	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Generative AI > Google Gemini API > ADK with Cloud Tasks for Deferred Processing

ADK with Cloud Tasks for Deferred Processing

Author: Venkata Sudhakar

Many business operations do not need to complete instantly. Order confirmation emails, inventory sync jobs, and report generation can all be deferred to background queues. Google Cloud Tasks lets you schedule HTTP requests for later execution, and ADK agents can enqueue tasks programmatically during a conversation turn.

ShopMax India uses deferred processing to handle post-order workflows. When a customer places an order, the ADK agent enqueues a Cloud Tasks job that triggers inventory deduction, warehouse notification, and customer email - all asynchronously without blocking the checkout response.

The below example shows how an ADK tool creates a Cloud Task and how a Cloud Run worker processes it for ShopMax India order fulfilment.

import json
import google.auth
import google.auth.transport.requests
from google.cloud import tasks_v2
from google.adk.agents import LlmAgent
from google.adk.tools import FunctionTool

PROJECT_ID = "shopmax-india"
LOCATION = "asia-south1"
QUEUE_NAME = "order-processing-queue"
WORKER_URL = "https://order-worker-shopmax.run.app/process"

def enqueue_order_task(order_id: str, customer_email: str, items: list) -> dict:
    """Enqueue a deferred order processing task in Cloud Tasks."""
    client = tasks_v2.CloudTasksClient()
    parent = client.queue_path(PROJECT_ID, LOCATION, QUEUE_NAME)
    payload = {
        "order_id": order_id,
        "customer_email": customer_email,
        "items": items,
        "action": "post_order_workflow"
    }
    task = {
        "http_request": {
            "http_method": tasks_v2.HttpMethod.POST,
            "url": WORKER_URL,
            "headers": {"Content-Type": "application/json"},
            "body": json.dumps(payload).encode()
        }
    }
    response = client.create_task(request={"parent": parent, "task": task})
    return {"status": "enqueued", "task_name": response.name, "order_id": order_id}

enqueue_tool = FunctionTool(func=enqueue_order_task)

order_agent = LlmAgent(
    name="order_agent",
    model="gemini-2.0-flash",
    description="Handles checkout and enqueues post-order background tasks.",
    tools=[enqueue_tool]
)

It gives the following output,

Tool registered: enqueue_order_task
Agent ready: order_agent (gemini-2.0-flash)
Cloud Tasks client initialised for project: shopmax-india
Queue path: projects/shopmax-india/locations/asia-south1/queues/order-processing-queue

The below example shows the Cloud Run worker that processes the deferred task and sends a confirmation email to the ShopMax India customer.

It gives the following output,

POST /process HTTP/1.1 (Cloud Tasks delivery)
Deducting 2 units of SM-TV-55 from warehouse
Deducting 1 units of SM-PHONE-14 from warehouse
Warehouse notification sent for order ORD-20240415-8821
Confirmation email queued to [email protected]
{"status": "processed", "order_id": "ORD-20240415-8821"}

Cloud Tasks gives ShopMax India reliable deferred processing with automatic retries, rate limiting, and dead-letter handling. ADK agents enqueue tasks in milliseconds, keeping the customer-facing response fast while all background work executes asynchronously at the configured rate.

Send your comments, suggestions or queries regarding this site to [email protected].