Skip to content

Workflow tutorial

Use an editable checkout example to test connected APIs, configure real or simulated nodes, and inspect success, low-stock, and unavailable-service outcomes.

Load the checkout example in /services. Review its service definitions, operations and graph before running. The draft is editable; loading it invokes no model.

Create separate cases for successful checkout, insufficient stock, and an unavailable dependency. Bind actual upstream outputs to downstream inputs. Keep expected outputs in assertions, not in the service instructions. Each case creates a fresh session, so a simulated reservation in one case cannot affect another.

To mix real and simulated calls, select real on a node and configure its HTTP endpoint and method. Select simulate for a missing dependency. Browser runs require browser-accessible endpoints; use the local CLI when the real service is on your computer or needs local secrets.

Inspect the graph after each run. A expected error is only a passed case when its assertions pass. A disconnected real service and a valid simulated error response are different results. Historical native-command recordings are retained as evidence of earlier implementations, not replayed as live results.

Open /gallery to explore checkout, inventory session, and shipping templates. Use Fork workflow to start an independent copy, or Reuse an Eve-simulated component in the editor to add a service agent to the current graph. Save the draft after signing in to keep it in your dashboard. Historical trajectories have their own evidence labels and never substitute for a new run.