[BUG]: Inifinite loop when using code_executor #3921
Description
Describe the bug
When the LLM returns a code block to execute, it goes to an infinite loop and never returns. This used to work in v1.16.0 and doesn't work on main.
To Reproduce
Run this agent:
from google.adk import Agent
from google.adk.code_executors import UnsafeLocalCodeExecutor
root_agent = Agent(
model='gemini-2.0-flash',
name='hello_world_agent',
description=(
'hello world agent.'
),
instruction="""
you are an agent that knows how to produce python code.
""",
code_executor=UnsafeLocalCodeExecutor(),
)and give a prompt that would make it return a code block. For example:
write python code that computes 2+2.You will see this repeats ∞ times in the logs.
2025-12-15 13:50:54,034 - INFO - _client.py:1740 - HTTP Request: POST https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash:generateContent "HTTP/1.1 200 OK"
2025-12-15 13:50:54,034 - INFO - google_llm.py:246 - Response received from the model.
2025-12-15 13:50:54,076 - INFO - google_llm.py:181 - Sending out request, model: gemini-2.0-flash, backend: GoogleLLMVariant.GEMINI_API, stream: False
2025-12-15 13:50:54,077 - INFO - models.py:7039 - AFC is enabled with max remote calls: 10.
2025-12
Expected behavior
Agent returns the executed code.
Screenshots
N/A
Desktop (please complete the following information):
- OS: Linux Fedora 43, Docker (python:3.13-slim)
- Python Python 3.14.2 and 3.13
- ADK version: this bug happened when i updated to v1.18. also tested on 1.20 and 1.21.
Model Information: - Are you using LiteLLM: Yes/No
- Which model is being used(e.g. gemini-2.5-pro)
Additional context
The issue seems to happen because the code execution event from the session doesn't make it back to the LLM when the content is prepared in