We've recently released v0.2 of our LangSmith SDKs (Python and TypeScript) with a number of quality-of-life improvements to make the evals experience more intuitive and flexible:

What’s New

• Simpler Evaluators: Define custom evaluators as basic functions with cleaner arguments, making them easier to write and understand.

• Unified Evaluation: One method to rule them all—use evaluate() for all evaluation types, including pairwise and existing experiments.

• Local Mode: Run evaluations without uploading results—perfect for quick iterations or testing before committing to LangSmith.

• Better Performance: Python SDK now processes large examples ~30% faster.

• Revamped Docs: Explore clearer, more actionable guides and expanded references to level up your workflow.

Together, these improvements allow you to evaluate and refine your applications with less friction. The cleaner syntax and unified methods save time and local evaluations help you test faster without cluttering your experiment history.

Read about the changes in our blog post: https://blog.langchain.dev/easier-evaluations-with-langsmith-sdk-v0-2/

Check out the docs: https://docs.smith.langchain.com/evaluation