Let user select and configure a model in AI agent

[Hooghof]

As a user, I want to be able to configure a LLM provider and attach it to an Agent component, so that I can select and configure an available model that my component will use to generate a response

Acceptance criteria

1. I'm able to configure and select a LLM provider from a predefined set of defaults. For example OpenAI, Mistral or Groq.
2. When the credentials are configured and saved I get a clear message whether the credentials worked.
3. When picking a model I see the models that are available to me from that provider.
4. I can easily see which provider / model is used.
5. Each Agent can use its own provider/model configuration without affecting other Agents.
6. If no provider / model is configured then the Agent component stays invalid.
7. I'm able to add additional API options for example like:

• temperature
• max number of tokens
• thinking (if available)
  In my mind it's a mechanism that refreshes the auth, list the models with a /models call and then wraps the request and serves it to the AI providers /chat/completions API.

If we compare this with the n8n world:

• n8n has a concept of credentials that are created on a tenant level.
• The credential can then be selected in an agent component
• The agent component then loads all the available models
• A model can be selected and configured

I also added some screenshots of the n8n functionality.

Jira: https://dovetailworld.atlassian.net/browse/FG-2634