What problem does it solve? Managing evaluation datasets for LLM experiments requires repetitive CLI work: creating datasets from files, appending examples, exporting versions, and verifying counts. This Skill provides complete operational guidance for the Arize ax CLI so dataset tasks are executed correctly the first time. ## Core Features & Use Cases - Dataset CRUD: List, get, create, export, and delete datasets in an Arize space, with support for CSV, JSON, JSONL, and Parquet files or stdin piping. - Example Management: Append examples inline via JSON or from files, with schema validation guidance to avoid silent field mismatches. - Versioned Exports: Export specific dataset versions, bulk-export datasets over 500 examples with --all, and verify row counts against server metadata. - Use Case: You need a golden test set for prompt evaluation. Create a dataset from a JSON file, append new examples as you collect them, then export a specific version to feed into an experiment run. ## Quick Start Ask the AI to create an Arize dataset named eval-set-v1 in your space from your eval_data.csv file using the ax CLI.
