Search Toolkit helps teams determine whether an AI system's poor answers start with poor document retrieval. Mistral AI's public-preview announcement describes a shared framework for preparing documents, searching them, and evaluating the returned results. Teams can compare keyword search, meaning-based search, or combinations against their own test questions and relevance judgments. Because retrieval is measured separately from answer generation, developers can check whether the right evidence is reaching the model before changing its prompts.
Terms in this piece · Glossary
RAG — Retrieval-augmented generation — fetching relevant documents first and pasting them into the model's context so it answers from your data instead of memory.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
