Building Robust, Secure LLM and Agentic AI Workflows
summary
An open-source agent catalog built on Langraph registers tools and prompts into a vector database, uses semantic search to pick among 100+ tools, versions tools, and persists all agent responses and logs for audit and rollback; a Streamlit demo shows a song DB and a Spotify API tool that actually plays music rather than just returning text. Production concerns include monitoring with Prometheus/Grafana, zero-trust access controls, prompt-injection guardrails, hallucination detection, and options to self-host models for data privacy.
AI-written summary · about
chapters (13)
- 0:00Introduction and setup
- 1:07Era of AI agents
- 3:35What are AI agents
- 7:09Agentic workflows explained
- 8:35LangGraph framework
- 12:08Enterprise tool challenges
- 15:17Agent Catalog solution
- 16:42Demo with LangGraph
- 20:33Agent workflows summary
- 23:03Security for AI agents
- 26:39Zero trust for LLMs
- 28:02Private LLM deployment
- 29:49Q&A and closing
related talks
41mBuilding LLM Agents – Challenges and Learnings
40mAgents in Action: Advancing Open Source AI Missions
02h 08mIntroduction to AI Agents
17mHow to Make Useful AI Agents
35mBuilding LLM-based Agents
16mBuilding Agentic AI applications with Positron and AWS Strands Agents
29mBuilding Trust in AI Agents
19mBuilding Agents at Cloud Scale