Propagation

def create_initial_state(self, company_name: str, trade_date: str) -> AgentState:
    """Create the initial AgentState for the graph execution.

    Args:
        company_name (str): The name of the company or ticker symbol.
        trade_date (str): The trade date in YYYY-MM-DD format.

    Returns:
        AgentState: The initialized agent state.
    """
    try:
        parsed_date = date.fromisoformat(str(trade_date))
    except ValueError as exc:
        raise ValueError(f"trade_date must be in YYYY-MM-DD format: {trade_date!r}") from exc
    today = date.today()
    if parsed_date > today:
        raise ValueError(f"trade_date cannot be in the future: {parsed_date} > {today}")

    return AgentState(
        messages=[HumanMessage(content=company_name)],
        company_of_interest=company_name,
        trade_date=parsed_date.strftime("%Y-%m-%d"),
    )

def get_graph_args(self, callbacks: list | None = None) -> dict[str, Any]:
    """Get arguments for the graph invocation.

    Note: LLM callbacks are handled separately via LLM constructor.

    Args:
        callbacks (list | None, optional): Optional list of callback handlers for tool execution tracking. Defaults to None.

    Returns:
        dict[str, Any]: A dictionary containing stream mode and config arguments for graph execution.
    """
    config: dict[str, Any] = {"recursion_limit": self.max_recur_limit}
    if callbacks:
        config["callbacks"] = callbacks
    return {"stream_mode": "values", "config": config}