Pick the AI provider. Ollama and Custom run locally (no cloud account needed) — everything else needs an API key from that provider.
API key for the chosen provider. Stored in plain text, not hashed — it has to be readable, not just comparable.
The local server's address, e.g. http://localhost:1234/v1 for LM Studio or http://localhost:11434/v1 for Ollama's default port.
Model id for the chosen provider. Picked from the model registry (run bin/rails ruby_llm:load_models to populate it) when it has entries for that provider — for Ollama/Custom, which aren't in the registry, type it by hand.