A catalog of 168,605,740 public files (144 TB; size not published for 23,970,466 files) from 26 sources: web documents, Python packages, research data, chess games, forensic scenarios (disk, memory and network images), Q&A posts, single files inside zip archives, mixed file types, mixed documents, presentations, issue tickets, source packages, U.S. government documents, chess engine evaluations, spreadsheets, R packages, emails, forum posts, system logs, security vulnerability records (CVE), chess puzzles, chat logs, JavaScript packages and meeting transcripts. Every row says where one file can be downloaded, how large it is, which hashes verify it and under which license it is offered. This repository holds only the catalog (about 29.2 GB); the files stay with their publishers.

Sources

Source	Content	Files	Size	Training use	Redistribution	Reachable	Browse
PyPI packages	Python packages	21,693,722	47.6 TB	mixed	mixed	not checked	rows
End of Term Web Archive (in progress)	web documents	88,300,000	39.0 TB	yes	yes	180 / 180 (sample)	rows
Harvard Dataverse (in progress)	research data	354,800	35.7 TB	mixed	mixed	not checked	rows
Common Crawl office files (in progress)	web documents	18,000,000	13.9 TB	unknown	unknown	not checked	rows
Lichess open database	chess games, chess puzzles, chess engine evaluations	1,369	2.63 TB	yes	mixed	1,369 / 1,369	rows
NapierOne and Digital Corpora scenarios	forensic scenarios (disk, memory and network images), mixed file types	1,056	2.49 TB	mixed	mixed	304 / 307 (sample)	rows
Stack Exchange data dump (archive.org)	Q&A posts	3,629	960 GB	conditional	conditional	3,629 / 3,629	rows
GovDocs1 files (in progress)	single files inside zip archives	976,399	497 GB	yes	yes	not checked	rows
NapierOne files (in progress)	single files inside zip archives	400,566	416 GB	mixed	mixed	not checked	rows
GovDocs1 (Digital Corpora)	mixed documents	1,000	333 GB	yes	yes	1,000 / 1,000	rows
Zenodo10K (PPTAgent)	presentations	10,448	242 GB	mixed	mixed	not checked	rows
Debian bug reports	issue tickets	2,244,932	145 GB	unknown	unknown	58 / 60 (sample)	rows
Debian source packages	source packages	115,614	99.2 GB	yes	conditional	not checked	rows
govinfo bulk data	U.S. government documents	419,274	56.4 GB	yes	yes	31 / 351 (sample)	rows
CRAN packages	R packages	25,299	18.9 GB	yes	conditional	25,284 / 25,284 (sample)	rows
Enron Spreadsheets and Emails (figshare article 1221767)	spreadsheets, emails	14	9.85 GB	yes	yes	14 / 14	rows
Fuse (Zenodo record 581678)	spreadsheets	1	9.39 GB	yes	yes	1 / 1	rows
Discourse forum posts (in progress)	forum posts	8,998,807	6.84 GB	mixed	mixed	not checked (API source)	rows
Loghub: A Large Collection of System Log Datasets for AI-driven Log Analytics (Zenodo record 8196385)	system logs	19	6.00 GB	conditional	conditional	19 / 19	rows
NVD CVE vulnerability records	security vulnerability records (CVE)	381,690	1.95 GB	yes	yes	not checked (API source)	rows
Apache Jira issues	issue tickets	1,179,062	1.28 GB	unknown	unknown	not checked (API source)	rows
Zulip public chat archives	chat logs	1,300,638	239 MB	unknown	no	not checked (API source)	rows
Ubuntu IRC logs	chat logs	456,106	51.6 MB (size not published for 229,171 files)	conditional	no	8,834 / 8,846 (sample)	rows
npm packages (in progress)	JavaScript packages	22,370,107	not published	mixed	mixed	not checked	rows
UK Parliament Hansard debates	meeting transcripts	1,370,783	not published	yes	yes	not checked (API source)	rows
Linux kernel mailing list archives	emails	405	not published	no	no	0 / 405	rows
All sources		168,605,740	144 TB (size not published for 23,970,466 files)			40,723 / 41,465	rows

A file is one download, as the source publishes it; an archive counts once.

Training use and Redistribution summarize each license as yes, conditional, no or unknown (mixed when the files of a source differ); docs/licenses.md defines each tier and where the license of a file is written. This is a summary, not legal advice: check the upstream terms before use.

