Setup & Installation
What This Skill Does
Read-only API client for the Hugging Face Dataset Viewer. Fetches dataset splits, paginates rows, runs text search and filter queries, retrieves parquet shard URLs, and reads size and column statistics. Also supports SQL querying over parquet files and uploading datasets to the Hub.
Instead of downloading full dataset files to inspect or query them, you get direct row-level access and SQL queries against remote parquet shards without any local data movement.
When to use it
- Paginating through 100k+ rows of a public dataset without downloading the full file
- Searching a text classification dataset for rows matching a specific string
- Running SQL queries on Hub parquet shards to get aggregated statistics
- Listing all configs and splits of a dataset before writing a training script
- Uploading a local parquet folder to a new private Hub dataset repo