Package Adoption Tracker: npm & PyPI Download Trends

Keyless download data for npm and PyPI packages. No API key, no account. Stars are vanity and easy to game; installs are what people actually do.

• Trends — per package: downloads in the last 7 days, the 7 days before that, the growth percent between them, the 30 day total and the daily average. Direction, not just size.
• Discover — search npm by keyword, take the highest ranked packages and rank them by growth. This answers "what is actually rising in the vector database space", which is a different question from "what has the most stars".
• History — the daily download series, one row per package per day, ready to chart.

Who uses it

• Engineering leads choosing a dependency — you are picking something you will live with for years, and a library whose installs are falling is a maintenance problem waiting to happen.
• VCs and analysts — open source traction is a leading indicator, and download growth is the cleanest public measure of it.
• Developer tool marketers — track your own adoption against competitors, weekly, on a schedule.
• Package maintainers — see your own trend without wiring up analytics.

Pairs with our GitHub Trending Scraper for what is being starred and the Hugging Face AI Models Scraper for model adoption.

Input

Field	Description
mode	trends, discover, or history.
registry	npm (all three modes) or pypi (trends only).
packages	Names to track. Scoped npm names like @langchain/core are handled.
searchQuery	What space to search, in discover mode.
searchLimit	How many top results to pull download data for.
minWeeklyDownloads	Drop abandoned and toy packages.
includeMetadata	Add version, license, repository and last publish date.
maxRows	Row cap per run.

Output

• Trends and discover: packageName, registry, scoped, lastWeek, priorWeek, growthPercent, last30Days, averagePerDay, windowStart, windowEnd, url, plus searchScore and searchRank in discover mode and latestVersion, license, repository, lastPublishedAt, versionCount when metadata is on.
• History: packageName, day, downloads.
• PyPI trends: packageName, lastDay, lastWeek, lastMonth, url.

Notes on the data

• Growth is measured against the registry's own returned days, not against today's date. npm publishes with a lag of several days and the lag moves, so the series is sliced into its last 7 and prior 7 buckets from whatever window came back. windowStart and windowEnd tell you exactly what was compared.
• npm bulk lookups reject scoped packages. @scope/name is fetched individually and merged back in, so a mixed list works, it is just slightly slower per scoped name.
• A misspelled npm package is not an error. The registry returns a null value inside an otherwise successful response, so unknown names are logged and skipped rather than silently counted as zero.
• PyPI is trends only. It publishes no keyless search, and pypistats serves totals rather than a daily series, so PyPI rows carry no growthPercent and cannot be charted. It also rate limits hard, one package per request with spacing.
• Download counts include continuous integration and mirror traffic. They measure activity, not distinct humans, and that is true of every download figure anyone quotes.

Pricing

Pay per event: $0.003 per row. The first 2 rows of every run are free.

Data sources: npm registry and downloads API, pypistats.org.