ollama - 💡(How to fix) Fix Turbo migration persists invalid cloud model names [1 pull requests]

What is the issue?

The app migration for legacy Turbo settings constructs cloud model names by concatenating cloud directly onto selectedModel:

const cloudModel = `${settings.selectedModel}cloud`;

For gpt-oss:20b, this produces gpt-oss:20bcloud.

The frontend treats that string as cloud because its cloud checks use endsWith("cloud"), then persists SelectedModel: "gpt-oss:20bcloud" and clears TurboEnabled.

The backend model reference parser recognizes explicit source tags like gpt-oss:20b:cloud or legacy tag suffixes like gpt-oss:20b-cloud. It does not recognize gpt-oss:20bcloud, so subsequent requests use an invalid or nonexistent local model instead of cloud.

Steps to reproduce

1. Start with app settings where TurboEnabled is true and SelectedModel is gpt-oss:20b. The same issue applies to the other base models in the Turbo migration list.
2. Load the app with cloud enabled.
3. Observe that the selected model is migrated to gpt-oss:20bcloud and persisted while TurboEnabled is set to false.
4. Send a request with the migrated selected model.

Expected behavior

The migration should preserve explicit cloud source intent, for example gpt-oss:20b:cloud, which the backend already understands and normalizes to gpt-oss:20b-cloud where legacy pull names are needed.

Additional context

The backend already tests gpt-oss:20b:cloud and gpt-oss:20b-cloud. Repo history and docs use delimited cloud names, not gpt-oss:20bcloud.