Workshop

Building a RAG Agent with Groq API and AgentCore Memory

Overview

In this workshop, we will build a complete RAG (Retrieval-Augmented Generation) Agent with the following capabilities:

• Calling the Groq API to use high-performance LLM models
• Chunking & Embedding documents for optimized vector search
• AgentCore Memory to maintain long-term context across chat sessions
• Tool Integration so the agent can automatically search FAQs and reformulate queries

AgentCore provides a framework for building AI agents with memory persistence, middleware hooks, and tool orchestration — enabling the agent to “remember” conversation history and personalize responses.

Contents

1. Workshop Overview
2. Prerequisites
3. Architecture
4. Running AgentCore
5. Resource Cleanup

Tech Stack

Prerequisites

• Python 3.8+
• Groq API Key
• Basic understanding of RAG and LLMs
• Familiarity with vector embeddings

Expected Outcome

By the end of the workshop, you will have:

• An agent capable of answering FAQs via vector search
• A memory system that remembers user preferences and context
• Tool orchestration allowing the agent to decide when to use tools
• Production-ready code with logging and error handling

• 5.1. Introduction
• 5.2. Preparation Steps
• 5.3. RAG Architecture Deployed on AWS AgentCore
• 5.4. Run AgentCore
• 5.5. Clean Up