GitHub REST API (Repositories & Search)

The public side of GitHub is the largest collection of open-source application software in the world, and every bit of it is described by a machine-readable record: one flat JSON object per repository, more than 200 million of them. Each object bundles identity (numeric id, owner login, full name, maintainer-written description), popularity (stargazer, watcher and fork counts, open issue volume), classification (primary language as detected by GitHub's Linguist engine, lowercase maintainer-applied topics, a detected license, a declared homepage) and lifecycle (creation, update and last-push timestamps, an archived flag, visibility level).

Datadory turns that catalog into a delivered feed - filtered, shaped and shipped the way your pipeline wants it. If you size markets by counting repositories, score competitors by their stars-versus-issues posture, or screen dependencies before adoption, this is the base layer under all of it. Get a sample of this dataset and the dictionary below stops being hypothetical.

One captured record beats a paragraph of specification. Here is a real one - the top result when the catalog is queried for the application-software topic:

full_name   : rjlovespy/Mindnote
id          : 670418243
node_id     : R_kgDOJ_XFQw
description : An application software that is an advanced version of notepad
fork        : false

Yes - asked for application software, the corpus opens with an advanced version of notepad. That is the honest texture of the thing: hundreds of millions of repositories, most of them small, a few of them destined to run half the internet, all wearing the same twenty-field shape. The volume is exactly why the counters matter. Stars, forks and issue counts are how signal climbs out of it.

Twenty documented fields, definitions verified against live responses during cataloging. The dictionary below is the complete core schema: every repository record carries all twenty, and none of them require a secondary lookup to interpret.

Geography — global. The corpus is every public repository hosted on GitHub, in every country and every human language. Geography arrives implicitly, through maintainer names and descriptions, rather than as a dedicated column.

Temporal — every record self-describes its own history with three timestamps: created_at (project birth), updated_at (last metadata change) and pushed_at (most recent code push). Cohort any slice by creation month or filter by last-push recency and the timeline reconstructs itself, no external calendar table required.

Granularity — one flat record per repository, delivered in batches of up to 100 records. Nothing nested to unwind; roll up to owner or organization level yourself if the analysis calls for it.

API, files, or your warehouse. Daily, weekly, or hourly.

• Technology-landscape sizing — count repositories by primary language and topic to measure where a category is thick or thin before committing a roadmap to it.
• Competitor open-source tracking — stars, forks and open-issue counts turn rivals' public projects into a readable activity curve.
• Dependency due diligence — the archived flag, fork status, detected license and last-push timestamp screen a candidate library's health in one glance.
• Developer-tool market maps — maintainer-applied topics cluster the corpus into de facto categories and reveal the vocabulary a niche already uses for itself.
• Model training corpora — a stable twenty-field schema over hundreds of millions of behavior-labeled objects is ready-made input for ranking and recommendation systems.

Competitive intelligence and product teams read rivals' public repositories as a leading indicator - issue velocity and fork counts tend to move before launch announcements do. Data scientists and ML engineers get a clean, schema-stable corpus large enough to train on without augmentation. Developers and data-product builders wrap the flat record shape into dependency checkers and ecosystem dashboards without schema gymnastics. Investors and quant researchers use language- and topic-level repository growth as a bottom-up gauge of developer mindshare.

Three things worth knowing upfront. First, the headline corpus size - more than 200 million repositories - is a widely cited order of magnitude rather than a frozen inventory figure; the live total moves constantly and is reported precisely with every query result. Second, eleven of the twenty fields (the timestamps, counters and flags) ship without a single representative example value because they vary by orders of magnitude across repositories; request a sample and they populate immediately. Third, popularity metrics measure attention, not quality - a modestly starred library with recent pushes routinely beats a famous one whose archived flag flipped years ago.

Application Software catalog neighbors that answer different halves of the software-market question. Libraries.io Open Source Packages extends the open-source view into package-manager dependency graphs. G2 Software Categories & Product Pages and Capterra Software Directory cover the commercial, review-driven side of the market. Product Hunt - New Software Launches catches products on day one, before any counters accumulate.