Agentic AI Patterns
summary
Presentation claims a concrete library for Java (LangChain for Java + Quarkus) that implements four orchestration patterns — sequence loop parallel conditional — plus an LLM-driven supervisor planner and a pluggable planner API. It shows an agentic scope that shares state between agents, demos a €100 transfer composed of three agent invocations, and explains goal‑oriented planners that compute shortest agent paths. Implementation ships (introduced in 1.3.0, 1.11 releasing tomorrow), is composable with non‑AI and human‑in‑loop agents, and works in demos.
AI-written summary · about
chapters (15)
- 0:00Agentic AI patterns intro
- 3:05LLMs in Java integration
- 7:16AI services vs agents
- 8:42Programmatic agent workflows
- 12:30Composite agents with sequences
- 17:01Loop workflows for refinement
- 19:51Parallel agent calls
- 21:17Dynamic agent routing
- 23:07Agent scope and context
- 29:51Supervisor pattern demo
- 33:21Supervisor pattern demo
- 37:52Pluggable planner abstraction
- 41:52Goal-oriented orchestration
- 45:46Extending agentic patterns
- 46:09Q&A
related talks
47mAgentic AI Patterns
49mAgentic AI Patterns
35mAI Agent Control Logic with Pattern Matching
46mAgentic Design Patterns Where You Are the Hero
31mAgentic AI architecture patterns
27mEmerging Patterns Agentic AI
56mDesign Patterns for Multi Agent Systems
31mBuilding and Orchestrating Real Production-ready Agentic AI Systems