I have a question about structure outputs vs stucture data and costs. I've just created a new voice agent and it appears we are now being forced to use structred outputs as opposed to structured data. That's fine and I can understand the reasoning, however, in my experience structured data worked just fine. Now using the new system I can see my call costs are almost doubled, with 40% of those costs now attributable to processing structured output. Am I doing something wrong? Was it the case we were getting the API costs for processing the structured data for free, but now you're looking to claim those costs back. Hopefully this is just an issue my side, but I thought it worth asking.