How to Trace RAG Failures with Braintrust Logging and Playground
Evals tell you something is wrong. Traces tell you where. Log retrieval, reranking and generation as spans, then replay a bad trace in the playground.
Evals tell you something is wrong. Traces tell you where. Log retrieval, reranking and generation as spans, then replay a bad trace in the playground.
Braintrust turns prompt tweaks into measurable experiments. Set up a dataset, an eval, and a scored run, then compare two prompts side by side.