self.search_memory() method does not work inside an agent #2543
Description
Describe the bug
I tried to use self.search_memory() for agent to keep user preferences. I followed the code in ADK docs here in official Google ADK docs website.
To Reproduce
Just copy and paste the first code snippet in the documentation page.
from google.adk.agents import Agent
from google.genai import types
class MyAgent(Agent):
async def run(self, request: types.Content, **kwargs) -> types.Content:
# Get the user's latest message
user_query = request.parts[0].text
# Search the memory for context related to the user's query
search_result = await self.search_memory(query=user_query)
# Create a prompt that includes the retrieved memories
prompt = f"Based on my memory, here's what I recall about your query: {search_result.memories}\n\nNow, please respond to: {user_query}"
# Call the LLM with the enhanced prompt
return await self.llm.generate_content_async(prompt)
The code in the second documentation snippet works correctly (instantiating VertexAiMemoryBankService inside the Agent). I.e. replacing search_result... line above with this snippet fixes the problem.
# print(f"User query: {user_query}")
# self.vertexai_memorybank_service = VertexAiMemoryBankService(
# project=PROJECT_ID,
# location=LOCATION,
# agent_engine_id=ENGINE_NAME,
# )
# search_result = await self.vertexai_memorybank_service.search_memory(query=user_query)
Expected behavior
Based on the docs, I expect the self.search_memory() to be able to retrieve user preferences recorded by the runner. The first example in the docs is unable to do that, while the second example does it fine.
Desktop (please complete the following information):
OS: Mac Sequoia 15.4
Python version(python -V): 3.12.15
ADK version(pip show google-adk): 1.10.0
Model Information:
gemini-2.5-flash and VertexAiMemoryBankService (both enterprise)