Bug: `AdkApp` Deployment via `vertexai.Client` Fails to Inherit Location

[eliasecchig]

client.agent_engines.create does not pass the client's location to the AdkApp instance, causing deployment to use an incorrect or default region.

Problem:

The provided code demonstrates a conflict between the new client-based and the older global initialization methods in the Vertex AI SDK. While vertexai.Client is correctly instantiated with location="us-central1", the AdkApp object is created in a way that it remains unaware of this client-scoped configuration.

When client.agent_engines.create(agent=app, ...) is called, the app object does not carry the us-central1 location. Consequently, the deployment process falls back to a default location (or one set by a global vertexai.init call), leading to a regional mismatch and deployment failures. The operation should be self-contained within the client's configuration.

Code to Reproduce:

import vertexai
import asyncio
import logging

# Configure logging
logging.basicConfig(level=logging.INFO)

# --- Problem Area ---
# 1. Client is configured for "us-central1"
project_id = "my-project"
client = vertexai.Client(project=project_id, location="us-central1")

# 2. Agent is defined independently
from google.adk.agents import Agent
root_agent = Agent(
    name="root_agent",
    model="gemini-2.5-flash",
    instruction="You are a helpful AI assistant designed to provide accurate and useful information.",
)

# 3. AdkApp is instantiated without knowledge of the client's location.
# The global init() further complicates which configuration is used.
from vertexai.agent_engines import AdkApp
vertexai.init(project=project_id)
app = AdkApp(agent=root_agent)
# At this point, app's internal configuration does not know about "us-central1"

# 4. Deployment is called on the client, but the `app` object lacks the correct location.
# This is where the regional mismatch occurs.
remote_agent = client.agent_engines.create(
    agent=app,
    config={
        "staging_bucket": "gs://<YOUR_BUCKET_IN_US_CENTRAL1>",
        "requirements": ["google-adk", "google-cloud-aiplatform"]
    }
)

print("Remote agent created successfully:")
print(remote_agent)

# --- Remote Query (will fail if deployment used the wrong region) ---
async def run_remote_query():
    print("\nStreaming remote query...")
    async for event in remote_agent.async_stream_query(message="hi!", user_id="test"):
        print(event)

asyncio.run(run_remote_query())

Expected Behavior:

The client.agent_engines.create() method should ensure that the AdkApp instance (app) it is deploying inherits the configuration of the client object, specifically the location. The deployment should seamlessly proceed in us-central1 as defined in the client.

Actual Behavior:

The AdkApp object is unaware of the client's location. The deployment process initiated by client.agent_engines.create() uses a default or globally configured location, which conflicts with the client's explicit us-central1 setting, causing the deployment to fail.