When building a system around OAI, I found myself sending the same request multiple times as part of developing/testing some other part of the system. On top of wasting money in this way, I was also throwing away potentially useful later training data to specialize a smaller LLM for my use case.<p>I’m hosting an open server atm since I hit it from various different networks for my projects, or you easily enough run it as a local service.