Build a RAG Agent with NVIDIA Nemotron - A Developer's Guide to Agentic AI

Learn to build a powerful Retrieval-Augmented Generation (RAG) agent using NVIDIA Nemotron in this hands-on 11-minute workshop. Discover the fundamental differences between traditional RAG and agentic RAG systems while exploring advanced reasoning and tool-calling capabilities. Master the core concepts of agentic reasoning as you construct a multi-step agent with LangGraph to orchestrate complex tasks. Follow along with a detailed code walkthrough of the rag_agent.py file, learn to define retrieval chains with retrievers and rerankers, and understand how to create the agent's LLM and system prompt components. Experience practical deployment by building a functional IT help desk agent capable of answering complex user queries, complete with live demonstrations including password reset scenarios. Gain valuable debugging skills through log output examination and agent observability tools to trace performance and understand the agent's thought processes. Access the complete source code, technical blog post, and OpenRouter API resources to continue your development journey, whether you're a beginner starting with AI agents or an intermediate developer looking to enhance your RAG systems.

Syllabus

0:00 - Introduction to the RAG Agent Workshop: An overview of what will be covered in the workshop.
0:12 - Setting Up the Brev Environment: A step-by-step guide to deploying the development environment.
0:48 - Understanding RAG and Agentic AI: A foundational discussion of traditional RAG vs. agentic RAG.
2:48 - The Agenti-RAG Architecture: A look at the different components of the system, including the React Agent and Retrieval Chain.
3:07 - Building the Agent: Code Walkthrough: A detailed code review of the rag_agent.py file.
4:14 - Defining the Retrieval Chain: Explaining how to build the retrieval component with a simple retriever and a reranker.
5:01 - Creating the Agent's Brains: How to define the LLM and system prompt for the agent.
5:16 - Creating the LangGraph Flowchart: Connecting all the components into a functional graph.
5:44 - Running the Agent for Showtime: Deploying the agent behind an API and preparing to interact with it.
6:38 - Chat with the Agent: A live demonstration of how to interact with the agent.
7:27 - Examining the Log Output: How to debug and understand the agent's thought process.
7:56 - Live Demo: Resetting a Password: A practical example of the agent's reasoning in action.
9:47 - Agent Observability: An overview of using observability tools to trace the agent's performance.
10:09 - Conclusion and Next Steps: A wrap-up of what was learned and a call to action.