Using Poiesis

Poiesis is a TES server. Anything that speaks TES can drive it: workflow engines (Nextflow, Cromwell, Toil), client SDKs, or plain curl. This section covers two concrete scenarios.

Pick your client

• curl — fastest way to learn the API surface. The deployment guide has working examples.
• Workflow engines — see Nextflow. Most TES clients only need a base URL and (optionally) an S3 endpoint.
• Programmatic — the TES OpenAPI spec is the contract.

Common patterns

• No-IO task — just executors. Useful for smoke testing image pulls and Pod scheduling.
• S3 in, S3 out — inputs[] and outputs[] point at object storage; TIF stages onto /transfer, TOF uploads from /transfer.
• Inline content — inputs[].content carries the file body inline; no external storage needed.
• Multi-executor — executors run strictly sequentially through init-container ordering; the /transfer PVC persists between them.

What Poiesis doesn't do

• Schedule across clusters. One TaskPod, one cluster.
• Mutate your executor image. Commands run exactly as submitted.
• Cache anything. Each submission is independent.
• Run its own login flow. Poiesis is a resource server, not an OAuth client. It validates OIDC bearer tokens minted by an external IdP (Keycloak, Auth0, Dex, …). See Authentication for the operator configuration and the principal-claim model.