Issue with OpenAIAssistantAgent

I’m encountering issues with theOpenAIAssistantAgent for LlamaIndex. Regardless of the query, I’m repeatedly receiving the message, ‘The additional context provided does not contain relevant information to refine the answer to the original query.’ Additionally, the llm seems to have lost its ability to retrieve even basic information. Initially, the performance was pretty good, but after several runs, this unexpected behavior has started to occur.

What I did it: create an separate assistant agent over two query tool for each document

def create_agents_for_each_document(all_query_engine_tools):
    agents = {

    for document, tools in all_query_engine_tools.items():
        company, quarter = company_quarter.split('_')
        agent_name = f"QA Agent - {company} {quarter}"

        agent = OpenAIAssistantAgent.from_new(
            name=agent_name,
            instructions=instructions ,
            tools=tools,
            verbose=True,
            run_retrieve_sleep_time=1.0,
            model="gpt-4-turbo-preview",
        )

        agents[document] = agent

    return agents


agents = create_agents_for_each_document(all_query_engine_tools)