LangSmith adds tracing for LangGraph agents and automated RAG evaluation

On October 3, 2026, a post published on Juejin introduced LangSmith, presenting it as a path from tracing to automated RAG evaluation. The article opens with a problem that shows up soon after building a LangGraph Agent or a RAG application: the program runs, but what happens inside it is unclear. Which nodes did a single request pass through? What did the Retriever return?

What prompt did the model actually receive? LangSmith is aimed at exactly this class of questions. According to the post, it provides tracing for LangGraph Agent and RAG applications, making those steps in the chain visible. Beyond tracing, the article goes on to explain how to use LangSmith for automated RAG evaluation, extending the tool's use into the evaluation stage. As of this report, the verifiable material stops at that single introductory post: it does not mention a version number, a release timeline, or any follow-up updates, and no other parties are quoted. The story therefore currently rests at an introduction to LangSmith's tracing and automated RAG evaluation capabilities.