Configuring Google Vertex AI models in kagent

Documentation

Configuring Google Vertex AI

Google Vertex AI is supported for Gemini and Anthropic models.

Create the Google Application Default Credentials file and store it in a Kubernetes Secret:

kubectl create secret generic kagent-google-creds -n kagent --from-file=./google_creds.json

For Gemini models: create a ModelConfig resource using the GeminiVertexAI provider that references the secret and key name, and specify the Gemini model you want to use. Note the projectID and location are required:

apiVersion: kagent.dev/v1alpha1
kind: ModelConfig
metadata:
  name: gemini-model-config-vertexai
  namespace: kagent
spec:
  apiKeySecretRef: kagent-google-creds
  apiKeySecretKey: google_creds.json
  model: gemini-2.0-flash-lite
  provider: GeminiVertexAI
  geminiVertexAI:
    projectID: kagent-dev
    location: us-west1
    maxOutputTokens: 1000

For Anthropic models: create a ModelConfig resource using the AnthropicVertexAI provider that references the secret and key name, and specify the Anthropic model you want to use. Note the projectID and location are required:

apiVersion: kagent.dev/v1alpha1
kind: ModelConfig
metadata:
  name: anthropic-model-config-vertexai
  namespace: kagent
spec:
  apiKeySecretRef: kagent-google-creds
  apiKeySecretKey: google_creds.json
  model: claude-sonnet-4@20250514
  provider: AnthropicVertexAI
  anthropicVertexAI:
    projectID: kagent-dev
    location: us-east5