New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Nothing changed in my code, so why do the model's answers look different this month?
What you'll have at the end
Model calls pinned to one dated version, with a saved output set that flags a real change
You need
An LLM feature already live in production, calling the model by a plain or floating name from your own code or config, with the ability to change that call and pull a handful of the real inputs it already handles.
Not covered
Building a full evaluation framework or a CI gate for every prompt edit; this keeps one already-pinned model stable and catches the moment its behavior actually shifts.
Roll back a prompt change that broke your output
Go there if the answers changed right after you edited the prompt yourself, not after the provider changed anything.
See your LLM's cost, latency, and errors on one dashboard
Use this first if you have no way yet to notice a quality problem before a user reports it.
Checked 15 Aug 2026
Part of the LLMOps cookbook