Description

After initial setup, the model picker in the WebUI is effectively limited to OpenRouter and the provider configured during onboarding. Users who have multiple providers configured in ~/.hermes/config.yaml (Anthropic, Gemini, Groq, Mistral, custom endpoints, etc.) expect to be able to select models from all of them — matching what the Hermes CLI supports.

Two related gaps:

1. Post-setup model management is limited

There's no way to add a new provider, change API keys, or configure additional providers from within the WebUI after onboarding. Users have to edit config.yaml manually. The Settings panel should expose this.

See also #586 (allow provider key update) which covers part of this.

2. No API key validation before model selection

When a user selects a model that requires an API key they haven't configured, the selection silently fails or errors on the first message. The model picker should check whether a valid key/credential exists for the selected provider before allowing selection, and show a clear inline prompt to configure it if not.

Proposed

• Surface all configured providers (matching hermes-agent's full provider list) in the model picker
• Settings panel: add/remove/update providers and keys post-onboarding
• Model picker: grey out or flag models whose provider has no configured key, with an inline "Add API key" action