Non-deterministic? No problem! You can test it!
summary
Shows a working Quarkus demo that wires LLMs into an app, uses Playwright to run UI tests and record videos, and spins local models in CI via Testcontainers + Quarkus spins local LLMs in CI so tests don’t require cloud calls. Proposes validating non‑deterministic outputs with a separate validator agent checks LLM output and can re‑prompt (decoupled from the production model), adds guardrails, retries and observability, and recommends using smaller 7–32B models locally as drop‑in test proxies.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:48AI-integrated applications
- 3:10Large language models
- 7:34AI transforming developer roles
- 13:07Integrating AI into development
- 15:58Demo: AI insurance claim app
- 18:47Quarkus and AI integration
- 21:15Testing AI components
- 25:49AI integration types and testing
- 30:01Guardrails in AI applications
- 32:48AI for validation and testing
- 37:26Observability in AI production
- 44:26Local vs production AI environments
- 47:17Prompt engineering and migration
- 50:27Key takeaways
related talks
39mLLM Agents Gone Wild (And How to Tame Them with Quarkus)
48mAgentic AI with Quarkus, LangChain4j and vLLM
36mCrafting Intelligent Cloud Native applications with Quarkus
49mPushing LLMs over the Edge: Exploring the Limits of the Possible
55mTest Driven Development It’s Easier Than You Think
52mLivecoding: Creating Beautiful Java Code With LLM and Agents
45mComplementing LLM with agentic AI: a Quarkus story on AI
44mCrafting Intelligent Supersonic Subatomic applications with Quarkus