How I Built a Real-Time AI Budget Cap with Redis and PHP (Under 1ms)

When your AI proxy handles thousands of LLM calls per day, you need budget enforcement that doesn't add latency. A database query on every request isn't going to cut it. Here's how I built a sub-milli