Implements a Kafka client with basic functionality for producing and consuming records, supporting multiple brokers and partitions, and enabling parallel consumption through consumer groups.
Kotlin Multiplatform implementation of Kafka client, written from scratch.
This library is at the moment in early development stage and should not be used in production.
Used technologies:
net.socket from NodeJS wrapper for JS and WASMktor-network for other platformskotlinx-serialization (with custom serializer for Kafka protocol)There is a demo app built using Compose Multiplatform that runs on Android, iOS and Desktop: kotlin-kafka-client-demo
Library is published to Maven Central under name io.github.vooft:kotlin-kafka-client-core.
Add the dependency to your project:
kotlin {
...
sourceSets {
commonMain.dependencies {
implementation("io.github.vooft:kotlin-kafka-client-core:<version>")
}
}
}
Then in your code, start with class KafkaCluster:
Main goal for this library is to provide a simple library to use for typical Kafka use-cases, such as produce to a topic from one service and consume from it in another, optionally using groups to consume in parallel.
It is not a goal to implement all the features of Kafka protocol, for example to be used in Kafka Streams.
At the moment library supports only basic features of Kafka protocol:
fun main() {
// define bootstrap servers
val bootstrapServers = listOf(
BrokerAddress("localhost", 9092),
BrokerAddress("localhost", 9093),
BrokerAddress("localhost", 9094)
)
// create instance of KafkaCluster
val cluster = KafkaCluster(bootstrapServers)
// both producer and consumer are bound to a single topic
val producer = cluster.createProducer(KafkaTopic("my-topic"))
producer.send("my-key", "my-value")
// consumer can be used both with group and without (in this case group is not created and no offset remembered)
val consumer = cluster.createConsumer(KafkaTopic("my-topic"), GroupId("my-group"))
// low-level API is similar to the kafka-clients implementation, but Flow-based API is on the way
consumer.consume().forEach {
println(it)
}
}
Surfaced from shared tags and platforms — no rankings paid for.