PRODUCTION SCENARIO
A team tuned a prompt against the single example used to write it and now wants to know whether the app is ready to ship.
Which Foundry step comes next?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Run an evaluation over a representative test dataset
Pre-production evaluation validates performance through evaluation datasets, identifies edge cases, assesses robustness, and measures groundedness, relevance, and safety before anyone is exposed to the app. Continuous evaluation samples production traffic, which requires shipping first. Catalog benchmarks belong to base model selection, a decision already made. Tracing explains how a run behaved, not whether the app generalizes.