First Pipeline

Standalone example

The following example creates a pipeline that triggers every second (time adapter) and logs the current payload via SLF4J.

import io.pipelite.core.Pipelite;
import io.pipelite.core.context.PipeliteContext;
import io.pipelite.dsl.definition.FlowDefinition;

public class HelloPipelite {

    public static void main(String[] args) {

        FlowDefinition flow = Pipelite.defineFlow("hello-flow")
            .fromSource("time://ticker?period=1000")                  (1)
            .process("log-step", (exchange, contribution) -> {        (2)
                System.out.println("Tick: " + exchange.getInputPayload());
                exchange.setOutputPayload("processed");
            })
            .toSink("slf4j://output-logger")                          (3)
            .build();                                                 (4)

        PipeliteContext context = Pipelite.createContext();           (5)
        context.registerFlowDefinition(flow);
        context.start();                                              (6)
    }
}

1	The `time://` adapter generates a tick every 1000 ms.
2	An inline processor that receives the `Exchange` and a `ProcessContribution` - most processors ignore the latter, it exists for the rare case one needs to call `stopExecution()` and end the flow early from inside the processor itself.
3	The `slf4j://` sink logs the output message via SLF4J.
4	`build()` returns the immutable `FlowDefinition`.
5	The `PipeliteContext` is the runtime container.
6	`start()` starts all consumers and begins processing.

Spring Boot example

With the Spring Boot starter, flows are declared in a @FlowConfiguration class and activated via @EnablePipelite:

@SpringBootApplication
@EnablePipelite(flowConfigurations = IngressFlowConfiguration.class)   (1)
public class Application {

    public static void main(String[] args) {
        SpringApplication.run(Application.class, args);
    }
}

1	`flowConfigurations` lists the classes containing `@DefineFlow` methods.

@FlowConfiguration
public class IngressFlowConfiguration {

    @DefineFlow
    public FlowDefinition ingressFlow() {
        return Pipelite.defineFlow("ingress-flow")
            .fromSource("http://ingress")                 (1)
            .wireTap("audit-log", "slf4j://audit")        (2)
            .process("validate", (exchange, contribution) -> {
                // validation logic
                exchange.setOutputPayload(exchange.getInputPayload());
            })
            .toSink("slf4j://main-log")
            .build();
    }
}

1	The `http://ingress` source opens an HTTP listener on port 8080 at path `/ingress` (configurable - see HTTP Channel Adapter).
2	`wireTap` sends a copy of the message to the audit channel, then the main flow continues - currently on the same thread, so a slow audit target delays it too (issue #47).

For full details on the FlowConfiguration SPI — including dependency injection and multiple flow classes — see FlowConfiguration SPI.

Adding error handling

import io.pipelite.dsl.definition.builder.Backoff;
import java.time.Duration;

FlowDefinition flow = Pipelite.defineFlow("resilient-flow")
    .fromSource("kafka://my-topic")
    .process("transform", (exchange, contribution) -> {
        // processing that may throw exceptions
    })
    .toSink("slf4j://output")
    .withRetry(retry -> retry                                  (1)
        .maxAttempts(3)                                        (2)
        .backoff(Backoff.exponential(Duration.ofSeconds(2)))   (3)
        .onErrorChannel(err -> err.toDLQ()))                   (4)
    .build();

1	`withRetry(…)` resumes the flow at the step that failed, up to `maxAttempts` times.
2	Retries up to twice more after the original attempt (3 attempts total).
3	Waits between attempts, doubling each time (2s, 4s, …) - `Backoff.linear(…)` grows by a constant step instead.
4	An exhaustion action is mandatory: here, the built-in dead letter queue. `err.toChannel("queue://…")` routes to a named flow or channel adapter instead; `onExceptionHandler(…)` runs custom code instead of routing anywhere.

For error handling with no retry, use withErrorChannel(err → …) directly - same exhaustion-routing options, without the attempts/backoff in between. For full details see Error Handling.

Multi-channel concurrent example

pipelite-examples/pipelite-fooddelivery-application is a complete Spring Boot application showing several flows linked together, two of them running with concurrency > 1, fed by four different channel adapters:

A docker-compose.yml in the module starts a local Kafka broker. Run the app with mvn spring-boot:run and watch the console: a built-in load generator produces sample orders on its own, so concurrent processing is visible immediately with no manual input needed.