Module 20 Wrap-up: The High-Fidelity Assistant
Hands-on: Design an AgentCore workflow that uses a critic node to verify RAG output.
Module 20 Wrap-up: The Truth Seeker
You have reached the final technical milestone. You have combined the Search power of Knowledge Bases with the Control power of AgentCore. You understand that "Intelligent Systems" aren't just one API call—they are Orchestrations that prioritize accuracy and verification.
Hands-on Exercise: The Fact Checker
1. The Scenario
You are building an AI for a pharmaceutical company to answer questions about drug interactions. The margin for error is ZERO.
2. The Task
Design an AgentCore graph with 4 Nodes:
- Search Node: Finds the relevant pharmaceutical PDF.
- Synthesis Node: Generates a summary of the interaction.
- Fact-Check Node: A smart model (Sonnet 3.5) compares the summary against the PDF.
- Formatting Node: Converts the final verified text into a clean medical report.
How does this prevent a "Guess" if the PDF doesn't mention the drug?
Module 20 Summary
- Controlled RAG: Using refinement and verification nodes to improve search.
- Proposer-Critic: Building a multi-model logic team.
- Accuracy over Speed: Prioritizing verification for enterprise safety.
- AgentCore: The framework that makes this multi-stage logic possible.
Coming Up Next...
It's time. You have the foundations. You have the tools. You have the security and scaling strategies. It is time for your Capstone Project: Building an Enterprise AI Assistant on AWS Bedrock from scratch.
Module 20 Checklist
- I can describe the "Proposer-Critic" pattern.
- I understand how a "Refinement Node" improves Knowledge Base search.
- I have identified why multi-stage reasoning is more reliable than a single model.
- I know how to use AgentCore to link a model to a Knowledge Base.
- I am ready to design my final Capstone project!