LlmAgent.mode attribute is forbidden on the model but referenced in runners.py (regression)

[lvth3r]

That's the GitHub issue template — I'll fill it out for you. Paste this into the GitHub issue form:

Describe the Bug:
When deploying an LlmAgent (or Agent) to Vertex AI Agent Engine, every query against the deployed agent returns an empty HTTP response (Content-Length: 0). The underlying error in the agent runtime logs is:
AttributeError: 'LlmAgent' object has no attribute 'mode'. Did you mean: 'model'?
The error originates inside ADK's own runners.py at line 995, which accesses self.agent.mode. However, the LlmAgent Pydantic model rejects mode as an extra/forbidden field meaning the runtime expects a field that the model definition explicitly forbids. This is an internal inconsistency in ADK between the model schema and the runner code.
Steps to Reproduce:

pip install "google-cloud-aiplatform[adk,agent_engines]>=1.153.0" "google-adk>=1.20.0"
Create a minimal agent and deploy to Vertex AI Agent Engine (see Minimal Reproduction Code below)
Query the deployed agent via REST: POST https://us-central1-aiplatform.googleapis.com/v1beta1/{RESOURCE_NAME}:streamQuery?alt=sse with {"classMethod": "stream_query", "input": {"user_id": "test", "message": "Hi"}}
Response is HTTP 200 with Content-Length: 0 and empty body
Check Cloud Logging for the deployed ReasoningEngine's stderr — see the AttributeError traceback

Expected Behavior:
The agent should respond to queries with a streaming response containing the model's reply.
Observed Behavior:
The HTTP response returns immediately with an empty body. The agent never processes the query. Logs show:
File "/code/.venv/lib/python3.12/site-packages/google/adk/runners.py", line 995, in run_async
if self.agent.mode is None:
^^^^^^^^^^^^^^^
File "/code/.venv/lib/python3.12/site-packages/pydantic/main.py", line 1042, in getattr
raise AttributeError(f'{type(self).name!r} object has no attribute {item!r}')
AttributeError: 'LlmAgent' object has no attribute 'mode'. Did you mean: 'model'?
Attempting to add mode="interactive" (or any value) to the Agent(...) constructor fails Pydantic validation:
pydantic_core._pydantic_core.ValidationError: 1 validation error for LlmAgent
mode
Extra inputs are not permitted [type=extra_forbidden, input_value='interactive', input_type=str]
Environment Details:

ADK Library Version (pip show google-adk): latest from >=1.20.0 resolution at time of deploy (mid-June 2026)
Desktop OS: Cloud Shell (Linux)
Python Version (python -V): Python 3.12
Runtime: Vertex AI Agent Engine, us-central1

Model Information:

Are you using LiteLLM: No
Which model is being used: gemini-3.5-flash

Regression:
Yes. An identical agent definition deployed in early June 2026 (with google-cloud-aiplatform==1.112.0) works correctly. The bug appeared after upgrading to google-cloud-aiplatform>=1.153.0.
Logs:
textTraceback (most recent call last):
File "/usr/local/lib/python3.12/threading.py", line 1075, in _bootstrap_inner
self.run()
File "/usr/local/lib/python3.12/threading.py", line 1012, in run
self._target(*self._args, **self._kwargs)
File "/code/.venv/lib/python3.12/site-packages/google/adk/runners.py", line 932, in _asyncio_thread_main
asyncio.run(_invoke_run_async())
File "/usr/local/lib/python3.12/asyncio/runners.py", line 195, in run
return runner.run(main)
File "/usr/local/lib/python3.12/asyncio/runners.py", line 118, in run
return self._loop.run_until_complete(task)
File "/usr/local/lib/python3.12/asyncio/base_events.py", line 691, in run_until_complete
return future.result()
File "/code/.venv/lib/python3.12/site-packages/google/adk/runners.py", line 925, in _invoke_run_async
async for event in agen:
File "/code/.venv/lib/python3.12/site-packages/google/adk/runners.py", line 995, in run_async
if self.agent.mode is None:
^^^^^^^^^^^^^^^
File "/code/.venv/lib/python3.12/site-packages/pydantic/main.py", line 1042, in getattr
raise AttributeError(f'{type(self).name!r} object has no attribute {item!r}')
AttributeError: 'LlmAgent' object has no attribute 'mode'. Did you mean: 'model'?

[mahesh-github]

Can I work on this bug

[llalitkumarrr]

Hello @lvth3r ,

Kindly try specifying the versions during deployment. We successfully deployed a simple agent on the agent engine using the code below and executed curl commands async_create_session and async_stream_query without encountering any errors.

Here is the deployment code for your reference:

# agent.py
import vertexai
from google.adk.agents.llm_agent import Agent
from google.adk.apps import App
from vertexai import agent_engines
from google.adk.tools.function_tool import FunctionTool
import os 

PROJECT_ID = os.getenv("GOOGLE_CLOUD_PROJECT", "PUT_YOUR_PROJECT_ID_HERE")
LOCATION = os.getenv("GOOGLE_CLOUD_LOCATION", "PUT_YOUR_LOCATION_HERE")
STAGING_BUCKET = os.getenv("STAGING_BUCKET", "PUT_YOUR_BUCKET_HERE")

print(f"Initializing Vertex AI with project: {PROJECT_ID}, location: {LOCATION}, staging bucket: {STAGING_BUCKET}") 

vertexai.init(
    project=PROJECT_ID,
    location=LOCATION,
    staging_bucket=STAGING_BUCKET,
)






def tell_weather(city: str) -> str:
    """A function that tells the weather.
    args:
        city (str): The name of the city to get the weather for.
    """
    return f"The weather in {city} is sunny and warm today."

    return 
a2a_agent = Agent(
    model="gemini-2.5-flash",
    name="a2a_agent",
    description="Wether agent that can tell the weather in a city.",
    instruction="You are a helpful assistant that can tell the weather in a city. Use the function tell_weather to get the weather information.",
    tools=[FunctionTool(tell_weather)],)


# Wrap agent into AdkApp
app = agent_engines.AdkApp(
    app_name="sample_agent",
    agent=a2a_agent
)



remote_app = agent_engines.create(
    agent_engine=app,
    requirements=[
        "google-adk==1.35.2",
        "google-cloud-aiplatform[adk,agent_engines]==1.158.0"
    ],
    display_name="test_2",
)

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

Kindly let us know if this resolved your issue, if not feel free to reach out.