Langsmith Dataset

LangSmith datasets help you test and improve your AI agents. You can create, upload, and manage evaluation datasets. These datasets store example inputs and expected outputs. Use them to check if your agent behaves correctly. The langsmith CLI makes it easy to list, create, delete, or export datasets. You can also add examples and run experiments. This keeps your AI reliable and trustworthy.

There are four main dataset types. final_response tests the full conversation. single_step checks one part of an agent. trajectory verifies the series of tool calls. rag tests retrieval quality. Pick the type that fits your testing goal.

You need a LangSmith API key to get started. Set it as an environment variable or pass it with the CLI. Always check the project name to find the right traces. The CLI warns you before deleting anything. Stay safe and test often.