Build AI Agents with Open Source
summary
A step-by-step build of a Chilean sommelier AI agent implemented in Python using the Strends SDK and a local Olama model (Llama 3.1), then switched to Amazon Bedrock (Cloud Opus 4.6) by changing two lines. The agent includes two real tools—an inventory lookup and a food‑to‑wine pairing module—and a file‑based session memory so recommendations reflect cellar contents and user tastes; the demo shows end‑to‑end tool calling, memory working, and seamless provider swapping.
AI-written summary · about
chapters (16)
- 0:00Introduction to workshop
- 1:23What is an AI agent?
- 4:13Agent reasoning loop
- 8:10Basic agent with Strands
- 11:23Customize with system prompt
- 13:52Add tools to agent
- 16:25Debugging tools
- 17:29Agent with multiple tools
- 19:57Model memory limits
- 21:00Memory with session manager
- 24:13Final local agent
- 24:37Switch to Bedrock
- 26:29Alternative model providers
- 27:55Custom models per user
- 28:16Strands advanced features
- 29:22Invitation to build agents
related talks
40mAgents in Action: Advancing Open Source AI Missions
16mBuilding Agentic AI applications with Positron and AWS Strands Agents
14mIntroducing Strands Agents, an Open Source AI Agents SDK
53mHands On Building Agents
29mDemystifying AI Agents with Python Code
38mBuilding AI Agents and Assistants for DevOps Workflows
19mBuilding Agents at Cloud Scale
30mMind the Gap - A Developer's Roadmap To Building AI Agents