Skip to main content
sfe is SeekrFlow’s command-line tool. It generates your values file, creates SeekrFlow’s secrets and databases, and installs and upgrades SeekrFlow. Set it up on the machine you run the rest of this guide from.

​
Choose where to run commands

Run the rest of this guide from a machine that can reach the Kubernetes API and Seekr’s registry. If your laptop can’t reach the cluster’s API, use a machine inside your network, such as a bastion host. The identity you run commands as needs Kubernetes cluster-admin permissions. On Amazon Elastic Kubernetes Service (Amazon EKS), only the identity that created the cluster has them by default. Give any other identity an access entry. On Amazon Web Services (AWS), sfe signs in with your AWS Command Line Interface (AWS CLI) credentials and needs a default Region. If aws configure get region prints nothing, set one:
Installing with Helm is for experienced Helm and Kubernetes users who are building their own automation. If that’s you, you need sfe only to mirror container images. Otherwise, follow this page to install sfe.

​
Install the sfe CLI

Download sfe from Seekr’s registry with the registry username and token from your Seekr account team. Set REGISTRY_USERNAME and REGISTRY_PASSWORD to those values, then follow the steps for your machine’s operating system.
1

Download the binary

Replace <version> with the sfe version to install, such as 3.11.2:
2

Check the download

The command prints sfe-linux-amd64: OK. If it reports FAILED, download the binary again.
3

Install the binary

4

Confirm the version

If you don’t have admin rights on the machine, put sfe in a folder in your home directory that’s on your PATH instead. sfe doesn’t need the helm or aws commands, but other steps in this guide do. For every tool the guide uses, see Review SeekrFlow requirements.

​
Save a profile

A profile saves the settings you’d otherwise type on every sfe command.
1

Find your Kubernetes context

Copy the name it prints.
2

Create the profile

Replace <profile-name> with a name for this deployment, such as prod:
sfe asks for each setting. Enter these values:
3

Check the profile

The command prints the settings you entered.
Later commands in this guide use -p <profile-name> to apply these settings. Profiles don’t save the release version. Add it with -r on each command.

​
Sign in to the registry

Sign in with the same REGISTRY_USERNAME and REGISTRY_PASSWORD you used to download sfe:
sfe signs in to Seekr’s registry and saves the credentials for downloading the chart, installing, and upgrading.

​
Mirror container images

If your cluster can’t reach Seekr’s registry, copy SeekrFlow’s container images into a registry inside your network. You can also mirror them if you want your cluster to pull images only from your own registry. The machine you mirror from needs to reach Seekr’s registry and registry.k8s.io, which hosts the JobSet controller image.
1

Sign in to both registries

sfe reads registry sign-ins from Docker’s credentials file, not from sfe chart login. Sign in to Seekr’s registry and to your own:
On Amazon Elastic Container Registry (Amazon ECR), skip the second command. sfe signs in to Amazon ECR with your AWS credentials.
2

Preview the copy

To see what sfe will copy, run:
3

Copy the images

On Amazon ECR, each image needs its own repository. To create them as you copy, add --create-missing-repos. This requires the ecr:CreateRepository permission.
4

Confirm the copy

To confirm every image arrived, run:
To also confirm each copied image is identical to Seekr’s original, run:
5

Point SeekrFlow at your registry

Set global.repository in your values file to your registry. Every image in the chart pulls from there, including images for supporting components such as MLflow and Argo Workflows.
Last modified on October 16, 2026