graph invoke always throws error No message found in input

[malikusman]

I have created a Graph and added node (a tool) and another node (as a agent)

Some how its throwing error no matter what and how i query.

async def create_dynamic_agent_endpoint(data: dict):
    try:
         (some code)
        workflow = create_workflow(workflow_id, user.id, previous_conversations)
      
        response = workflow.invoke(query)

        save_conversation_history(db, user_id=user.id, query=query, response=response.get("output", ""))

        return {"response": response.get("output", ""), "products": parse_products(response.get("output", ""))}

    except Exception as e:
        logger.error(f"Error in create_dynamic_agent_endpoint: {e}")
        raise HTTPException(status_code=500, detail=str(e))

Above is the main where i am calling the function to create a graph

    def create_workflow(workflow_id: int, user_id: int, previous_conversations: Optional[List[dict]] = None) -> Graph:
    """
    Create and return a compiled LangGraph workflow for the given workflow and user.
    """
    print(f"in create_dynamic_workflow - workflow id: {workflow_id}")
    
    workflow = build_workflow(workflow_id, user_id, previous_conversations)
    
    print(f"workflow is: {workflow}")

    compiled_workflow = workflow.compile()

    return compiled_workflow

And this is how i am initializing graph

def build_workflow(workflow_id: int, user_id: int, previous_conversations: Optional[List[dict]] = None) -> Graph:
    sql_db = SQLDatabase(
        engine,
        ignore_tables=["malls", "products", "stores"]
    )

    memory = create_user_specific_memory(user_id, previous_conversations)

    llm = ChatOpenAI(model_name="gpt-4o-mini", temperature=0.0)

    dynamic_tools = generate_tools(workflow_id)

    toolkit = SQLDatabaseToolkit(db=sql_db, llm=llm)
    sql_tools = toolkit.get_tools()

    all_tools = dynamic_tools + sql_tools

    prompt = generate_dynamic_prompt(workflow_id, user_id)

    agent = create_openai_functions_agent(
        llm=llm,
        tools=all_tools,
        prompt=prompt
    )
    agent_executor = AgentExecutor(agent=agent, tools=all_tools)

    workflow = Graph()
    
    
    def initialize_state(input_data):
        print(f"initialze state: {input_data}")
        initial_state = {
            "input": input_data, 
            "results_found": False,
            "intermediate_results": None
        }
        return initial_state


    workflow.add_node("start", initialize_state)
    workflow.add_node("embedding_search", ToolNode(tools=dynamic_tools))
    workflow.add_node("sql_query", ToolNode(tools=sql_tools))
    workflow.add_node("agent", agent_executor)

    def route(state):
        if not state.get("results_found", False):
            return "sql_query"
        return "agent"

    workflow.add_edge("start", "embedding_search")
    workflow.add_conditional_edges("embedding_search", route)
    workflow.add_edge("sql_query", "agent")
    workflow.add_edge("agent", END)
    workflow.set_entry_point("start")

    return workflow

I am always getting `No message found in input` error (500)

where i am wrong?