Quick start

Browse, search and filter every row in the viewer at the top of this page, or load the catalog:

from datasets import load_dataset

catalog = load_dataset("Battam/terminal-data-pointers", split="train")  # all sources
pypi = load_dataset("Battam/terminal-data-pointers", "pypi", split="train")  # one source

usable = catalog.filter(lambda row: row["training_ok"] == "yes")
print(usable.num_rows, "files,", round(sum(b or 0 for b in usable["size_bytes"]) / 1e12, 2), "TB")

Then download each uri and check it against size_bytes and the hashes (see docs/retrieval.md).

What is inside

By content

Table: files and size by content kind

Content	Category	Files	Size	Sources
web documents	`web-documents`	106,300,000	53.0 TB	2
Python packages	`python-packages`	21,693,722	47.6 TB	1
research data	`research-data`	354,800	35.7 TB	1
chess games	`chess-games`	1,367	2.61 TB	1
forensic scenarios (disk, memory and network images)	`forensic-scenarios`	922	1.65 TB	1
Q&A posts	`qa-posts`	3,629	960 GB	1
single files inside zip archives	`archive-members`	1,376,965	913 GB	2
mixed file types	`mixed-files`	134	838 GB	1
mixed documents	`mixed-documents`	1,000	333 GB	1
presentations	`presentations`	10,448	242 GB	1
issue tickets	`issue-tickets`	3,423,994	146 GB	2
source packages	`source-packages`	115,614	99.2 GB	1
U.S. government documents	`government-documents`	419,274	56.4 GB	1
chess engine evaluations	`chess-evaluations`	1	22.1 GB	1
spreadsheets	`spreadsheets`	15	19.2 GB	2
R packages	`r-packages`	25,299	18.9 GB	1
emails	`emails`	419	9.85 GB (size not published for 405 files)	2
forum posts	`forum-posts`	8,998,807	6.84 GB	1
system logs	`system-logs`	19	6.00 GB	1
security vulnerability records (CVE)	`vulnerability-records`	381,690	1.95 GB	1
chess puzzles	`chess-puzzles`	1	304 MB	1
chat logs	`chat-logs`	1,756,744	290 MB (size not published for 229,171 files)	2
JavaScript packages	`js-packages`	22,370,107	not published	1
meeting transcripts	`meeting-transcripts`	1,370,783	not published	1

A file of several kinds is counted under each of them. The content column of the catalog holds the category.

Table: files and size by file extension

Extensions are those of the downloads as the sources publish them; an archive (zip, 7z, gz, zst and so on) counts under its own extension, not under the files inside it.

Extension	Files	Size	Sources
`xml`	42,410,604	1.13 TB	6
`pdf`	37,917,420	43.3 TB	7
`txt`	25,720,307	7.44 TB (size not published for 229,171 files)	10
`tgz`	22,370,427	422 GB (size not published for 22,370,107 files)	4
`whl`	12,892,237	37.9 TB	1
`md`	8,999,839	6.97 GB	3
`gz`	8,650,080	13.4 TB	8
`json`	1,760,789	30.2 GB (size not published for 1,370,783 files)	5
`log`	1,133,023	149 GB	4
`summary`	1,122,443	389 MB	1
`csv`	911,586	692 GB	6
`ics`	886,909	2.54 GB	3
`doc`	662,895	134 GB	6
`docx`	556,276	204 GB	6
`xls`	481,466	67.6 GB	5
other (10,852 more extensions)	2,128,320	39.3 TB (size not published for 405 files)	17

By license

Table: files and size by license tier

Training use	Files	Size	Sources
`yes`	128,143,700	101 TB (size not published for 19,397,225 files)	17
`conditional`	4,027,279	17.4 TB (size not published for 229,171 files)	8
`no`	108,609	81.4 MB (size not published for 405 files)	2
`unknown`	36,326,152	25.3 TB (size not published for 4,343,665 files)	11

Redistribution	Files	Size	Sources
`yes`	123,729,434	91.5 TB (size not published for 18,386,627 files)	14
`conditional`	7,915,324	25.5 TB (size not published for 1,010,598 files)	12
`no`	1,827,634	1.86 TB (size not published for 229,576 files)	5
`unknown`	35,133,348	25.3 TB (size not published for 4,343,665 files)	9

A source whose files carry different licenses is counted in every tier it has files in.

