I know OpenPipe is popular for making LLM workflows easier, but I'm worried about the "low-code" promise. In my AWS experience, when you hide the infrastructure, you can get stuck when something breaks.
For example, if a prompt change suddenly costs 10x more or times out, how do you debug it? With Terraform, I can trace everything. With a low-code layer, I feel like I'm just hoping it works. Has anyone hit a wall where the abstraction leaked and you needed to see the actual API calls? 😅
Maybe I'm just used to IAC, but I need to see the moving parts. Like, if I was building a retrieval system, I'd want to define the security groups and VPC endpoints myself for cost and control. Does OpenPipe give you that visibility, or is it all managed behind their UI?