Adk Function Call Warning

[vernondmonte]

🔴 Required Information

Please ensure all items in this section are completed to allow for efficient
triaging. Requests without complete information may be rejected / deprioritized.
If an item is not applicable to you - please mark it as N/A

Describe the Bug:
When an ADK Agent invokes a registered tool, the model returns a response containing a function_call part. The ADK framework internally accesses .text on this response (likely for logging or introspection). The google.genai SDK fires a UserWarning from types.py whenever .text is accessed on a response that contains non-text parts. The warning is emitted on every single tool call, flooding logs with noise. The tool itself executes correctly — the warning is a false positive caused by an internal .text access at the ADK call site.

Steps to Reproduce:
Please provide a numbered list of steps to reproduce the behavior:

1. Install pip install google-adk google-genai google-cloud-aiplatform
2. Create an ADK Agent with a registered async tool (see Minimal Reproduction Code below)
3. Set GOOGLE_GENAI_USE_VERTEXAI=1 and valid GOOGLE_CLOUD_PROJECT / GOOGLE_CLOUD_LOCATION env vars
4. Run adk web and ask the agent a question that triggers the tool call
5. Observe the following warning printed to the console on every tool invocation:
   Github Bug Report Adk Function Call Warning
   less than a minute ago

Review
🔴 Required Information
Please ensure all items in this section are completed to allow for efficient triaging. Requests without complete information may be rejected / deprioritized. If an item is not applicable to you - please mark it as N/A

Describe the Bug: When an ADK
Agent
invokes a registered tool, the model returns a response containing a function_call part. The ADK framework internally accesses .text on this response (likely for logging or introspection). The google.genai SDK fires a UserWarning from types.py whenever .text is accessed on a response that contains non-text parts. The warning is emitted on every single tool call, flooding logs with noise. The tool itself executes correctly — the warning is a false positive caused by an internal .text access at the ADK call site.

Steps to Reproduce:

Install pip install google-adk google-genai google-cloud-aiplatform
Create an ADK
Agent
with a registered async tool (see Minimal Reproduction Code below)
Set GOOGLE_GENAI_USE_VERTEXAI=1 and valid GOOGLE_CLOUD_PROJECT / GOOGLE_CLOUD_LOCATION env vars
Run adk web and ask the agent a question that triggers the tool call
Observe the following warning printed to the console on every tool invocation:

WARNING - types.py:6814 - Warning: there are non-text parts in the response:
['function_call'], returning concatenated text result from text parts.
Check the full candidates.content.parts accessor to get the full model response.

Expected Behavior:
No warning should be emitted during tool invocation. The ADK framework should check response.candidates[0].content.parts directly when working with responses that may contain function_call parts, rather than accessing the .text property which is documented to warn on mixed-part responses.

Observed Behavior: A UserWarning is emitted on every tool call via the google.genai types.py:6814 .text property getter. The warning originates inside the ADK framework (google_llm.py), not in user code. The tool call completes successfully regardless — the warning is noise only.

Observed Behavior:
What actually happened? Include error messages or crash stack traces here.

Environment Details:

ADK Library Version (pip show google-adk): 1.26.0
Desktop OS: Windows 11
Python Version (python -V): 3.13.0

Model Information:

• Are you using LiteLLM: No
• Which model is being used: gemini-2.0-flash

---

🟡 Optional Information

Providing this information greatly speeds up the resolution process.

Regression:
have not tested against earlier ADK versions.

Logs:
Please attach relevant logs. Wrap them in code blocks (```) or attach a
text file.
``2026-03-03 11:54:10,767 - INFO - google_llm.py:185 - Sending out request, model: gemini-2.0-flash, backend: GoogleLLMVariant.VERTEX_AI, stream: False
2026-03-03 11:54:14,921 - INFO - google_llm.py:250 - Response received from the model.
2026-03-03 11:54:14,922 - WARNING - types.py:6814 - Warning: there are non-text parts in the response: ['function_call'], returning concatenated text result from text parts. Check the full candidates.content.parts accessor to get the full model response.

**Screenshots / Video:**
If applicable, add screenshots or screen recordings to help explain
your problem.

**Additional Context:**
The .text property in google.genai.types is a convenience accessor that concatenates text parts and explicitly warns when non-text parts (such as function_call) are present. The ADK framework should not be accessing this property on responses it knows may contain function calls. The internal call site should instead iterate over response.candidates[0].content.parts directly, or guard with a check before calling .text.

Suggested fix at the ADK call site in google_llm.py:

# Instead of (triggers warning when parts contain function_call):
text = response.text

# Use (safe, no warning):
parts = response.candidates[0].content.parts if response.candidates else []
text = "".join(p.text for p in parts if hasattr(p, "text") and p.text)


**Minimal Reproduction Code:**
Please provide a code snippet or a link to a Gist/repo that isolates the issue.
```import os
from google.adk.agents import Agent
from google.adk.tools import ToolContext
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService
from google.genai import types

async def my_tool(question: str, tool_context: ToolContext) -> str:
    """A simple tool that answers questions."""
    return f"Answer to: {question}"

root_agent = Agent(
    model=os.getenv("ROOT_AGENT_MODEL", "gemini-2.0-flash"),
    name="test_agent",
    instruction="Use my_tool to answer all questions.",
    tools=[my_tool],
)

session_service = InMemorySessionService()
session = session_service.create_session(app_name="test", user_id="user1")

runner = Runner(
    app_name="test",
    agent=root_agent,
    session_service=session_service,
)

content = types.Content(role="user", parts=[types.Part(text="What is 2+2?")])
for event in runner.run(user_id="user1", session_id=session.id, new_message=content):
    # WARNING fires here on every tool invocation:
    # "Warning: there are non-text parts in the response: ['function_call']"
    pass

How often has this issue occurred?:

• Always (100%)

[surajksharma07]

@vernondmonte The warning originates in _build_response_log (google_llm.py:567), where resp.text is accessed even when the response contains only function_call parts — the genai SDK is built to warn in exactly that case.

Worth trying a workaround of iterating resp.candidates[0].content.parts directly instead of using .text, which avoids triggering the accessor altogether; could you give that a shot and let us know if it stops the flooding on your end?

There's also already a community PR (#4689) targeting this exact spot — probably worth validating the workaround against your setup before raising anything separately.