Source details

PyPI packages: 21,693,722 files, 47.6 TB, Per project (license expression, classifiers or license text)

Release files (sdists, wheels and others) of all projects on PyPI (pypi.org). The file list comes from the Index API (PEP 691 JSON); license information comes from the JSON API info of each project's latest release. One pointer per release file. Packages can contain malicious code; treat files as data and run nothing outside a sandbox.

Upstream title: PyPI

Batch	Files	Size	File types
`all`	21,693,722	47.6 TB	`whl`, `gz`, `egg` and 8 more

License: Per project (license expression, classifiers or license text). Each project declares its own license. Per file: license_override.notes gives the license expression, the License classifiers and the first 200 characters of the license field of the project's latest release; license_override.spdx gives the SPDX expression where one is known. Check the upstream terms before use.

End of Term Web Archive (in progress): 88,300,000 files, 39.0 TB, No restrictions (Creative Commons Zero)

Collection in progress: 88,300,000 files so far.

Office and document files (PDF, Word, Excel, PowerPoint, CSV, plain text and similar) captured by the End of Term Web Archive, which crawls U.S. federal government websites at each change of administration (terms 2004 to 2024). Each pointer is one HTTP response record inside a WARC file, retrieved by byte range; the hash covers the HTTP response body. Payloads are deduplicated across terms by hash.

Batch	Files	Size	File types
`all`	88,300,000	39.0 TB	`xml`, `pdf`, `txt` and 19 more

Link check: 180 of 180 files reachable with matching size. The check covered a random sample of 180 of the 88,300,000 files.

License: No restrictions (Creative Commons Zero). The dataset page states: "There are no restrictions on the use, access, and/or download of data from the End of Term Web Archive Dataset." It is marked Creative Commons Zero. Third-party content inside archived pages keeps its own terms. The project asks users to cite the End of Term Web Archive. Check the upstream terms before use.

Harvard Dataverse (in progress): 354,800 files, 35.7 TB, Per dataset (CC0 1.0, CC BY 4.0 and others)

Collection in progress: 354,800 files so far.

Public files published on Harvard Dataverse (dataverse.harvard.edu), enumerated through the Search API by license facet. One pointer per file; for ingested tabular files the pointer is to the original upload (?format=original). Files in datasets that require a guestbook form are listed with access.needs_auth set to guestbook.

Batch	Files	Size	File types
`all`	354,800	35.7 TB	`jpg`, `pdf`, `stl` and 2480 more

License: Per dataset (CC0 1.0, CC BY 4.0 and others). Each file carries the license of its dataset. Per file: license_override.notes gives the Dataverse license name; license_override.spdx gives the SPDX id where one exists. Check the upstream terms before use.

Common Crawl office files (in progress): 18,000,000 files, 13.9 TB, Common Crawl Terms of Use

Collection in progress: 18,000,000 files so far.

Office and document files (including ICS, EML and MBOX) captured by Common Crawl in crawl CC-MAIN-2026-34. Each pointer is one HTTP response record inside a WARC file, retrieved by byte range; the hash covers the HTTP response body. Payloads are deduplicated within the crawl by hash. Common Crawl truncates each response at 1 MiB; truncated records are marked in source_meta.content_truncated.

Upstream title: Common Crawl

Batch	Files	Size	File types
`2026-34`	18,000,000	13.9 TB	`pdf`, `txt`, `xml` and 19 more

License: Common Crawl Terms of Use. Each page remains under the copyright of its owner. The Terms of Use say: "YOU AGREE TO RESPECT THE COPYRIGHTS ... OF THIRD PARTIES". Check the upstream terms before use.

Lichess open database: 1,369 files, 2.63 TB, CC0 1.0 (broadcast games: CC BY-SA 4.0)

Monthly PGN files of standard games and 8 variants, monthly PGN files of broadcast games, the puzzle database (CSV) and the engine evaluation database (JSONL) from database.lichess.org, all zstd-compressed. One pointer per published file.

Batch	Files	Size	File types
`all`	1,369	2.63 TB	`zst`

Link check: 1,369 of 1,369 files reachable with matching size.

License: CC0 1.0 (broadcast games: CC BY-SA 4.0). The site states: "Database exports are released under the Creative Commons CC0 license". Broadcast games are marked CC BY-SA 4.0. Per file: Broadcast rows carry license_override with CC BY-SA 4.0. Check the upstream terms before use.

