Chat memory with LangChain4j and Ollama

This example shows how a bounded message window carries conversational state across turns.

The first call states a fact, and the second one relies on the memory window to recall it.

Requirements

• Java 21+
• Maven 3.9+
• Ollama installed locally
• A pulled chat model such as llama3.2:1b

Steps for running this example in the shell

1. Start Ollama and pull the model:

ollama serve
ollama pull llama3.2:1b

2. Run the application:

mvn compile exec:java

Output

Nice to meet you, Boni.
Your name is Boni.