Quarkus LLM Demos

This repository contains a set of demos demonstrating some Quarkus Langchain4J features:

• Declarative AI services
• Tools and Agents
• Observability
• Building chatbots
• Implementing a RAG (Retrieval Augmented Generation) pattern

Prerequisites

• Java 21 or later
• Apache Maven 3.9.6 or later
• Podman or Docker (Podman recommended)
• A working Docker Compose installation (working with Podman recommended)

Running the Demos

The repo follows an incremental structure. Each demo builds on the previous one. To run a demo, navigate to its directory and run mvn quarkus:dev. The UI is available at http://localhost:8080.

Demo 1: Declarative AI Services

This demo (contained in the 0-summarizer directory) demonstrates how to create a declarative AI service using Quarkus and Langchain4J. The service provides a REST API to summarize insurance claims (you can find claim examples in the claim-samples directory). It also extracts the policy number and analyzes the sentiment of the claim.

Demo 2: Tools and Agents

This demo (located in the 1-agent directory) extends the previous one but uses a tool to detect if the new document is related to an existing claim. If so, the application updates the existing claim.

Demo 3: Observability

This demo (located in the 2-observability directory) extends the previous one by adding observability features to the application.

You can run the observability stack using:

> podman compose -f infrastructure/observability.yaml up

Then use the application. The Jaeger UI is available at http://localhost:16686. To demonstrate Prometheus, use the dev-ui (available at http://localhost:8080/q/dev).

Demo 4: Building Chatbots

This demo (located in the 3-chatbot directory) extends the previous one by adding a chatbot to the application. This chatbot is very simple and uses a tool to find a given claim.

Demo 5: Implementing a RAG (Retrieval Augmented Generation) Pattern

This demo (located in the 4-rag directory) extends the previous one by adding a RAG pattern to the application.

License

This project is licensed under the Apache License, Version 2.0.