README.md

metadata

language:
  - en
license: apache-2.0
size_categories:
  - 100K<n<1M
task_categories:
  - tabular-classification
  - text-classification
tags:
  - pypi
  - python
  - software-engineering
  - metadata
pretty_name: PyPI Download and Package Analysis
configs:
  - config_name: default
    data_files:
      - split: train
        path: packages_data/*.parquet
dataset_info:
  features:
    - name: metadata_version
      dtype: string
    - name: name
      dtype: string
    - name: version
      dtype: string
    - name: summary
      dtype: string
    - name: description
      dtype: string
    - name: description_content_type
      dtype: string
    - name: author
      dtype: string
    - name: author_email
      dtype: string
    - name: maintainer
      dtype: string
    - name: maintainer_email
      dtype: string
    - name: license
      dtype: string
    - name: keywords
      dtype: string
    - name: classifiers
      sequence: string
    - name: platform
      sequence: string
    - name: home_page
      dtype: string
    - name: download_url
      dtype: string
    - name: requires_python
      dtype: string
    - name: requires
      sequence: string
    - name: provides
      sequence: string
    - name: obsoletes
      sequence: string
    - name: requires_dist
      sequence: string
    - name: provides_dist
      sequence: string
    - name: obsoletes_dist
      sequence: string
    - name: requires_external
      sequence: string
    - name: project_urls
      sequence: string
    - name: uploaded_via
      dtype: string
    - name: upload_time
      dtype: timestamp[us]
    - name: filename
      dtype: string
    - name: size
      dtype: int64
    - name: path
      dtype: string
    - name: python_version
      dtype: string
    - name: packagetype
      dtype: string
    - name: comment_text
      dtype: string
    - name: has_signature
      dtype: bool
    - name: md5_digest
      dtype: string
    - name: sha256_digest
      dtype: string
    - name: blake2_256_digest
      dtype: string
    - name: license_expression
      dtype: string
    - name: license_files
      sequence: string
    - name: recent_7d_downloads
      dtype: int64

PyPI Download and Package Analysis

A comprehensive snapshot of the Python Package Index (PyPI), covering 690,775 packages published from April 2005 through February 2026. Each row represents a published package release, enriched with full metadata from the PyPI API and recent download statistics from the PyPI BigQuery public dataset.

Dataset at a Glance

Stat	Value
Total packages	690,775
Date range	April 2005 – February 2026
Total 7-day downloads	~13.3 Billion
Format	Parquet (15 shards)
License	Apache-2.0

Data Fields

Field	Type	Description
name	string	Package name on PyPI (unique identifier)
version	string	Release version string (PEP 440)
summary	string	One-line description of the package
description	string	Full project description / README text
description_content_type	string	MIME type of the description (e.g. text/markdown)
author	string	Primary author name
author_email	string	Primary author email
maintainer	string	Maintainer name (if different from author)
maintainer_email	string	Maintainer email
license	string	License string as declared by the author
keywords	string	Space- or comma-separated keywords
classifiers	list[string]	PyPI trove classifiers (e.g. Programming Language :: Python :: 3)
platform	list[string]	Target platforms declared by the author
home_page	string	Project homepage URL
download_url	string	Direct download URL (if provided)
requires_python	string	Python version constraint (e.g. >=3.8)
requires	list[string]	Runtime dependencies
project_urls	list[string]	Additional URLs (source, docs, tracker, etc.)
upload_time	timestamp	UTC timestamp of when this release was uploaded
size	int64	Size of the distribution file in bytes
packagetype	string	Distribution type: sdist, bdist_wheel, etc.
metadata_version	string	Metadata specification version
recent_7d_downloads	int64	Total downloads in the most recent 7-day window

Usage

from datasets import load_dataset

ds = load_dataset("semvec/pypi-packages")
df = ds["train"].to_pandas()

# Top 10 most downloaded packages
df.sort_values("recent_7d_downloads", ascending=False).head(10)[["name", "summary", "recent_7d_downloads"]]

Example Use Cases

• Trend Analysis — Track adoption of ecosystems (AI/ML, web frameworks, DevOps tooling) by filtering classifiers and plotting upload_time vs. cumulative package count.
• Package Classification / NLP — Use summary and description to train text classifiers or summarization models that categorize packages by domain.
• Dependency Graph Research — Parse requires to construct a directed dependency graph of the entire Python ecosystem.
• Popularity Modeling — Predict recent_7d_downloads from metadata features like requires_python, classifiers, description length, and age.
• License Compliance — Audit license diversity across the ecosystem and identify packages with missing or ambiguous license declarations.
• Author & Maintainer Analysis — Study open-source contribution patterns, prolific authors, and package maintainer turnover over time.

Data Collection

Metadata was fetched from the PyPI JSON API for every package listed in the PyPI simple index. Download counts were sourced from the PyPI public BigQuery dataset (bigquery-public-data.pypi.file_downloads), aggregated over the 7 days preceding the collection date (February 2026).

Citation

If you use this dataset in your research, please cite it as:

@dataset{pypi_packages_2026,
  title   = {PyPI Download and Package Analysis},
  author  = {semvec},
  year    = {2026},
  url     = {https://huggingface.co/datasets/semvec/pypi-packages},
  license = {Apache-2.0}
}