Experiments without Labels for RAG

All you need to run experiments on your RAG pipeline is a CSV file with a set of questions. You don't need labeled answers, and you can easily obtain the questions from search history. The experiment will show you your RAG pipeline's query latency, groundedness ratio, and no answer ratio.

To learn more about the metrics, see Metrics for RAG pipelines.
For instructions on how to prepare datasets, see Evaluation datasets in deepset Cloud.
If you need more information on Experiments, see About Experiments and Evaluating Your Pipeline.

Chat in Playground

Chat is now accessible in Playground. You no longer need to use a special URL to test your chat pipelines. Just go to Playground and choose your chat pipeline to try it out.