NapierOne and Digital Corpora scenarios: 1,056 files, 2.49 TB, Edinburgh Napier University License Agreement plus per-subset upstream terms; CC0 1.0 (Digital Corpora terms)

NapierOne mixed file subsets (one pointer per subset zip) and the Digital Corpora forensic scenarios (one pointer per file), both listed from public AWS buckets.

Upstream title: NapierOne (Edinburgh Napier University)

Batch	Files	Size	File types
`napierone`	134	838 GB	`zip`
`scenarios`	922	1.65 TB	`zip`, `e01`, `xml` and 44 more

Link check: 304 of 307 files reachable with matching size. Other results: 3 error. The check covered a random sample of 307 of the 1,056 files.

License: Edinburgh Napier University License Agreement plus per-subset upstream terms. NapierOne is released under an MIT-style license from Edinburgh Napier University. Each subset also carries the terms of the material it was built from (for example the UK Open Government Licence v3.0, Linux distribution licenses or the Common Crawl terms of use), stated in the subset's description PDF. The AWS Open Data registry entry allows commercial and noncommercial use with acknowledgment. Acknowledge NapierOne and cite S. R. Davies, R. Macfarlane and W. J. Buchanan, NapierOne: A modern mixed file data set alternative to Govdocs1, Forensic Science International: Digital Investigation 40 (2022) 301330, doi:10.1016/j.fsidi.2021.301330. Per file: Subsets gathered from websites (names containing FROM-WEB) carry license_override with training_ok and redistribution_ok unknown: copyright stays with each website. CC0 1.0 (Digital Corpora terms). Digital Corpora releases its original content, including the scenarios, under CC0. Material that explicitly claims copyright, such as third party software inside disk images and network captures, is not covered by CC0. Per file: The terms ground CC0 on authorship by US Government employees. Rows from scenario folders made by others (magnet, 2025-cn) carry license_override with training_ok and redistribution_ok unknown. Check the upstream terms before use.

Stack Exchange data dump (archive.org): 3,629 files, 960 GB, CC BY-SA 2.5, 3.0 or 4.0 (by contribution date)

Stack Exchange data dumps on archive.org: one pointer per 7z archive (one archive per site). The main batch is the last official upload (2024-04-02); later community re-uploads are separate batches.

Batch	Files	Size	File types
`2024-04-02`	371	99.1 GB	`7z`
`2024-06-30-revised`	362	83.5 GB	`7z`
`2024-09-30`	362	96.2 GB	`7z`
`2024-12-31`	362	94.5 GB	`7z`
`2025-03-31`	362	97.1 GB	`7z`
`2025-06-30`	362	97.3 GB	`7z`
`2025-09-30`	362	97.4 GB	`7z`
`2025-12-31`	362	98.3 GB	`7z`
`2026-03-31`	362	98.4 GB	`7z`
`2026-06-30`	362	98.5 GB	`7z`

Link check: 3,629 of 3,629 files reachable with matching size.

License: CC BY-SA 2.5, 3.0 or 4.0 (by contribution date). Content is licensed CC BY-SA; the version (2.5, 3.0 or 4.0) depends on when it was contributed. The ContentLicense field of each post in the dump gives the license of its current revision. Attribution must name Stack Exchange as the source, link directly to the original post, show the author's name and link to the author's profile. Check the upstream terms before use.

GovDocs1 files (in progress): 976,399 files, 497 GB, Digital Corpora terms

Collection in progress: 976,399 files so far.

Single files inside the GovDocs1 zip archives (real files collected from U.S. .gov websites, mixed types), one pointer per file, each retrievable with HTTP Range requests on its archive.

Upstream title: GovDocs1 files (Digital Corpora)

Batch	Files	Size	File types
`all`	976,399	497 GB	`pdf`, `html`, `jpg` and 60 more

License: Digital Corpora terms. The site states the files are "freely available for research and may be (to the best of our knowledge) freely redistributed". Check the upstream terms before use.

NapierOne files (in progress): 400,566 files, 416 GB, Edinburgh Napier University License Agreement plus per-subset upstream terms

Collection in progress: 400,566 files so far.

Single files inside the NapierOne subset archives (the total archive of each subset), one pointer per file, each retrievable with HTTP Range requests on its archive.

