Design Patterns for Multi Agent Systems
summary
Compares three concrete orchestration styles by building a runnable demo that wires Spring AI, LangChain4j, and Embabel into separate services talking to a Redis memory server and a curated slice of the Olist dataset (≈100k orders). Shows routing-first pipelines, tool calls via MCP, a generator‑critic loop for quality with an observed 75% enterprise failure baseline, and goal‑oriented planning that assembles workflows on the fly — all three approaches run and expose tradeoffs in latency, determinism, and operability.
AI-written summary · about
chapters (9)
- 0:00Introduction and motivation
- 2:15Agentic systems overview
- 10:45Design patterns history
- 22:45Demo setup and frameworks
- 26:22Routing tools and memory
- 35:39Tool use and MCP integration
- 43:23Multi-agent coordination
- 49:27Goal oriented planning
- 55:10Decision framework summary
related talks
41mPatterns & Practices for building Multi-Agent Systems
25mBeyond Orchestration: An Open Source Architecture for Agentic Systems
50mHarnessing Event-Driven and Multi-Agent Architectures for Complex Workflows in Gen...
32mEvent-Driven & Multi-Agent Architectures for Gen AI Workflows
49mAgentic AI Patterns
46mAgentic Design Patterns Where You Are the Hero
45mAn Agent Walks Into a Bar... and Meets Another Agent: Multi-Agent Systems in Action
47mAgentic AI Patterns