[DEVTALKS] From AI to Agent with Langchain4J

You’ve built AI features into your application. Models are wrapped in services, RAG is in place, tools are wired, and calls are flowing. Then requirements evolve. A single response is no longer enough. You need steps that follow each other, branches based on decisions, retries when things fail, and sometimes multiple actions happening at the same time. At that point, it becomes unclear how to structure the logic without losing control or readability.
In this talk, we move from AI calls to agentic systems using LangChain4j. We’ll explore common workflow patterns such as sequential, loop, parallel, and conditional flows, along with error handling as a first-class concern. From there, we’ll zoom out to agentic concepts like shared state with an AgenticScope, goal-oriented agents, and pure agentic setups. We’ll also look beyond AI-only approaches, mixing in non-AI agents and human-in-the-loop interactions. All of this will be built live, step by step, to show how these ideas work in practice.
You’ll leave with a concrete understanding of when to use workflows and when to rely on agents, how to combine both without confusion, and how to keep state, goals, and responsibilities explicit. The goal is not to build smarter prompts, but systems that are easier to reason about, easier to evolve, and better aligned with real-world application needs.

Transcript

1. From AI to Agent with Langchain4J @LoMagnette
2. Loïc Magnette @LoMagnette lomagnette.github.io
3. AI Service @LoMagnette
4. AI Service @UserMessage(""" Turn this message into a sitter card with exactly these five lines. Invent nothing — if the message does not say, write "not given": Dog: Meals: Walks: Watch out for: Vet: Message: {{message}}""") String card(@V("message") String message); @LoMagnette
5. AI Service var aiService = AiServices.builder(NoteRetriever.class) .chatModel(model) .build(); var notes = aiService.card(input); @LoMagnette
6. Agent @LoMagnette
7. Agent An agent is a service that talk to an AI model to perform a goalbased operation using the tools and context it has. @LoMagnette
8. Agent @Agent(description = "Turns a rambling message about the dog into a structured sitter card") @UserMessage(""" Turn this message into a sitter card with exactly these five lines. Invent nothing — if the message does not say, write "not given": Dog: Meals: Walks: Watch out for: Vet: Message: {{message}}""") String card(@V("message") String message); @LoMagnette
9. Agent var clerk = AgenticServices.agentBuilder(SitterCardClerk.class) .chatModel(model) .name("SitterCardClerk") .outputKey("Notes") .build(); @LoMagnette
10. Workflow Workflow @LoMagnette
11. State @LoMagnette
12. State @LoMagnette
13. State @LoMagnette
14. State Agentic Scope @LoMagnette
15. Sequence @LoMagnette
16. Sequence @LoMagnette
17. Loop @LoMagnette
18. Loop @LoMagnette
19. Parallel @LoMagnette
20. Parallel @LoMagnette
21. Routing @LoMagnette
22. Routing Police Herding Speed @LoMagnette
23. Typed key @LoMagnette
24. Typed key public record Checklist() implements TypedKey<String>{} var list = AgenticServices.agentBuilder(FridgeChecklist.class) .chatModel(model) .name("FridgeChecklist") .outputKey(Checklist.class) .build();
25. Human in the loop @LoMagnette
26. Human in the loop @LoMagnette
27. Observability / Monitoring @LoMagnette
28. Agentic @LoMagnette
29. Supervisor @LoMagnette
30. Goal Oriented @LoMagnette
31. Where to start WORKFLOW GOAP, P2P, Voting,… SUPERVISOR Control High Medium Low (LLM) Flexibility Low Medium-High High Cost Low Medium High Auditability Excellent Good Difficult Best for Stable pipeline Variable deps Adaptive systems Deterministic Autonomous Start here @LoMagnette
32. @LoMagnette
33. RESOURCES @LoMagnette lomagnette.github.io @LoMagnette