Учёт расхода токенов и затрат ИИ-агентов в Sheets

## Limitations ### 1. This workflow doesn't account for Prompt caching If you're consecutively sending similar prompts to OpenAI it'll automatically use Cached Tokens to reduce the cost of your requests. So our cost estimates will have a higher value than the actual cost. Other providers like Anthropic and Google have similar mechanisms. ### 2. Not tested with audio or video files This workflow was tested with text and images, but no tests were made with audio files or videos. ### 3. The cost is an estimate In the spreadsheet you can see the total cost of the requests, but that is only an estimate. If you do batch requests, prompt caching or other techniques to reduce cost, the estimate might be higher than the actual cost.