Server always runs in router mode: every [section] in models.ini is served ([*] = global defaults: model, alias, n-gpu-layers, ctx-size...). Save, then restart llama.cpp above to apply.
no task running