How Are Indirect LLM Call Costs, Like Web Search, Reflected and Planned for Budgeting?

Hey everyone, i have an overall question that is connected to estimation of costs for indirect calls of llm, like when google uses google search or url context, are these reflected inside of the models costs or how does this setup work? if not are there any plans either for openinference or for phoenix itself to implement this costs determination? My current workaround is by using custom attributes but i assume that there might be an overall usage as large companies have these privately, mainly the websearch for OpenAI and google and the implementation locally maybe makes no sense but it would be a good use case for budgeting to determine where are the costs then going. Thank you for the response Cheers