Upstream title: NapierOne files (Edinburgh Napier University)

Batch	Files	Size	File types
`all`	400,566	416 GB	`jpg`, `7z`, `docx` and 8335 more

GovDocs1 (Digital Corpora): 1,000 files, 333 GB, Digital Corpora terms

About 986,000 real files collected from U.S. .gov websites, distributed as 1,000 zip archives of about 1,000 files each; mixed types (pdf, html, jpg, doc, xls, ppt, txt, xml, gif, ps, csv and more). One pointer per zip archive, the smallest unit the source distributes.

Batch	Files	Size	File types
`all`	1,000	333 GB	`zip`

Link check: 1,000 of 1,000 files reachable with matching size.

License: Digital Corpora terms. The site states the files are "freely available for research and may be (to the best of our knowledge) freely redistributed". Check the upstream terms before use.

Zenodo10K (PPTAgent): 10,448 files, 242 GB, CC BY 4.0 (other licenses per file)

Native .pptx files collected from Zenodo, published as the Hugging Face dataset Forceless/Zenodo10K; the directory name of each file is its license. One pointer per .pptx file.

Batch	Files	Size	File types
`all`	10,448	242 GB	`pptx`

License: CC BY 4.0 (other licenses per file). Each file sits in a directory named after its license; most files are CC BY 4.0. Per file: Rows whose license is not CC BY 4.0 carry license_override with the license from the directory name and its own training_ok and redistribution_ok. Check the upstream terms before use.

Debian bug reports: 2,244,932 files, 145 GB, No content license stated

Raw data of the Debian bug tracking system (bugs.debian.org) from the official rsync mirror bugs-mirror.debian.org (bts-spool-db, bts-spool-archive: open and recently closed bugs, and archived bugs). Each bug has two pointers: the full mail log of the bug (.log) and its metadata (.summary, with package, title, severity and tags).

Batch	Files	Size	File types
`all`	2,244,932	145 GB	`log`, `summary`

Link check: 58 of 60 files reachable with matching size. Other results: 2 unreachable. The check covered a random sample of 60 of the 2,244,932 files.

License: No content license stated. Debian states that all bug tracker correspondence, including names and email addresses, is publicly archived (https://www.debian.org/legal/privacy). Messages remain the copyright of their senders; no content license is stated. Check the upstream terms before use.

Debian source packages: 115,614 files, 99.2 GB, Debian Free Software Guidelines (main)

All source files (.dsc, upstream tarballs and Debian tarballs) of Debian source packages (trixie; main). Lists and hashes come from Sources.xz on the mirror network, checked against Release first.

Batch	Files	Size	File types
`all`	115,614	99.2 GB	`xz`, `dsc`, `gz` and 2 more

License: Debian Free Software Guidelines (main). Everything in main meets the Debian Free Software Guidelines; per-package licenses are in debian/copyright. Per file: Rows from contrib or non-free carry license_override describing that area. Check the upstream terms before use.

govinfo bulk data: 419,274 files, 56.4 GB, U.S. Government work (public domain, 17 U.S.C. 105)

Files from the govinfo Bulk Data Repository of the U.S. Government Publishing Office: bill texts, bill status and summaries, the Federal Register, the Code of Federal Regulations (annual and electronic editions), public laws, the Statutes at Large, presidential documents and more. One pointer per file in the public JSON listings. The source gives sizes but no hashes; compute them at retrieval.

Batch	Files	Size	File types
`all`	419,274	56.4 GB	`xml`, `zip`, `html` and 9 more

Link check: 31 of 351 files reachable with matching size. Other results: 320 error. The check covered a random sample of 351 of the 419,274 files.

License: U.S. Government work (public domain, 17 U.S.C. 105). Works of the U.S. Government are not subject to copyright in the United States (17 U.S.C. 105). The govinfo copyright notice adds that Government publications may contain copyrighted material used with the permission of the copyright owner. Check the upstream terms before use.

CRAN packages: 25,299 files, 18.9 GB, Per package (GPL, MIT, Apache and others)

Source packages of all current R packages on CRAN (src/contrib). The list and md5 come from the PACKAGES index; sizes come from HEAD requests.

Upstream title: CRAN

Batch	Files	Size	File types
`all`	25,299	18.9 GB	`gz`

Link check: 25,284 of 25,284 files reachable with matching size. The check covered a random sample of 25,284 of the 25,299 files.

