Skip to main content

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

Goal​

This file contains the complete, step-by-step guide to creating the "Researcher" agent.

Goal​

You will build a new agent that can answer questions about current events and topics beyond the LLM's training data by giving it the ability to search the web using the built-in google_search tool.

Prerequisites​

  • Note on Costs: Enabling the Vertex AI API and using the google_search tool may incur small costs on your Google Cloud bill. Please be aware of the pricing for these services.

Step 1: Create the Researcher Agent Project​

  1. Navigate to your training directory:

    Open your terminal and make sure you are in the adk-training directory.

    cd /path/to/your/adk-training

  2. Create the agent project (Python approach):

    Use the adk create command to create a new Python agent named researcher_agent.

    adk create --type=python researcher_agent

    Alternative (YAML approach):

    adk create --type=config researcher_agent

To use the google_search tool, you need to enable the Vertex AI API in your Google Cloud project, as this is the service that provides the grounding functionality.

  1. Enable the Vertex AI API:

    • Go to the Vertex AI API page in the Google Cloud Console.
    • Make sure your current project is selected.
    • Click the "Enable" button if it's not already enabled.

  2. Set up your environment variables:

    Navigate into the researcher_agent directory and open the .env file. For the search tool to work, you must configure your agent to use Vertex AI.

    Update the .env file to look like this, filling in your project ID and desired location:

    GOOGLE_GENAI_USE_VERTEXAI=1
    GOOGLE_CLOUD_PROJECT=<your_gcp_project>
    GOOGLE_CLOUD_LOCATION=us-central1
    • Important: The google_search tool does not work with a simple Google AI Studio API key; it requires a full Google Cloud project setup.

  3. Define the agent's behavior and add the tool (Python approach):

    Open the agent.py file inside researcher_agent and replace its contents with the following:

    from google.adk.agents import LlmAgent
    from google.adk.tools import google_search

    root_agent = LlmAgent(
        name="researcher_agent",
        model="gemini-2.5-flash",
        description="An agent that can research current events using Google Search.",
        instruction=(
            "You are a helpful research assistant. "
            "Your job is to answer the user's questions accurately. "
            "If the question is about a recent event, a specific person, or anything that might require up-to-date information, you MUST use the `google_search` tool. "
            "Do not rely on your own knowledge for topics that could have changed since your training."
        ),
        tools=[
            google_search
        ],
    )

    Alternative (YAML approach):

    Open the root_agent.yaml file and replace its contents with the following:

    # yaml-language-server: $schema=https://raw.githubusercontent.com/google/adk-python/refs/heads/main/src/google/adk/agents/config_schemas/AgentConfig.json
    name: researcher_agent
    model: gemini-2.5-flash
    description: An agent that can research current events using Google Search.
    instruction: |
      You are a helpful research assistant.
      Your job is to answer the user's questions accurately.
      If the question is about a recent event, a specific person, or anything that might require up-to-date information, you MUST use the `google_search` tool.
      Do not rely on your own knowledge for topics that could have changed since your training.
    tools:
      - name: google_search

Step 3: Test Your New Agent​

  1. Navigate back to the parent directory:

    cd ..

  2. Start the web server:

    From the adk-training directory, run adk web.

    adk web

  3. Interact with the Researcher Agent:

    • Open the Dev UI in your browser.
    • Select researcher_agent from the dropdown menu.
    • Ask a question that the LLM wouldn't know from its training data:
      • "Who won the last Super Bowl?"
      • "What are the latest headlines about space exploration?"
    • Examine the Trace View: Click on the "Trace" tab. You will see a new step in the execution flow: execute_tool. Expand it to see that the google_search tool was called, confirming the agent is using its new tool correctly.

Lab Summary​

You have successfully given an agent a powerful new capability. You have learned how to:

  • Enable the necessary Google Cloud APIs for advanced features.
  • Configure an agent to use a built-in tool like google_search (using both Python and YAML approaches).
  • Write instructions that guide the agent on when to use its tools.
  • Verify that a tool was used by inspecting the Trace View in the Dev UI.