Module 11 Wrap-up: The Toolmaker

You have moved from a "Reasoning" brain to a brain that has "Hands." You understand the relationship between an OpenAPI Schema (which describes what a tool does) and a Lambda Function (which actually does the work). You are now ready to build agents that solve real-world business automation tasks.

---

Hands-on Exercise: The Delivery Tracker

1. The Goal

Create a logic plan for an agent that can track a Fedex/UPS package.

2. The Task

1. Write the Function Description: (e.g., "Use this tool to convert a human tracking ID into a status string from our logistics database.")
2. Define the Parameters: (e.g., tracking_number - string).
3. Draft the Lambda logic: Write a simple Python function that returns a hardcoded status like "In Transit" if the ID starts with "123" and "Delivered" otherwise.

---

Module 11 Summary

• Action Groups: Collections of tools the agent can use.
• API Schemas: The "Instruction Manual" for each tool.
• AWS Lambda: The compute engine that executes the tool logic.
• JSON Response: The specific format required to send data back to the agent brain.
• Discovery: How the agent uses reasoning to match a user query to a schema description.

Coming Up Next...

In Module 12, we put it all together. We will look at Agent Reasoning—how the agent plans out a multi-step mission, recovers from errors, and finally delivers an answer after using multiple tools.

---

Module 11 Checklist

• I can describe the purpose of an OpenAPI schema in Bedrock.
• I have written a small lambda_handler that parses parameters.
• I understand why descriptions in the schema are critical for AI.
• I know how to grant Bedrock permission to call my Lambda.
• I have identified the TEXT or JSON output format for a tool response.