Add a region setting for Mistral models with three options:
• Global — api.mistral.ai
• European Union — api.eu.mistral.ai
• United States — api.us.mistral.ai
We currently use Mistral’s global endpoint, which does not guarantee where inference is processed. Explicit EU or US selection would support data-location and latency requirements, subject to regional model/feature availability and the 10% pricing premium.[mistral]
We should first confirm that our model proxy can preserve explicit regional routing without falling back to the global endpoint.
