infrastructure-agent Example Repository

This example repository includes demo code for:

• LangFlow, OpenSearch, Ollama (with Granite 4) deployed on AWS EKS
• Agents using Agent2Agent protocol and HashiCorp Vault as an OIDC provider

Set up infrastructure

Go to the /terraform directory.

Set up credentials for AWS.

Run terraform init, terraform plan, and terraform apply.

Note: This will create an EKS cluster in auto-mode and HCP Vault cluster. It was originally deployed with HCP Terraform.

Go to the /kubernetes directory.

Run kubectl apply -f auto-mode.yaml. This will set up the rest of auto-mode.

LangFlow

Update build.sh to the correct ECR repository.

Build the images and push them to ECR.

bash build.sh

Go to /kubernetes and deploy the following:

kubectl apply -f ollama.yaml
kubectl apply -f langflow-ingress.yaml
kubectl apply -f terraform-mcp-server.yaml

Go to /helm and deploy the following:

bash langflow.sh
bash opensearch.sh

This creates a set of pods for LangFlow, Ollama, OpenSearch, and the Terraform MCP Server.

Agent authorization with Vault

Update agents/build.sh to the correct ECR repository.

Build the images and push them to ECR.

cd agents
bash build.sh

Go to the /kubernetes directory.

Deploy the following:

kubectl apply -f helloworld-agent.yaml
kubectl apply -f test-client.yaml

If you don't want to deploy the test-client to Kubernetes, you can run it locally with:

OPENID_CONNECT_PROVIDER_NAME=agent OPENID_CONNECT_CLIENT_NAME=agent AGENT_URL=${KUBERNETES_INGRESS_FOR_HELLOWORLD} python agents/test-client/__main__.py