What it does
Low confidence switches to a hard-coded “opus” model missing from the default list. The configured fallback handles missing Choice answers, not failed requests; the sample does not measure routing quality.
Integrations
Jev Model Router
This WebAssembly sample asks Jev which model should answer a prompt, then forwards that prompt to Ollama.
Bookmark: Jev Model Router
Keep this in your collection.
Leave a noteWhat would you try with this?
: Jev Model Router
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
How you can use it
Borrow the two-step design when your app can choose between several AI models. Ask your developer to list the models, ask Jev which should answer the user’s prompt, then send it to that model. Try prompts from your own app and check whether the choices make sense.
OpenRouter sends prompts to Jev for the choice; Ollama runs the chosen model and answers. Your developer must connect both services using their access keys. Ollama may run elsewhere, not on your computer. The sample chooses “opus” when unsure, although its default list lacks that model. Failed choice requests return an error.