# Design a Multi-Service Workflow Engine Design and demonstrate a workflow engine whose independent services run on separate ports and call one another to complete a workflow. The result should make service boundaries, workflow state, failures, and local end-to-end verification explicit. ### Constraints & Assumptions - Each service owns one bounded responsibility and exposes a documented interface. - Calls may fail, time out, or be retried; duplicate work must be considered. - The design must be runnable locally as multiple processes, but no particular language or framework is required. ### Clarifying Questions to Ask - Is the workflow synchronous, asynchronous, or a mix of both? - Which steps have irreversible side effects and need compensation? - What completion result must the caller observe? ```hint Define the state machine Name each workflow state and transition before choosing transports or frameworks. ``` ### What a Strong Answer Covers - Clear service boundaries and interface contracts. - An orchestration or choreography choice justified for the workflow. - Durable state, correlation identifiers, idempotency, retry, timeout, and compensation behavior. - A local multi-port test that proves both the success path and at least one failed dependency path. ### Follow-up Questions 1. How would you resume a workflow after the coordinator restarts? 2. When would an event broker be preferable to direct service calls?