Lab 35: Deploying an Agent to Agent Runtime Challenge

Goal

In this lab, you will deploy the same multi-agent Customer Support system from Modules 32 and 33 to Google Cloud's Agent Runtime, using both the recommended Accelerated method and the manual Standard method — completing the "same application, three platforms" arc for Part 6.

Prerequisites

A Google Cloud Project with billing enabled.
gcloud CLI installed and authenticated (gcloud auth application-default login).
A GCS bucket for the Standard Deployment part (gsutil mb -p YOUR_PROJECT_ID -l us-central1 gs://YOUR_UNIQUE_BUCKET_NAME).
Python Environment: A Python version between 3.9 and 3.13.
UV Tool: For managing the Python environment. See Install UV.
Required APIs: Ensure the following APIs are enabled in your project:
- Vertex AI API
- Cloud Build API
- Artifact Registry API
- Cloud Resource Manager API
Set GCP Project: Before starting, ensure your gcloud CLI is configured to the correct project:
```
gcloud config set project YOUR_PROJECT_ID
```

Part 1: Accelerated Deployment (Recommended)

This method uses the Agents CLI to add deployment artifacts to your existing ADK project and deploy it.

Step 1: Prepare the Agent Project

Re-create the Customer Support Agent: We'll use the same multi-agent Customer Support system from Modules 32 and 33. Create a fresh directory and the same three YAML files:

mkdir support_agent
cd support_agent

billing_agent.yaml:

name: billing_agent
model: gemini-3.5-flash
description: "Handles questions about billing, invoices, and payments."
instruction: "You are a billing support agent. Politely answer questions about billing and payment issues."

tech_support_agent.yaml:

name: tech_support_agent
model: gemini-3.5-flash
description: "Handles technical support questions and troubleshooting."
instruction: "You are a technical support agent. Help users troubleshoot technical issues and provide clear solutions."

root_agent.yaml:

name: router_agent
model: gemini-3.5-flash
description: "The main customer support router."
instruction: |
  You are the customer support router.
  Your job is to understand the user's request and delegate it to the correct specialist agent.
  - If the user has a question about billing, delegate to the `billing_agent`.
  - If the user has a technical problem, delegate to the `tech_support_agent`.
sub_agents:
  - config_path: billing_agent.yaml
  - config_path: tech_support_agent.yaml

Navigate to the Parent Directory: In your terminal, navigate to the parent directory that contains support_agent/.
Scaffold the Project: Run the Agents CLI scaffold enhance command to add the required deployment files to your project. Pass --agent-directory support_agent so the CLI correctly detects your YAML config agent in support_agent/ — without it, the CLI can't find your agent code and silently generates an unrelated generic stub agent instead.
```
uvx google-agents-cli scaffold enhance -d agent_runtime --agent-directory support_agent
```
The command should report Found support_agent/root_agent.yaml (YAML config agent) and generate a support_agent/agent.py shim (via config_agent_utils.from_config(...)) that loads your router_agent and its billing_agent/tech_support_agent sub-agents — that's how you know it picked up the right project.
Follow the Prompts: The tool will ask you several questions. You can accept the defaults, but ensure you select a supported region for Agent Runtime (e.g., us-central1).

Step 2: Connect to Your Google Cloud Project

Login to Google Cloud:
```
gcloud auth application-default login
```

Set Your Project ID:

gcloud config set project your-project-id-xxxxx

Verify the Project:
```
gcloud config get-value project
```

Step 3: Deploy the Agent

Ensure you are in the parent directory (containing support_agent/).
Run the Deployment Command: This command uses the files added by the Agents CLI to provision the cloud infrastructure and deploy your agent. This process can take several minutes.
```
uvx google-agents-cli deploy
```
Find Your Agent: Once the build is complete, navigate to Agent Platform -> Agent Runtime in the Cloud Console to find your deployed agent and its ID.

Part 2: Standard Deployment (Manual)

This method involves writing a custom Python script to deploy the agent. Unlike Part 1, this script needs to import your root_agent directly as a Python object — so this part uses a Python-code version of the same Customer Support system, instead of the YAML config version.

Step 1: Prepare the Agent Project

Create the Python version of the Customer Support agent:

mkdir deploy_manual
cd deploy_manual
mkdir support_agent
touch support_agent/__init__.py

Action: Create support_agent/agent.py, translating the same router + two specialists design from Module 15/32 into Python Agent objects (instead of YAML config). Use the skeleton below and fill in the # TODO sections.

# In support_agent/agent.py
from google.adk import Agent

# TODO: 1. Define billing_agent, an Agent with name="billing_agent",
# the same model/description/instruction as billing_agent.yaml above.

