Temp prefixed state key not available on after_agent_callback #3047
Description
Describe the bug
The documentation states that state variables prefixed with temp: should be available during the entire invocation process:
temp: Prefix (Temporary Invocation State):
Scope: Specific to the current invocation (the entire process from an agent receiving user input to generating the final output for that input).
However, if the agent output_key is set with a temp: prefixed key, this key is not present anymore in the state during after_agent_callback
To Reproduce
The following code works as expected
from google.adk.agents import LlmAgent
from google.adk.agents.callback_context import CallbackContext
from google.genai import types
def _callback_test(callback_context: CallbackContext) -> types.Content | None:
print(callback_context.state["test_output"])
return None
root_agent = LlmAgent(
name="test_agent",
model="gemini-2.5-flash",
instruction="You are a helpful assistant that can generate a 10 words random text.",
output_key="test_output",
after_agent_callback=_callback_test,
)
But the following code fails, because it can't find the key in the after_agent_callback execution
from google.adk.agents import LlmAgent
from google.adk.agents.callback_context import CallbackContext
from google.genai import types
def _callback_test(callback_context: CallbackContext) -> types.Content | None:
print(callback_context.state["temp:test_output"])
return None
root_agent = LlmAgent(
name="test_agent",
model="gemini-2.5-flash",
instruction="You are a helpful assistant that can generate a 10 words random text.",
output_key="temp:test_output",
after_agent_callback=_callback_test,
)
Expected behavior
Temp keys should remain available during the entire invocation process
Desktop (please complete the following information):
- OS: windows, mac
- Python version: 3.13
- ADK version: 1.15.1
Model Information:
- Are you using LiteLLM: No
- Which model is being used: gemini-2.5-flash