Workflows that
survive crashes.
Write a workflow as ordinary async code in Rust, Python or TypeScript. Orcher journals every step, so after a crash or a deploy a run resumes from its last completed step.
#[workflow(name = "order")]async fn order(ctx: WorkflowContext, order_id: String) -> Result<String> { let charged: String = ctx.execute_task(charge, order_id.clone()).await?; // A durable timer: the engine keeps it, so a worker restart doesn't reset it. ctx.sleep(Duration::from_secs(10)).await?; let shipped: String = ctx.execute_task(ship, order_id).await?; Ok(format!("{charged}, then {shipped}"))}@workflow(name="order")async def order(ctx: WorkflowContext, order_id: str) -> str: charged = await ctx.execute_task(charge, order_id=order_id) # A durable timer: the engine keeps it, so a worker restart doesn't reset it. await ctx.sleep(timedelta(seconds=10)) shipped = await ctx.execute_task(ship, order_id=order_id) return f"{charged}, then {shipped}"@Workflow({ name: 'order' })export class Order { async run(ctx: WorkflowContext, orderId: string): Promise<string> { const charged = await ctx.executeTask(orderTasks.charge, orderId); // A durable timer: the engine keeps it, so a worker restart doesn't reset it. await ctx.sleep(Duration.fromSeconds(10)); const shipped = await ctx.executeTask(orderTasks.ship, orderId); return `${charged}, then ${shipped}`; }}The core model
Six building blocks of a durable workflow.
Ordinary async functions whose progress survives crashes and restarts. Code-first, no YAML.
Where side effects go: each task has its own retries and timeouts, and its result is journaled.
Every step is recorded, so a run replays to where it stopped without repeating completed steps.
Send an event to a running workflow, and have a workflow wait for one, with or without a timeout.
State a workflow keeps as it runs, and how callers read it from outside.
Long-lived entities with their own state, exclusive and shared operations.
Your language, first-class
Native, idiomatic APIs for Rust, Python, and TypeScript.
#[workflow] and #[task] macros, typed task references, and registration at startup with no manual wiring.
$ cargo add orcher-sdk@workflow and @task decorators on plain async functions, a worker builder and a typed client.
$ pip install orcher-sdkFunctional or decorator-based definitions, typed task references, and dependency injection for tasks.
$ npm install @orcher/sdkPatterns & use cases
Common workflow shapes, and the systems they are used in.
Workflow patterns
Use cases
Guides for your role
Core concepts, building patterns, distributed systems, and production readiness.
5 guidesETL pipelines, batch processing, parallel execution, and production data workflows.
5 guidesDeploying, operating, scaling, and monitoring the Orcher platform.
5 guidesRun and operate
Deploy, configure and observe the engine.