Ensemble definitions are saved "recipes" that run multiple decision app configurations simultaneously and return the best result based on predefined rules within the Nextmv platform. This approach solves the problem of manual retuning by allowing a platform to try various settings, such as different solver strategies or trade-offs, in parallel.
Ensembles consist of run groups, which are specific configurations with overridden options, and lexicographic selection rules that determine the winner from the generated candidates. These rules are applied sequentially, with ties broken by subsequent rules. Users can set these up via the console or programmatically for parameter sweeps, de-risking deployments, or managing multi-objective trade-offs.
