Creating Categorical Prompt Properties in Deepchecks

You’ll see how to define categories, set clear guidelines, and use few-shot examples to steer the evaluation process. The video also covers how to use the UI to review, test, and refine your setup, ensuring that your evaluation logic aligns with your specific application goals.

This functionality is especially useful for evaluation tasks where interactions need to be bucketed into clearly defined categories—such as sentiment analysis, quality grading, or intent detection. By combining structured labeling with guideline-driven consistency, Deepchecks makes it easy to analyze LLM interactions at scale and spot recurring issues or trends.