and a regression test covering label resolution, the raw-token fallback, the no-override case, and the default-model case. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Co-authored-by: multica-agent <github@multica.ai>