Module 11 Lesson 1: Action Groups and Schemas
Defining the Capability. How to describe your APIs to a Bedrock Agent using OpenAPI schemas.
Defining the "Hands": Action Groups
An agent with instructions but no tools is just a "Chatter." In Bedrock, we give agents "Hands" using Action Groups. An Action Group is a collection of functions that the agent is allowed to call.
To help the agent understand what each function does, we use an API Schema (OpenAPI/Swagger).
1. How the Agent "Sees" a Tool
The agent reads your schema to understand:
- Function Name: e.g.,
get_weather - Description: e.g., "Retrieves the current temperature for a given city."
- Parameters: e.g.,
city(string).
2. Example: OpenAPI Schema Snippet
{
"name": "get_stock_price",
"description": "Get the current stock price for a ticker symbol",
"parameters": {
"ticker": {
"type": "string",
"description": "The stock ticker (e.g., AMZN, AAPL)"
}
}
}
3. Visualizing Discovery
graph TD
User[What is the price of AMZN?] --> Agent[Agent Brain]
Agent --> S[Read Action Schema]
S -->|Match| Tool[Action: get_stock_price]
Agent -->|Execute| Tool
4. Why Descriptions are Mandatory
In traditional coding, a function name is enough. In AI agents, the Description is what the model uses to "Decide."
- Wrong:
check_db(str text) - Right:
check_customer_record- "Use this function to retrieve the lifetime value and address for a customer ID."
Summary
- Action Groups are the set of tools available to an agent.
- API Schemas define the name, purpose, and parameters of each tool.
- Descriptions are the most important part—they guide the model's reasoning.
- Without a schema, the agent has no way to know your external APIs exist.