Skip to main content

EVA Agent Installation

The EVA Agent is the core component responsible for intelligent inference and data processing within the EVA service. Please follow the steps below to proceed with the configuration.


🛠️ Prerequisites

Before starting the installation, ensure that the following tools are ready:

  • Kubernetes Cluster: A cluster where EVA will be deployed.
  • kubectl: Go to Installation Guide
  • Helm: Go to Installation Guide
  • AWS CLI Configuration: AWS credentials are required for ECR image access. Run the following command in your terminal to configure:
aws configure
# Enter your AWS Access Key ID, Secret Access Key, etc., as prompted.



🚀 Installing EVA Agent

Step 1: Register and Update Helm Repository

Add and update the repository to fetch the latest version of the charts.

helm repo add eva-agent https://mellerikat.github.io/eva-agent
helm repo update



Step 2: Download Configuration (values) Templates

Download the necessary template files for your environment. Choose the files that match your setup (AWS or K3s).

# 1. Download common basic settings and secret templates
curl -L "https://raw.githubusercontent.com/mellerikat/eva-agent/chartmuseum/release/2.3-a3.0/1/eva-agent/values.yaml" -o values.yaml
curl -L "https://raw.githubusercontent.com/mellerikat/eva-agent/chartmuseum/release/2.3-a3.0/1/eva-agent/secret-values.yaml.tpl" -o values-secret.yaml

# 2. For Cloud (AWS) environment
curl -L "https://raw.githubusercontent.com/mellerikat/eva-agent/chartmuseum/release/2.3-a3.0/1/eva-agent/values-aws.yaml" -o values-aws.yaml

# 3. For On-premise (K3s) environment
curl -L "https://raw.githubusercontent.com/mellerikat/eva-agent/chartmuseum/release/2.3-a3.0/1/eva-agent/values-k3s.yaml" -o values-k3s.yaml



Step 3: Update Settings for Your Environment

Open the values.yaml and values-secret.yaml files to modify the key environment variables.

CategoryNameDescriptionOn-premise (K3s)Cloud (AWS)
Runtimeimage.tagImage tag"2.3-a3.0""2.3-a3.0"
RuntimereplicaCountNumber of replicas11
RuntimenodeSelectorNode selectoreks.amazonaws.com/nodegroup: ng-an2-eva-agent
Env Varenv.MODEL_CATALOG_FILEPath to the available model list configconfig/model_catalog.jsonconfig/model_catalog.json
Env Varenv.VLLM_BASE_URLvLLM endpoint (Leave blank if not used)http://eva-agent-vllm-service/v1http://eva-agent-vllm-service/v1
Env Varenv.LLM_PROVIDERProvider for the LLM servicevllmvllm
Env Varenv.LLM_MODELLocal model to be servedqwen3-vl-8b-awqqwen3-vl-8b-awq
Env Varenv.QDRANT_HOSTQdrant DB server addresshttp://eva-agent-qdranthttp://eva-agent-qdrant
Env Varenv.QDRANT_PORTQdrant DB port number"6333""6333"
Env Varenv.OPENAI_API_TYPEOpenAI API typeazureazure
Env Varenv.OPENAI_API_VERSIONOpenAI API version2025-04-01-preview2025-04-01-preview
Env Varenv.AZURE_OPENAI_ENDPOINTAzure OpenAI API endpointhttps://xxxxhttps://xxxx
Env Varenv.DEPLOYMENT_NAMEAzure model deployment namegpt-5-chat-2025-08-07gpt-5-chat-2025-08-07
Env Varenv.AZURE_MANAGEMENT_BASE_URLAzure Management Endpointhttps://management.azure.comhttps://management.azure.com
SecretsecretEnv.OPENAI_API_KEYOpenAI API Key (Dummy value required if unused)
SecretsecretEnv.AZURE_OPENAI_API_KEYAzure OpenAI API Key
SecretsecretEnv.AZURE_CLIENT_SECRETAzure App Registration Client Secret
SecretsecretEnv.CLOVA_API_KEYNaver Hyper Clova API Key

⚠️ Warning: For Secret items such as OPENAI_API_KEY, you must provide a dummy value even if they are not in use to avoid errors. (Do not comment them out).




Step 4: Run EVA Agent Installation

Once all configuration files are ready, proceed with the installation using the command below. The Helm installation begins after the ECR login.

# 1. Login to AWS ECR (To secure Image Pull permissions)
NS=eva-agent
ACCOUNT_ID=339713051385
AWS_ECR_REGION=ap-northeast-2
AWS_ECR_HOST=${ACCOUNT_ID}.dkr.ecr.${AWS_ECR_REGION}.amazonaws.com
aws ecr get-login-password --profile default | docker login --username AWS --password-stdin $AWS_ECR_HOST

# 2. Execute Helm Install (Match the yaml files for your environment)
docker_config_file="$HOME/.docker/config.json"
values_file="$docker_config_file-values.yaml"
cat > "$values_file" << EOF
dockerConfig:
json: $(cat "$docker_config_file" | base64 -w0)
EOF

# --version is the chart version (not the image tag)
helm install eva-agent eva-agent/eva-agent --version=2.1.2 -n eva-agent \
-f values.yaml \
-f values-k3s.yaml \
-f values-secret.yaml \
-f $values_file

rm -f "$values_file"

(Note: If you are in a AWS environment, use values-aws.yaml instead of values-k3s.yaml.)




Step 5: Verify Installation Status

After the installation is complete, verify that all services are up and running.

kubectl get all -n eva-agent