kafka-streams-topology

# Kafka Streams Topology Skill

Expert knowledge of Kafka Streams library for building stream processing topologies in Java/Kotlin.

## What I Know

### Core Abstractions

**KStream** (Event Stream - Unbounded, Append-Only):
- Represents immutable event sequences
- Each record is an independent event
- Use for: Clickstreams, transactions, sensor readings

**KTable** (Changelog Stream - Latest State by Key):
- Represents mutable state (compacted topic)
- Updates override previous values (by key)
- Use for: User profiles, product catalog, account balances

**GlobalKTable** (Replicated Table - Available on All Instances):
- Full table replicated to every stream instance
- No partitioning (broadcast)
- Use for: Reference data (countries, products), lookups

**Key Differences**:
```java
// KStream: Every event is independent
KStream<Long, Click> clicks = builder.stream("clicks");
clicks.foreach((key, value) -> {
    System.out.println(value); // Prints every click event
});

// KTable: Latest value wins (by key)
KTable<Long, User> users = builder.table("users");
users.toStream().foreach((key, value) -> {
    System.out.println(value); // Prints only current user state
});

// GlobalKTable: Replicated to all instances (no partitioning)
GlobalKTable<Long, Product> products = builder.globalTable("products");
// Available for lookups on any instance (no repartitioning needed)
```

## When to Use This Skill

Activate me when you need help with:
- Topology design ("How to design Kafka Streams topology?")
- KStream vs KTable ("When to use KStream vs KTable?")
- Stream operations ("Filter and transform events")
- Joins ("Join KStream with KTable")
- Windowing ("Tumbling vs hopping vs session windows")
- Exactly-once semantics ("Enable EOS")
- Topology optimization ("Optimize stream processing")

## Common Patterns

### Pattern 1: Filter and Transform

**Use Case**: Clean and enrich events

```java
StreamsBuilder builder = new StreamsBuilder();

// Input stream
KStream<Long, Clic