* 1. Take the user's query as-is. * 2. Embed it using an embedding model. * 3. Use the query's embedding to search an embedding store (containing small segments of your documents) * for the X most ...