Hey everyone, I saw the announcement about Traceloop flagging high-cost prompts now. That sounds super useful for keeping cloud bills down 😅.
Has anyone tried it yet? I'm nervous about trusting it fully. Like, does it actually catch the real expensive calls, or does it give a lot of false positives? I work mostly with AWS and Terraform, so I'm wondering if it's tuned for common IaC patterns. Any examples of what it flagged for you would be awesome.
I've been testing it for a week. It flagged a few calls to list all instances across every region as high cost, which was correct. But it also flagged some simple DescribeVolumes calls that I know are cheap. So yeah, some false positives.
For Terraform, it seems okay at spotting loops that could generate massive API calls, but I wouldn't rely on it alone. You still need to check the logic yourself.
Have you run it against any of your existing code yet? I'm curious if your experience matches mine.
Oh, that's a great question. I'm new to monitoring costs directly in the pipeline like this. I haven't tried the Traceloop feature yet, but your point about it being tuned for IaC patterns makes a ton of sense. A false positive on a DescribeVolumes call would be annoying.
I wonder if the accuracy depends a lot on how you've configured your cost profile, like assigning the right weights to different API services? Or maybe it's better at spotting certain patterns, like loops in Terraform as mentioned, versus judging individual calls. Have you found any docs on how it actually calculates the "high cost" risk?
rookie