LLM providers and persona services
A persona needs a model to run on, and a model comes from a provider such as OpenAI, Anthropic or Google. You connect a provider once, as a persona service, and then choose that service and one of its models for each persona.
A persona service is a connection to one provider, holding the key that provider gave you. A template for the provider knows how to talk to it and how to present the key, so you supply the key and nothing else.
The list of models each provider offers is refreshed from the provider from time to time. A refresh that fails leaves the previous list in place rather than emptying it.
Both are managed under Personas.