API gateways for LLM calls

Most LLM projects start the same way: one OpenAI API key, one SDK, one model. Almost every tutorial on YouTube shows that same setup, and I've started almost every project this way too. For a prototyp