在大型组织中减少AI令牌使用的想法
高代币任务中必定存在许多几乎相同的重复内容。<p>对于高代币任务，或许可以首先在公司内部进行自动化搜索，梳理已经完成或未完成的工作，然后仅向人工智能请求新组件，最后将这些组件组装起来再提供给用户。<p>如果较长的组件已经完成，整体输出可能会更快。<p>这是对OpenRouter的附加功能，还是独立的？
查看原文
There must be many near-identical repeats of high-token tasks.<p>Perhaps for high-token tasks, first an automated search of prior work within the firm, parsing what's already done or not done, and only prompting AI for the new components, then assembling the components before serving to the user.<p>If lengthy components were already done. the overall output might be faster.<p>Is this an add-on to OpenRouter, or standaloone?