Skip to content

projectEureka Docs

How to use projectEureka projects, apps, flows, and storage, and how to administer them.

projectEureka runs your team’s work either as cloud VMs on AWS or Google Cloud, or on your own Kubernetes cluster, where machines run as KubeVirt VMs. You open it in a web browser, sign in, and work inside projects. In a project you can start apps such as remote desktops, Jupyter, code-server (VS Code in the browser), RStudio Server, and ParaView (desktops include Firefox), build flows that run jobs step by step, submit Slurm batch jobs directly, and keep files in shared storage. Each project card on the home page shows what is running and an estimated cost, so costs are tracked per project.

These docs are for two kinds of readers. Users create or join projects, then launch apps, run flows, and manage files. Within a project, a member is an Owner, Editor, or Reader. System admins manage users, the apps and machine types people can launch, AI services, storage, and time limits for everyone.

  • Projects: a project holds members, apps, flows, storage, and credentials. The cost of everything you run is tracked against, and limited by, the project it runs in. Billing itself usually goes through your organization’s procurement.
  • Apps: a launcher is a saved app setup. Launching it starts a session, such as a desktop, Jupyter, or code-server that you open in the browser. A project can also have one shared desktop that one member drives while others watch.
  • Flows: a flow is a list of stages that run in order, such as CPU or GPU jobs, file copies, language-model calls, refine (RFN) stages that repeat a model call until its reply passes your checks, and quantum (Q) stages that run circuits on IonQ or Amazon Braket (Braket on AWS only; on Kubernetes, IonQ with an API key). Each submit is a run, and its files are kept in the project’s storage.
  • Jobs: projectEureka accepts Slurm jobs directly. From a terminal, sbatch (or srun) submits an ordinary #SBATCH script, -p/--partition picks the project, and projectEureka starts machines for the job; squeue, scancel, scontrol show job, sacct, and sinfo work on those jobs, and ccqsub is also accepted. Multi-node, MPI, and array jobs run under a small Slurm cluster created just for that job.
  • Storage: each project has your home directory, a project directory (prj-storage), and S3 or GCS buckets, all on the Storage page.
  • Credentials: project credentials are API keys that flow stages use to call AI services. A system admin decides which services can be chosen.

New here? Get started walks through your first sign-in screen by screen. Then create a project or ask a project Owner to add you to theirs. If you are walking through the SC26 demos, see the SC26 demo walkthrough. If you are a system admin, start with Manage users.