Requirements

The tools, cluster, and values you need before deploying Meko BYOC

Tools

Run the install from a workstation or jump host that can reach your cluster's API server and the internet. It needs:

  • kubectl, with a context for the cluster.
  • helm 3.8 or later. The charts are pulled from an OCI registry.
  • AWS CLI v2, signed in to the AWS account that will hold the buckets and IAM users.
  • openssl and curl.

No Git checkout is needed. The install pulls the charts from oci://quay.io/yugabyte/charts.

Cluster

Meko runs on Amazon EKS. The install creates a YugabyteDB universe (replication factor 3) in the cluster, beside the Meko stack, so size the nodes for both.

poc prod
Database nodes 3 × 8 vCPU / 32 GiB 3 × 16 vCPU / 64 GiB
Stack nodes 3 × 8 vCPU / 32 GiB 3 × 16 vCPU / 64 GiB
  • The database runs one tablet server and one master per node, so it needs at least three nodes. The database nodes and stack nodes can be one node group or two. Two groups let you keep the database off the stack's nodes.
  • The stack nodes also carry YugabyteDB Anywhere, which manages the database (about 4.5 vCPU requested), Valkey, the telemetry collectors, and the reconcile agent. Leave room for one node to fail.
  • Database disk per node is 200 GiB for poc and 500 GiB for prod. A volume can grow but not shrink. As a guide, 5,000 datapacks use about 8 GB per node, and 100,000 use about 73 GB.
  • The cluster needs a storage class that provisions block volumes in each zone you use. You name it in the database definition when you deploy.

Values

The install command starts with a header of values to fill in. This table lists them, who supplies each, and where this guide covers it.

Variable What it is Supplied by Covered in
Meko license file The credentials that pull the Meko charts and images from Quay, and that ship metrics and logs. It replaces the Quay robot account, the metrics password, and the log key Meko support Apply your Meko license
AGENT_API_KEY The reconcile agent's API key, shown once when you register the environment Meko Deploy Meko on AWS
CHART_VERSION, API_TAG, MCP_TAG, UI_TAG, RAG_TAG, LF_TAG, CONSUMER_TAG, AGENT_TAG, GW_TAG Chart version and image tags of the release Meko, filled in for you Deploy Meko on AWS
BEDROCK_KEY, BEDROCK_SECRET Access key of the IAM user for Bedrock and the knowledge-base bucket You Create the IAM users
BEDROCK_REGION AWS Region where Bedrock Titan embeddings are enabled. Default us-east-1 You Create the IAM users
KB_S3_BUCKET, KB_S3_REGION Knowledge-base bucket and its Region. The Region must equal BEDROCK_REGION You Create the S3 buckets
LANGFUSE_S3_BUCKET, LANGFUSE_S3_REGION Langfuse bucket and its Region You Create the S3 buckets
LANGFUSE_S3_KEY, LANGFUSE_S3_SECRET Access key of the IAM user for the Langfuse bucket. Blank with IRSA You Create the IAM users
OPENAI_KEY OpenAI API key You Get an OpenAI key
OIDC_ISSUER_URL, OIDC_CLIENT_ID, OIDC_CLIENT_SECRET Your identity provider's OIDC client You Configure your identity provider
PUBLIC_URL, ACM_CERT_ARN Production only: the DNS name for Meko and its ACM certificate You Prepare the cluster and network
LANGFUSE_OWNER_EMAIL Email address for the Langfuse owner login You Deploy Meko on AWS
FOUNDATION_CR Path to your edited database definition You Deploy Meko on AWS
NS The namespace Meko is installed into. The console fills in a default, and your license must be applied in the same namespace You Apply your Meko license
SIZE poc or prod You Deploy Meko on AWS
EXTRA_VALUES Your own -f files and --set flags, applied last. Every rerun must pass them again. You Upgrade

Next

Create the S3 buckets.