License: Per package (GPL, MIT, Apache and others). CRAN accepts only packages under open source licenses; each package declares its own. Per file: license_override.notes gives the License field from PACKAGES; license_override.spdx gives the SPDX id where one is recognised. Check the upstream terms before use.

Enron Spreadsheets and Emails (figshare article 1221767): 14 files, 9.85 GB, CC BY 4.0

Files of figshare article 1221767: Enron Spreadsheets and Emails (DOI 10.6084/m9.figshare.1221767.v2). One pointer per file.

Batch	Files	Size	File types
`all`	14	9.85 GB	`7z`

Link check: 14 of 14 files reachable with matching size.

License: CC BY 4.0. License in the figshare article metadata: CC BY 4.0. Check the upstream terms before use.

Fuse (Zenodo record 581678): 1 file, 9.39 GB, cc-by-4.0

Files of Zenodo record 581678: Fuse (DOI 10.5281/zenodo.581678). One pointer per file.

Batch	Files	Size	File types
`all`	1	9.39 GB	`zip`

Link check: 1 of 1 file reachable with matching size.

License: cc-by-4.0. License in the Zenodo record metadata: cc-by-4.0. Check the upstream terms before use.

Discourse forum posts (in progress): 8,998,807 files, 6.84 GB, Per forum: CC BY-NC-SA 3.0, CC BY-SA 4.0, the Swift license, or none stated

Collection in progress: 8,998,807 files so far.

Posts from 36 public Discourse forums of open source and technology communities (such as Python, Rust, Julia, Fedora, Home Assistant and Swift), listed newest first through each forum's /posts.json API. One pointer per post; the collection is the forum host name. The hash covers the post's raw Markdown text as returned by the forum API, as UTF-8 bytes. Topic, reply position and times are in source_meta.

Batch	Files	Size	File types
`all`	8,998,807	6.84 GB	`md`

Link check: not checked (API source).

License: Per forum: CC BY-NC-SA 3.0, CC BY-SA 4.0, the Swift license, or none stated. Each forum sets its own terms. Forums on the default Discourse terms (such as the linked page) license user posts under CC BY-NC-SA 3.0; the Fedora and GNOME forums use CC BY-SA 4.0; the Swift forum uses the Swift license (Apache License 2.0 with the Runtime Library Exception); the Julia forum uses CC BY-NC-SA 3.0 up to 2020-03-06 and CC BY-SA 4.0 after; the remaining forums state no content license. The Docker forum terms forbid training competing AI models without consent, so its posts are marked training_ok no. Per file: license_override gives the license of the forum (and, for Julia, of the post date): spdx when one is stated, and notes quoting the forum terms page. Check the upstream terms before use.

Loghub: A Large Collection of System Log Datasets for AI-driven Log Analytics (Zenodo record 8196385): 19 files, 6.00 GB, Loghub license: research or academic use (Zenodo metadata: CC BY 4.0)

Files of Zenodo record 8196385: Loghub: A Large Collection of System Log Datasets for AI-driven Log Analytics (DOI 10.5281/zenodo.8196385). One pointer per file.

Batch	Files	Size	File types
`all`	19	6.00 GB	`gz`, `zip`

Link check: 19 of 19 files reachable with matching size.

License: Loghub license: research or academic use (Zenodo metadata: CC BY 4.0). The Zenodo record metadata gives CC BY 4.0, while the record description and the LICENSE file of the Loghub repository (https://github.com/logpai/loghub) state: "The datasets are freely available for research or academic work, subject to the following condition: For any usage or distribution of the loghub datasets, please refer to the loghub repository URL (https://github.com/logpai/loghub) and cite the loghub paper (Loghub: A Large Collection of System Log Datasets for AI-driven Log Analytics) where applicable." Training use and redistribution are therefore marked conditional. Refer to https://github.com/logpai/loghub, cite the Loghub paper (ISSRE 2023) and include the license notice in all copies. Check the upstream terms before use.

NVD CVE vulnerability records: 381,690 files, 1.95 GB, Public domain (NIST); CVE Terms of Use (MITRE)

CVE vulnerability records from the National Vulnerability Database (NVD, run by NIST), listed through the CVE API 2.0 by publication date (published before 2026-10-01). One pointer per CVE; the hash covers the CVE record (vulnerabilities[0].cve) as returned by the API. CVSS scores, CWE ids, status and the CISA KEV date are in source_meta.

