AI-driven translation

The AI Tutor asks a language model for a translation plan covering every relationship. Each suggestion comes with a rationale, the risks and a confidence level.

Run it

1. Click AI Tutor in the translation stage to open it. NoER analyzes the model and displays the questions.
2. Choose a model and the global assumptions. Pick the language model, the main priority (read performance, write performance, consistency or storage), the expected data scale and the importance of consistency.
3. Answer for each relationship. Move through the relationships with Next and Back and answer the questions on each one: how often both sides are read together, the growth of related objects, independent updates and the most frequent navigation direction. Questions can be left unanswered.
4. Click Get recommendations to receive the suggested pattern for each relationship.
5. Apply or refine. Apply the patterns one at a time with Apply on each card, or all at once with Apply all. To change a suggestion, use Review answer, adjust it and run the tutor again.

For relationships involving a time entity, the questions address the interval between event arrivals, the maximum number of events per bucket, the append-only nature of events, and the need for summary statistics.

Available models

What the model receives

NoER builds the request automatically. It contains a summary of the model (entities, attributes, relationships and cardinalities), the valid patterns for each relationship, the global assumptions and the answers, and the required answer format. The model must choose among the valid alternatives: suggestions that don’t match one are discarded and never applied.