Agent observability field: LangSmith vs Braintrust vs Langfuse vs Arize

LangSmith = deepest LangChain/LangGraph integration. Braintrust = eval-first, trace-to-eval, a GitHub Action that blocks merge below score thresholds. Langfuse = best self-hosted/open-source option.

What it means Informs deterministic-vs-LLM-judge layering. Braintrust's merge-blocking eval-action is a pattern to study — but LLM-judge stays post-hoc, never replacing deterministic CI gates.