# TODO: 2. Define tech_support_agent, an Agent with name="tech_support_agent",
# the same model/description/instruction as tech_support_agent.yaml above.

# TODO: 3. Define root_agent, an Agent named "router_agent" with the same
# instruction as root_agent.yaml, and sub_agents=[billing_agent, tech_support_agent].

Install Dependencies:

pip install "google-cloud-aiplatform[adk,agent_engines]>=1.111"

Step 2: Create the Deployment Script

In the deploy_manual directory, create a new file named deploy.py.

Action: Write the Python code to deploy the agent. Use the skeleton below and fill in the # TODO sections. You will need to:

Import vertexai, agent_engines, and your root_agent.
Initialize the Vertex AI SDK with your project details.
Wrap your root_agent in an agent_engines.AdkApp.
Call agent_engines.create to deploy the app.

# In deploy.py
import vertexai
from vertexai import agent_engines
from support_agent.agent import root_agent # Make sure this import is correct

# TODO: Fill in these values for your project
PROJECT_ID = "your-gcp-project-id"
LOCATION = "us-central1"
STAGING_BUCKET = "gs://your-gcs-bucket-name"

# Initialize the Vertex AI SDK
vertexai.init(
    project=PROJECT_ID,
    location=LOCATION,
    staging_bucket=STAGING_BUCKET,
)

# Wrap the agent in an AdkApp object
app = agent_engines.AdkApp(agent=root_agent)

# TODO: Call agent_engines.create() to deploy the app.
# Pass the `app` object to the `agent_engine` parameter.
# Also provide a `requirements` list: ["google-cloud-aiplatform[adk,agent_engines]"]
remote_app = None # Replace this

print(f"Deployment finished!")
print(f"Resource Name: {remote_app.resource_name}")

Step 3: Test Agent Locally (Optional but Recommended)

Action: Before deploying, add the local testing code from the lab-solution.md to your deploy.py script (before the deployment call) to test the AdkApp locally. This helps you catch errors before starting the lengthy deployment process.

Step 4: Deploy the Agent

Run the deployment script. This will take several minutes.

python deploy.py

Step 5: Interact with the Deployed Agent

Create an interact.py script (code available in lab-solution.md).
Action: Configure the script with your PROJECT_ID, LOCATION, and the AGENT_ENGINE_ID from the deployment output.
Run the script to test your deployed agent — try a billing question and a technical question, and confirm each routes to the correct specialist, exactly like in Modules 32 and 33:
```
python interact.py
```

Cleanup (Important!)

Follow the cleanup instructions in lab-solution.md to delete the Agent Runtime instances and GCS buckets to avoid incurring costs.

Self-Reflection Questions

What are the primary advantages of using the Accelerated Deployment method with the Agents CLI compared to the Standard Deployment method for production use?
Agent Runtime is a managed backend. How does this simplify the development of complex clients (e.g., web or mobile applications) that interact with your agent?
For what scenarios might the Standard Deployment method (using deploy.py and the Vertex AI SDK) still be advantageous, even if Accelerated Deployment is generally recommended?

🕵️ Hidden Solution 🕵️

Looking for the solution? Here's a hint (Base64 decode me): L2RvYy1hZGstdHJhaW5pbmcvbW9kdWxlMzUtZGVwbG95bWVudC1hZ2VudC1ydW50aW1lL2xhYi1zb2x1dGlvbg==

The direct link is: Lab Solution

Goal​

Prerequisites​

Part 1: Accelerated Deployment (Recommended)​

Step 1: Prepare the Agent Project​

Step 2: Connect to Your Google Cloud Project​

Step 3: Deploy the Agent​

Part 2: Standard Deployment (Manual)​

Step 1: Prepare the Agent Project​

Step 2: Create the Deployment Script​

Step 3: Test Agent Locally (Optional but Recommended)​

Step 4: Deploy the Agent​

Step 5: Interact with the Deployed Agent​

Cleanup (Important!)​

Self-Reflection Questions​

🕵️ Hidden Solution 🕵️​

Goal

Prerequisites

Part 1: Accelerated Deployment (Recommended)

Step 1: Prepare the Agent Project

Step 2: Connect to Your Google Cloud Project

Step 3: Deploy the Agent

Part 2: Standard Deployment (Manual)

Step 1: Prepare the Agent Project

Step 2: Create the Deployment Script

Step 3: Test Agent Locally (Optional but Recommended)

Step 4: Deploy the Agent

Step 5: Interact with the Deployed Agent

Cleanup (Important!)

Self-Reflection Questions

🕵️ Hidden Solution 🕵️