Sources

Sources implement the source.Source interface and push records into the pipeline.

Generator source

Use generator source for local examples and tests.

records := []types.Record{
    types.NewRecord([]byte("a"), []byte(`{"word":"hello"}`)),
    types.NewRecord([]byte("b"), []byte(`{"word":"mailer"}`)),
}

src := source.NewGeneratorSource(records)

Slice source

Slice source is similar to generator source and is useful in tests where you already have records in memory.

Kafka source

Kafka is the production streaming source.

src := source.NewKafkaSource(
    source.KafkaBrokers("localhost:9092"),
    source.KafkaTopic("orders"),
    source.KafkaGroupID("order-processor"),
    source.KafkaStartFrom(source.OffsetEarliest),
    source.KafkaFetchBytes(1, 10*1024*1024),
    source.KafkaCommitBatch(100),
)

Core options

Option	Purpose
`KafkaBrokers(...string)`	Required bootstrap brokers.
`KafkaTopic(topic)`	Single topic.
`KafkaTopics(...string)`	Multiple topics in consumer-group mode.
`KafkaGroupID(id)`	Consumer group id and broker offset tracking.
`KafkaStartFrom(offset)`	`OffsetEarliest` or `OffsetLatest`.
`KafkaFetchBytes(min, max)`	Fetch request sizing.
`KafkaCommitBatch(n)`	Commit offsets every N messages; `0` commits per message.
`KafkaParallel()`	Per-partition readers; disables consumer-group auto-balancing.

Authentication

src := source.NewKafkaSource(
    source.KafkaBrokers("localhost:9092"),
    source.KafkaTopic("orders"),
    source.KafkaSASL(auth.SASLConfig{
        Mechanism: auth.SASLPlain,
        Username: "user",
        Password: "pass",
    }),
    source.KafkaTLS(auth.TLSConfig{
        CAFile: "ca.pem",
    }),
)

Deserialization

src := source.NewKafkaSource(
    source.KafkaBrokers("localhost:9092"),
    source.KafkaTopic("orders"),
    source.KafkaDeserialize(source.NewJSONDeserializer[Order]()),
    source.KafkaDeserializeFailurePolicy(source.DeserFailureDLQ),
    source.KafkaDeserializeDLQ(dlq),
)

The deserialized value is stored in Record.Parsed. The raw bytes remain in Record.Value.

Watermarks

src := source.NewKafkaSource(
    source.KafkaBrokers("localhost:9092"),
    source.KafkaTopic("orders"),
    source.KafkaWithWatermarks(5*time.Second),
    source.KafkaWatermarkInterval(500*time.Millisecond),
)

KafkaWithWatermarks uses bounded out-of-orderness: watermark equals max event timestamp seen minus the configured lateness.

Exactly-once source mode

src := source.NewKafkaSource(
    source.KafkaBrokers("localhost:9092"),
    source.KafkaTopic("orders"),
    source.KafkaGroupID("orders-eo"),
    source.KafkaExactlyOnce(),
)

Exactly-once mode disables eager broker offset commits and uses read_committed isolation. Offsets are committed only after coordinated checkpoints complete.

On this page