Building Vertical Agent with AG2 AgentOS
summary
Open source AG2 AgentOS claims that composing small specialized agents is often faster and cheaper than one big model and demonstrates that with working demos: a Kaggle AutoML pipeline that executes code in a stateful Jupyter executor, a Google Drive toolkit integration that downloads files, and a nested deep research agent that spawns summarize/decompose/critic subagents and saves Markdown reports. The framework ships captain agent tooling to auto-generate teams, supports Olama and Google Cloud backends, and includes evaluation utilities (agent_val and critic agents) used in live demos.
AI-written summary · about
chapters (10)
- 0:00AI agents and AG2 AgentOS
- 2:13AG2 creation and vision
- 10:45Conversible agents and programming
- 17:38AG2 evolution and features
- 25:56Live demos of AG2
- 31:28Kaggle AutoML with AG2
- 43:59Google Drive tools with AG2
- 1:02:56Deep research agent patterns
- 1:13:27Advanced AG2 applications
- 1:22:46Q&A
related talks
25mFrontiers of Agentic AI
01h 42mBuilding Responsible AI Agents in Decentralized Open-Source Environments
26mAI powered automation & multi-agent orchestration
40mAI Agents in Practice
25mHow Google DeepMind Runs Agents at Scale
34mBuilding Multi-Agent Generative-AI Applications with AutoGen
40mAgents in Action: Advancing Open Source AI Missions
26mThe World of Vertical AI Agents