Batch	Files	Size	File types
`all`	381,690	1.95 GB	`json`

Link check: not checked (API source).

License: Public domain (NIST); CVE Terms of Use (MITRE). NIST publications are in the public domain under Title 17 of the United States Code. CVE records are licensed under the CVE Terms of Use (https://www.cve.org/Legal/TermsOfUse): a perpetual, worldwide, royalty-free, irrevocable license to reproduce, prepare derivative works of and distribute them, provided the MITRE copyright notice and that license are reproduced. Display this notice: This product uses data from the NVD API but is not endorsed or certified by the NVD. Do not present modified records as NVD content. Check the upstream terms before use.

Apache Jira issues: 1,179,062 files, 1.28 GB, No license stated for issue content

Issues of all public projects on Apache Jira (issues.apache.org). One pointer per issue; the hash covers the issue description (fields.description) as returned by the API, as UTF-8 bytes. Title, type, status and attachment counts are in source_meta.

Batch	Files	Size	File types
`all`	1,179,062	1.28 GB	`txt`

Link check: not checked (API source).

License: No license stated for issue content. The issue tracker states no license for issue text. Check the upstream terms before use.

Zulip public chat archives: 1,300,638 files, 239 MB, No content license stated; Zulip Cloud Terms of Service

Messages in the web-public channels of public Zulip organizations (channels anyone can read without logging in). One pointer per message; the hash covers the message text (message.content, raw Markdown) as returned by the API, as UTF-8 bytes. Channel, topic, time and the numeric sender id are in source_meta.

Batch	Files	Size	File types
`all`	1,300,638	239 MB	`txt`

Link check: not checked (API source).

License: No content license stated; Zulip Cloud Terms of Service. The terms reserve ownership of messages to their authors and grant no license for training. The Zulip Cloud Terms of Service allow exporting content through the official API for legitimate purposes, such as a public archive approved by the organization's owners; other uses need written permission from Zulip. Check the upstream terms before use.

Ubuntu IRC logs: 456,106 files, 51.6 MB (size not published for 229,171 files), Canonical website terms (personal, educational and non-commercial use)

Daily logs of the Ubuntu IRC channels published at irclogs.ubuntu.com, from 2004 on (the channels moved from Freenode to Libera.Chat in 2021). One pointer per channel per day, to the plain text log.

Batch	Files	Size	File types
`all`	456,106	51.6 MB (size not published for 229,171 files)	`txt`

Link check: 8,834 of 8,846 files reachable with matching size. Other results: 12 unreachable. The check covered a random sample of 8,846 of the 456,106 files.

License: Canonical website terms (personal, educational and non-commercial use). The site is covered by the Canonical website terms, which allow displaying, downloading and printing its pages for personal, education and non-commercial use only. Messages remain the copyright of their authors. Check the upstream terms before use.

npm packages (in progress): 22,370,107 files, not published, Per version (license field of package.json)

Collection in progress: 22,370,107 files so far.

Tarballs of every version of every package in the npm registry. The package list comes from the replicate.npmjs.com changes feed; versions and hashes come from the package documents on registry.npmjs.org. Packages can contain malicious code; treat files as data and run nothing outside a sandbox.

Upstream title: npm registry

Batch	Files	Size	File types
`all`	22,370,107	not published	`tgz`

License: Per version (license field of package.json). Each package version declares its own license. Per file: license_override.notes gives the license field of each version; license_override.spdx gives the SPDX expression where one is recognised. Check the upstream terms before use.

UK Parliament Hansard debates: 1,370,783 files, not published, Open Parliament Licence v3.0

Debates of the House of Commons and the House of Lords from 1803 on, from the official UK Parliament Hansard API (hansard-api.parliament.uk). One pointer per debate section; uri is the section's debate API address. House, section, sitting date and title are in source_meta.

Batch	Files	Size	File types
`all`	1,370,783	not published	`json`

Link check: not checked (API source).

License: Open Parliament Licence v3.0. Parliamentary information may be copied, published, adapted and used commercially under the Open Parliament Licence v3.0. Personal data, third-party rights and the Royal Arms and Crowned Portcullis are not covered. Include the statement: Contains Parliamentary information licensed under the Open Parliament Licence v3.0. Do not suggest that Parliament endorses your use. Check the upstream terms before use.

