Should I start with an online or local model connection?
Compare where online and local model connections run, how they use credentials, and whether the desktop app must stay open.
Maintained by the 路墨 Lumoswitch product team
Online configurations run through Lumoswitch's service; local LiteLLM connects directly to a provider and requires the desktop app while in use.
How do the two connection types differ?
Online model configurations use a Lumoswitch-hosted runtime. The local “This device” option runs bundled LiteLLM in the desktop app and sends requests directly to the provider you configure. Keep Lumoswitch running while using a local connection; provider keys are encrypted with system secure storage and stay on this device. The two paths differ in configuration limits, runtime requirements, and who bills for requests.
How should a new user decide?
If you already have an available online model or Credits in Lumoswitch, start online and check authorization and model status in the app. If you want to connect your own provider account directly and can keep the desktop app running, create a local configuration under Models → This device. The same model name does not guarantee the same capabilities or price on both paths.
Does browsing the model tabs change my active connection?
Browsing online or local model tabs does not switch the agent's active model or rewrite agent data. To apply a connection, select and confirm it in Agent configuration. Reopen the agent afterward and check its model status or make a safe trial request.