oberservability
Framework-agnostic eval toolkit for AI agents — capture traces, judge outputs, and evolve prompts.
Reliability-first observability for LLM workflows — trace steps, model calls, retries, and costs
