How do you do analytics on API usage for your products?

I personally don’t like the “usage” tab of the OpenAI analytics page. I personally would prefer something more granular, but would love to get a better idea from everyone here in case this is already a solved problem.

How do you all do analytics? Do you dump log to S3 buckets? Do you log API calls into a database?

Also, what metrics do you track? Things like # of requests, common prompts, latency, etc?

Hoping to better understand what good analytics looks like for the OpenAI API!

1 Like

Welcome to the forum!

Yeah, they used to have more data, but the page would take forever to load, and I bet it was hard on the servers. That said, I’m not tracking all API call data personally.