ReasoningDetails records which reasoning levels a model accepts, but not its default behavior when a request omits the level (whether it reasons by default, and at which level).
Several sources report that default:
Ollama: thinking.default from /api/show.
OpenRouter: default_effort and default_enabled on /api/v1/models.
Hosted catalogs, from the provider docs: Gemini 4 Neon and Hydrogen default to high, Carbon and Argon to medium.
Recording the default lets JP determine whether auto (reasoning enabled at a model-selected level) requires an explicit enable flag for a given model, and lets JP display the effective reasoning level for a request.
Populate both from Ollama, OpenRouter (making default_enabled optional in jp_openrouter), and the hosted catalogs.
Use the default enabled state when a provider determines whether auto requires an explicit enable flag.