Description
** Please make sure you read the contribution guide and file the issues in the right place. **
Contribution guide.
Describe the bug
We have Google App Engine running on Python 3.12 and React front end. I have wrote a basic chat interface for Agent we created using ADK library and just wanted to add it to that app.
All works as expected on local dev environment we manage sessions, load them, and can have conversation wit the Agent.
We use: google-adk== 1.2.1
The below is examle code how we use deployed Agent to Agent Engine in our App
import vertexai
from vertexai import agent_engines
from vertexai.preview.reasoning_engines import AdkApp
vertexai.init(
project=PROJECT_ID,
location=LOCATION,
staging_bucket=STAGING_BUCKET,
)
adk_app = AdkApp(agent=root_agent)
The issue we are facing is that after deploying to Google App Engine, the App wont load at all.
Checked log App Engine Logs and there was only one info that could mean something:
Exceeded hard memory limit of 768 MiB with 780 MiB after servicing 0 requests total. Consider setting a larger instance class in app.yaml
We used class instance B2 and increased it to B4, but then we got slightly different info:
Exceeded hard memory limit of 1536 MiB with 1542 MiB after servicing 0 requests total. Consider setting a larger instance class in app.yaml.
Once google-adk== 1.2.1 is removed from requirements.txt and adk.py commented out then re-deployed to App Engine, the app loads ok on Google App engine.
Could it be some sort of memory leak on AppEngine where once adk-python is installed it will just eat up all available memory ?
To Reproduce
Deploy basic AdkApp that uses Agent Engine Agent to Google App Engine Standard
Expected behavior
App Engine Application will load.
Desktop (please complete the following information):
- Latest Mac Pro