SteveSandersonMS
commented
5 months ago - During data generation, produces a corresponding set of question/answer pairs about the products based on pre-ingestion source data
- Ingests the q/a pairs into seed data
- Adds a console app that loops over these q/a pairs and calls the backend assistant API to ask the question, then uses another LLM prompt to score the quality of the answers (in batches) vs the ground truth.