Support for Anthropic/AzureAI models

[peterj]

This PR updates the ModelConfig CRD and adds support for Anthropic and Azure OpenAI (still needs to be tested). I've also updated the Agent CRD to reference the ModelConfig + other changes that were necessary to dynamically create the model in autogen based on the ModelConfig resource.

There are also bunch of changes in the UI to support the backend changes.

Here's how to test this out:

1. Create the secret wit the API key:

kubectl create secret generic kagent-anthropic -n kagent --from-literal ANTHROPIC_API_KEY=<key>

2. Apply the ModelConfig referencing the secret and setting the provider to Anthropic:

apiVersion: kagent.dev/v1alpha1
kind: ModelConfig
metadata:
  name: claude-model-config
  namespace: kagent
spec:
  apiKeySecretName: kagent-anthropic
  apiKeySecretKey: ANTHROPIC_API_KEY
  model: claude-3-sonnet-20240229
  provider: Anthropic

Similar for Azure:

1. Secret first:

kubectl create secret generic kagent-azureopenai -n kagent --from-literal AZURE_OPENAI_API_KEY=<key>

2. Then the modelconfig with azure-specific config:

apiVersion: kagent.dev/v1alpha1
kind: ModelConfig
metadata:
  name: azureopenai-model-config
  namespace: kagent
spec:
  apiKeySecretName: kagent-azureopenai
  apiKeySecretKey: AZURE_OPENAI_API_KEY
  model: gpt-4o-mini
  provider: AzureOpenAI
  azureOpenAI:
    azureEndpoint: "https://<endpointname>.openai.azure.com/"
    apiVersion: "2025-03-01-preview"
    azureDeployment: "gpt-4o-mini"
    azureAdToken: <token>

closes: #174

[EItanya]

My main comment on this PR has to do with the use of map[string]interface{}. Totally makes sense why these would look like a good option coming from a JS/Python world. But typically in go you should create a type to represent it when using it within the code. map[string]interface{] is typically used to represent JSON you don't know the format of, but we know the format of all of the JSON in this case.

[Jmadih]

Hello,
Maybe in the future, you could check out LiteLLM (litellm.ai) for managing all LLMs in the OpenAI format.

[EItanya]

Looking better, but still have the same major comment about the API

[peterj]

Looking better, but still have the same major comment about the API

I think I address all feedback with the latest commit.