Run evaluations and generate synthetic datasets: Review

My first `evaluators.summarize` call threw a key error. Turns out my local environment was missing a dependency that the sandbox had pre-installed—a good reminder that these one-hour modules are tightly scripted. One step off the path and you’re on your own.

This is mostly a code walkthrough for Azure AI's evaluation SDK. It shows how to run built-in checks for things like groundedness and relevance, and then track them in AI Studio. The part on creating a custom evaluator with the `.prompty` format was new to me and genuinely useful. But the 'generate synthetic datasets' section felt tacked on at the end, just enough to make you aware the feature exists.

I got the code snippets I needed. It’s a dense hour if you're already in the Azure ecosystem, but don't expect it to teach you the theory behind the metrics themselves.