Deployment steps
1
Plan
Review security, confirm your environment meets the requirements, and size the compute your deployment needs.
2
Provision
Create the cluster, prepare GPU nodes and networking, create the database, search, and storage SeekrFlow depends on, then set up cloud identity and secret management, and collect your credentials and values.
3
Prepare to install
Set up the
sfe CLI, generate the values file, add your models, and choose the model each service uses. Then connect cloud identity, create secrets, and connect PostgreSQL.4
Deploy
Run preflight checks, install SeekrFlow, verify the deployment, and create the first administrator.
Installation overview
Before you install, Seekr gives you credentials for its artifact registry, which holds the Helm chart and container images, and thesfe command-line interface (CLI). SeekrFlow installs from the Helm chart, using a values file that describes your environment. If your environment has no outbound internet access, you first mirror the container images into your own registry.
You can install with sfe or with Helm directly:
Use
sfe unless you need to control each step yourself. The Helm path has more manual steps and none of sfe’s checks.
For how SeekrFlow runs once it’s installed, see Architecture. For what the deployment connects to and what you are responsible for securing, see Security considerations.
Supported deployment environments
Every environment follows the same deployment steps. What changes is which services fill each role, such as Amazon Simple Storage Service (Amazon S3) or an S3-compatible store for object storage. See Architecture.