New

  • Annotation queues — review and annotate spans with a dedicated queue workflow. Assign evaluators per queue item.
  • Evaluator versioning — evaluators now support versioning with backward-compatible migration.
  • Web search support — models with web search capabilities are now supported through the gateway.
  • Low credit webhook — get notified via webhook when your credit balance drops below threshold.

Improved

  • Histogram endpoint now supports boolean evaluators
  • Dataset log enrichment with evaluator metadata