Webinar 01: From development to production.

AI agents act weird in production. In this session, we covered how to:

• Score your app before it ships: upload a dataset of goldens, create a metric collection, and score your test cases against it, so every run gets the same yardstick.
• Catch regressions and ship the winner: run a regression test on a new version, compare the two runs side by side, and deploy the one that scores better.
• Keep scoring in production: set up tracing on the deployed app, create an evaluation rule, reuse the same metric collection, and score live traces as they come in.
• Close the loop: send the traces that failed back into your dataset as new goldens, so the next test run covers what production just taught you. Then run the cycle again.