Tuxery
Docs· Open data

Open data

Everything Tuxery shows comes from public sources, and everything it gathers is public too. Here is what you can get today, and what is planned.

Download, now

Per-source data

Each source's cleaned listing is committed to the catalog repository, one file per source: src/sources/cache/ on GitHub. Download a file from there, or clone the repository to get all of them (about 200 MB).

  • Format: NDJSON — one JSON object per line, one line per package as the source lists it (before any merging). Each source keeps its own fields, described in src/sources/<source>/types.ts; there is no shared schema across sources, because upstream shapes differ.
  • Metadata: next to each <source>.ndjson, a <source>.meta.json says when it was fetched (fetchedAt), from which URL, and how many entries it holds.
  • Biggest files: the AUR (about 37 MB), Ubuntu (29 MB) and Debian (27 MB). Everything else is smaller.
  • Update frequency: every source is refetched weekly (Saturday, 04:00 UTC); the changes land through a pull request, so a file's fetchedAt is the date to trust.
# One source, straight from GitHub
curl -LO https://raw.githubusercontent.com/tuxery/catalog/main/src/sources/cache/flatpak-flathub.ndjson

Classifications

The apps-or-games verdicts and categories Tuxery assigned with the help of language models (about 164,000 entries) are committed too: config/llm-classifications.json. Each entry carries the model that produced it and a confidence level — see the data philosophy for how they are used and checked.

The merged catalog

The merged dataset — one record per app, with its packages, editions, add-ons and related apps, as the site shows them — is not downloadable yet. It is rebuilt from the files above by the catalog's pipeline, and where to publish it (GitHub Releases is the likely candidate) is still being decided. Until then, you can rebuild it yourself from the catalog repository.

API, later

There is no public API today. One is planned, but serving queries to everyone costs more than the free tiers Tuxery runs on, so it will come once that is funded — see Status. Tell us what you would use it for on the catalog's issues: it shapes what gets built first.

Licenses

  • Tuxery's code (connectors, pipeline, this site) is AGPL-3.0-or-later.
  • The data itself comes from each source, and each source's own terms still apply to it. AppStream metadata (Flathub, elementary AppCenter, the distributions' AppStream files) is published under permissive metadata licenses, usually CC0-1.0; distribution package indexes and store APIs each have their own terms. Check the source's terms before redistributing its data. Tuxery has not yet chosen a license for what it adds on top (merges, classifications); until it does, ask before reusing those in bulk.