Inspect Parquet storage metadata locally

iParq is a free, open-source command-line tool and Agent Skill for understanding how Parquet files were written. It reports metadata without querying row data and without uploading files to a remote service.

What iParq reveals

Inspect compression codecs, encodings, physical and logical types, row groups, min/max and null statistics, dictionary pages, column and offset indexes, Bloom-filter metadata, compressed sizes, and compression ratios. Compare several files or restrict an inspection to one column when diagnosing storage layout and missing optimizations.

Run it without installing

uvx --refresh iparq inspect FILE.parquet --format json --details --sizes

The JSON output is designed for scripts and agent workflows. A single file produces one object; multiple files produce an array with a file field. Diagnostics go to stderr, and unreadable inputs return a non-zero exit status without corrupting successful JSON output.

When agents should use it

Use iParq when a task asks how a local Parquet file is encoded or compressed, whether it contains page indexes or Bloom filters, how row groups and statistics are arranged, or how storage metadata differs across files. Observed metadata is evidence about the file structure; it is not proof that a particular query engine will use an optimization.

Agent discovery

Agents can discover iParq through the ARD catalog, read the portable Parquet inspection Skill, or start with the machine-readable documentation index. Discovery does not create a network service: inspections still run locally through the published Python package.

Documentation and source

Read the CLI and agent documentation, install the package from PyPI, or review the implementation, tests, and MIT license in the GitHub repository. iParq is free to use; see the concise pricing and license summary.