Skip to main content

Lab 3 Challenge: Build and Run the "Echo" Agent

Goal​

Your task is to create, configure, and run a simple "Echo" agent using the ADK.

The Challenge: Unlike a standard chatbot, this agent must act like a parrot. It should never answer questions or provide information; it must only repeat the user's input exactly as it was received.

Expected Behavior​

User InputAgent Response (Correct)Agent Response (Wrong)
"Hello!""Hello!""Hi there, how can I help you?"
"What is the capital of France?""What is the capital of France?""The capital of France is Paris."
"12345""12345""You entered the numbers 1 through 5."
e

Requirements​

  1. Use the adk create command to scaffold a new agent named echo_agent.
  2. Follow the Python Approach below to define the agent's behavior.
  3. Instruction Strategy: Craft an instruction that forces the agent to only echo and explicitly forbids it from answering questions or being helpful.
  4. Configure the .env file with your Google API key or Google Cloud project details.
  5. Run the agent using the adk web command.
  6. Interact with the agent in the Dev UI to verify it passes the "Expected Behavior" tests.

Python Approach (Primary)​

Modify your agent.py file to look like the following, filling in the TODO sections.

from google.adk.agents import LlmAgent

# TODO: Define the root_agent for your application.
# You will need to provide a name, a model, and the instruction for the agent.
root_agent = LlmAgent(
    name=...,  # TODO: Give your agent a name (e.g., "echo_agent")
    model=..., # TODO: Specify the model to use (e.g., "gemini-2.5-flash")
    instruction=...,  # TODO: Provide the instruction for the echo agent.
)

Alternative Approach: Using YAML Configuration​

If you prefer a simpler, config-based agent, you can use a YAML file instead of Python.

  1. Create your agent using adk create --type=config echo_agent. This will create a root_agent.yaml file instead of agent.py.
  2. Fill in the TODO sections in your root_agent.yaml file.
# The first line is an optional schema definition that provides
# auto-completion and validation in compatible code editors.
# yaml-language-server: $schema=https://raw.githubusercontent.com/google/adk-python/refs/heads/main/src/google/adk/agents/config_schemas/AgentConfig.json

# TODO: Define the root_agent for your application.
name: ... # TODO: Give your agent a name (e.g., "echo_agent")
model: ... # TODO: Specify the model to use (e.g., "gemini-2.5-flash")
instruction: ... # TODO: Provide the instruction for the echo agent.

Note: If both agent.py and root_agent.yaml exist, the ADK will use the root_agent.yaml file.

Self-Reflection Questions​

  • What are the advantages of defining an agent in a Python script versus a YAML file?
  • Why is it important to keep API keys and other secrets in a .env file instead of directly in your agent's code?
  • Explore the "Events" tab in the Dev UI after running your agent. What information does it provide, and how could this be useful for debugging a more complex agent?

🕵️ Hidden Solution 🕵️​

Looking for the solution? Here's a hint (Base64 decode me): L2RvYy1hZGstdHJhaW5pbmcvbW9kdWxlMDMtZmlyc3QtYWdlbnQtZWNoby9sYWItc29sdXRpb24=

The direct link is: Lab Solution