Intro to RAG with Amazon AgentCore

In this hands-on workshop, you'll go from raw embeddings to a fully deployed, production-ready RAG agent on AWS — no prior RAG or AgentCore experience needed.
We'll build a real document Q&A system, stage by stage:

• Basic RAG from scratch — embed documents yourself with Amazon Titan Text Embeddings V2, store them in a local FAISS index, retrieve with cosine similarity, and generate answers with Claude.
• Bedrock Knowledge Bases — swap the DIY pipeline for a fully managed service that handles chunking, embedding, and indexing (OpenSearch Serverless) from an S3 data source, with hybrid search and metadata filtering. Compare the results against your hand-built version.
• AgentCore Agent — wrap your RAG into a Strands agent, containerize it, push to ECR, and deploy to AgentCore Runtime with session isolation, auto-scaling, and streaming.
• Production hardening — add AgentCore Memory for conversation continuity, instrument with OpenTelemetry for traces in CloudWatch, run a RAG evaluation suite, and expose your knowledge base as a Gateway MCP tool.

By the end you'll have a deployed, invokable RAG agent on AgentCore Runtime — with memory, full observability, and a real feel for the tradeoffs at every layer, from local FAISS to managed production.
What to bring:

• Laptop with Python 3.11+ (managed via uv) and AWS CLI v2 installed
• An AWS account

Skill level: Intermediate — some Python and basic AWS familiarity helpful. No prior RAG or agent development experience required.