Episode

Building with Agentic RAG and Cohere

with Korey Stegared-Pace, Sandra Kublik

Make Azure AI Real

First, we'll deep-dive into agentic RAG with Cohere models, covering the transition from vanilla to agentic RAG. Then we'll show how an agent can dynamically route queries to the best data sources for smarter, faster retrievals. Using a notebook example walkthrough, participants will learn to set up and run a simple agent that powers context-aware applications.

Chapters

  • 00:05 - Introduction and welcome
  • 02:35 - Agentic RAG and Cohere
  • 02:53 - Single-shot RAG is not enough - user examples
  • 05:20 - Agentic RAG is an expansion
  • 08:07 - Tools in Agentic RAG workflow
  • 10:45 - Question from the audience - keeping user information confidential
  • 13:11 - Question from the audience - graphic/PDF support
  • 14:51 - Question from the audience - R+ model example
  • 17:20 - Example notebook
  • 22:13 - Tool selections, instructions to agent
  • 30:01 - Step-by-step breakdown
  • 37:29 - Human monitoring for LLMs
  • 41:00 - Question from the audience - limitation for tools?
  • 45:23 - Dealing with multiple results
  • 49:20 - Citations
  • 56:09 - Resources, summary, what's next