pub fn prime_free_model_cache()Expand description
Prime the free-model cache synchronously at startup.
Called from Router::new() via spawn_blocking so the initial
blocking subprocess doesn’t stall the Tokio runtime.
Runs run_opencode_models_discovery_async() in a dedicated thread with its
own tokio runtime, avoiding block_on restrictions when called from within
an existing runtime (e.g. #[tokio::test] or a nested runtime context).