Add the full conversation transcript. The system will extract questions, classify them, prompt them, search for context, and show question-wise answers.
Enter one question. This uses the existing process_question flow and returns classification, prompted queries, retrieved context, chunks, and the final answer.
Download a CSV for evals with bucket, question, answer, profile, and chunks_json. Only rows with non-empty answers are exported.
Select one or more buckets. Matching uses ANY of the selected buckets.