I replaced the part that used the OpenAI model in AgentMemory with Gemini.
memory = AgentMemory(
summarizer=MemorySummarizer(model=gemini),
manager=MemoryManager(model=gemini),
classifier=MemoryClassifier(
model=gemini,
),
db=PgMemoryDb(
db_url="postgresql+psycopg://postgres:psw@localhost:5432/agno_storage",
table_name="memory_test",
schema="ai",
),
create_user_memories=True,
update_user_memories_after_run=True,
create_session_summary=True,
update_session_summary_after_run=True,
)
Then I built the agent and used it in the playground.
agent = Agent(
model=gemini,
memory=memory,
enable_agentic_memory=True,
enable_user_memories=True,
markdown=True,
monitoring=True,
)
playground_app = Playground(agents=[agent])
app = playground_app.get_app()
if __name__ == "__main__":
playground_app.serve("monitor:app",port=7777, reload=True)
Problems arose when I was happily conversing with my Agent in the playground.
Each time the Agent discovers that an update to the memory is necessary, it will repeat the answer twice in the dialogue box.
If memory is not updated, everything is harmonious and normal.
Is there a feasible way to solve this problem?