Datasets
These datasets let you study PyPI without downloading and parsing every package yourself.
py-code.org import index
What the latest release of each indexed project imports, and which packages reference every module, function and class. Updated monthly, free to use, as static JSON.
PyPI metadata in BigQuery
PyPI publishes download events and the metadata of every release in Google BigQuery, in the public bigquery-public-data.pypi dataset. Queries need a Google Cloud account.
ClickHouse PyPI demo
ClickHouse hosts PyPI download counts that you can query with SQL in the browser, without an account. Results are capped, so page through large queries.
PyPI's own APIs
The JSON API gives one project's releases and files. The Simple API lists every project name. The changelog lists every upload, yank and removal in order.
The PyPI Data datasets (2023–2026)
The original project on this domain published Parquet files with metadata about every file uploaded to PyPI: project, version, path, size, line count and SHA256 hash. Those releases were removed along with the project's GitHub organisation in 2026.
The project's author shared a snapshot of the data from June 2026, about 100 GB of Parquet, in the discuss.python.org thread “You can now download PyPI locally”. The thread has the download link.
DuckDB reads Parquet directly, so you can query the snapshot without loading it into a database:
$ duckdb -c "select project_name, path, lines from '*.parquet' order by lines desc limit 5"