I’m sharing scenario-go, a Go library for automated end-to-end testing of conversational agents. You can define scenarios with success and failure criteria, and a testing agent simulates your users until goals are met or issues surface.

We include a connector for OpenAI, but it's trivial to support an LLM via the simple LLMCompletion interface.

Just to note: I work at LangWatch, but this library does not use our product, it just came out of one of our hack days, and the code is MIT licensed.