Cadence Workflow Engine

Cadence is a fault-oblivious stateful code platform developed at Uber Engineering in 2017, designed to orchestrate long-running business processes and microservice workflows with guaranteed execution and automatic retry. Created by Maxim Fateev and Samar Abbas, Cadence solves distributed orchestration by treating workflows as deterministic, replayable functions. Key features: workflow-as-code (write business logic as regular code in Go and Java, with the Cadence runtime handling persistence, retries, timeouts, and failure recovery). Deterministic replay records workflow executions as event histories, enabling recovery and debugging, ensuring workflows resume from exact state after failures. Activities are non-deterministic operations like API calls and database queries, executed with configurable retry policies, timeouts, and heartbeats. External signals can be sent to running workflows to modify behavior, while queries allow reading workflow state without affecting history. Compose complex workflows from smaller child workflows with independent failure and retry semantics. Cron-like scheduled workflows for recurring business processes. Searchable workflow execution history with Elasticsearch integration for filtering by workflow ID and status. Workflows persist indefinitely surviving process crashes, machine failures, and deployments, with guaranteed at-least-once execution. Workflow code versioning with auto-reset capability for safe deployment. Go and Java SDKs with Python and TypeScript community SDKs.