LangChain invites teams to try Trajectories in LangSmith

LangChain says teams can try Trajectories in LangSmith. It describes the feature as a way to debug long-running agents faster without reconstructing a session, and to give subject-matter experts a readable view for reviewing agent behavior.

LangChain also says Trajectories can score agent behavior with online evals and turn production sessions into fine-tuning datasets. Further details are in its LangSmith Trajectories tracing post.