End-to-end LLM evaluation and observability platform for AI agents — trace, monitor, and continuously improve production agent pipelines with auto-evals and human review.