Skip to main content

Lab 8: Creating a "Researcher" Agent with Google Search Challenge

Goal​

Your task is to build and configure a new agent that can search the web to answer questions about current events.

Lab Tasks​

  1. Create a new agent named researcher_agent using the uv run adk create command, specifying the Python type.

  2. Enable the Vertex AI API in your Google Cloud project.

  3. Configure the .env file inside the researcher_agent directory to use Vertex AI, providing your project ID and a location.

  4. Modify the agent.py file (Python approach):

    # In agent.py
    from google.adk import Agent
    from google.adk.tools import google_search

    # TODO: Define the root_agent node
    # - name: "researcher_agent"
    # - model: "gemini-3.5-flash"
    # - instruction: Tell it to use search for current events.
    # - tools: Add the google_search tool.
    root_agent = Agent(...)

  5. Run the agent from your main adk-training directory using the uv run adk web command (without specifying the agent name).

  6. Test the agent by asking it a question about a recent event (e.g., "Who won the last major sports championship?").

  7. Verify that the google_search tool was used by inspecting the "Trace" view in the Dev UI.

Self-Reflection Questions​

  • Why is it important to explicitly instruct the agent when to use the google_search tool? What might happen if you just gave it the tool with no instructions?
  • The google_search tool requires an Agent Platform configuration. Why do you think this is a requirement, as opposed to using a simple Google AI Studio API key?
  • How does giving an agent access to real-time information fundamentally change the kinds of problems it can solve compared to an agent that only relies on its internal knowledge?

🕵️ Hidden Solution 🕵️​

Looking for the solution? Here's a hint (Base64 decode me): L2RvYy1hZGstdHJhaW5pbmcvbW9kdWxlMDgtaW50cm8tdG8tdG9vbHMvbGFiLXNvbHV0aW9u

The direct link is: Lab Solution