Linux kernel mailing list archives: 405 files, not published, No content license stated

Public-inbox git repositories behind lore.kernel.org, listed from the grokmirror manifest that the official mirror source erol.kernel.org publishes for mirroring. One pointer per list per git epoch.

Upstream title: Linux kernel mailing list archives (lore.kernel.org)

Batch	Files	Size	File types
`all`	405	not published	`git`

Link check: 0 of 405 files reachable with matching size. Other results: 405 error.

License: No content license stated. Messages remain the copyright of their senders; the archive states no content license. The robots.txt of erol.kernel.org disallows all automated crawlers and names AI training crawlers. Not for training. Check the upstream terms before use.

Using the catalog

manifests/<source>/<batch>/: the source of truth. collection.json describes the batch and part-NNNNN.jsonl.gz holds its pointers, one line per file.
catalog/<source>.parquet: a derived view of the manifests, one flat table per source, shown in the viewer above; where it differs from the manifests, the manifests are right. Sources published while they are collected have the table split into files of up to 100 parts instead, under catalog/<source>/<batch>/.
catalog/curated/ and catalog/starter/: per use tag, the pointers left after curation, and a starter pack drawn from them.
assets/: the figures on this page.

Every field of the pointers, the batch headers and the catalog tables is described in docs/fields.md.

How to download a file and verify it against its size and hashes, including byte ranges, http_payload and api_body, is in docs/retrieval.md.

The license tiers, and the license of this catalog itself, are defined in docs/licenses.md.

To pick files by use, file type, license or size, start from the curated tables catalog/curated/<tag>.parquet or the starter packs catalog/starter/<tag>.parquet: docs/selecting.md has DuckDB queries and the download command, and docs/curation.md says what the curated tables leave out and why.

Changelog

Date	Change
2026-10-01	First release: 7 sources, 16,480 files, 4.19 TB
2026-10-01	Added Apache Jira issues: 42,021 files, 40.7 MB
2026-10-01	Added PyPI packages: 367,331 files, 579 GB
2026-10-01	Added Harvard Dataverse: 218,098 files, 24.2 TB
2026-10-01	Updated Apache Jira issues: 44,520 files, 48.6 MB (previously 42,021 files, 40.7 MB)
2026-10-01	Updated PyPI packages: 1,255,518 files, 2.39 TB (previously 367,331 files, 579 GB)
2026-10-01	Updated Harvard Dataverse: 238,095 files, 25.0 TB (previously 218,098 files, 24.2 TB)
2026-10-01	Updated Apache Jira issues: 226,438 files, 238 MB (previously 44,520 files, 48.6 MB)
2026-10-01	Added Debian bug reports: 841,630 files, 55.9 GB
2026-10-01	Added Discourse forum posts: 41,988 files, 38.8 MB
2026-10-01	Added Ubuntu IRC logs: 21,968 files, 1.62 MB
2026-10-01	Added NVD CVE vulnerability records: 26,835 files, 103 MB
2026-10-01	Added UK Parliament Hansard debates: 17,452 files, 0 B
2026-10-01	Added govinfo bulk data: 380,890 files, 7.22 GB
2026-10-01	Updated Debian bug reports: 2,244,932 files, 145 GB (previously 841,630 files, 55.9 GB)
2026-10-01	Added NapierOne and Digital Corpora scenarios: 1,056 files, 2.49 TB
2026-10-01	Added Linux kernel mailing list archives: 405 files, 0 B
2026-10-01	Updated Apache Jira issues: 1,179,062 files, 1.28 GB (previously 226,438 files, 238 MB)
2026-10-01	Updated govinfo bulk data: 419,274 files, 56.4 GB (previously 380,890 files, 7.22 GB)
2026-10-01	Added Zulip public chat archives: 1,300,638 files, 239 MB
2026-10-02	Added Debian source packages: 115,614 files, 99.2 GB
2026-10-02	Added CRAN packages: 25,299 files, 18.9 GB
2026-10-02	Updated Ubuntu IRC logs: 456,106 files, 51.6 MB (previously 21,968 files, 1.62 MB)
2026-10-02	Updated NVD CVE vulnerability records: 381,690 files, 1.95 GB (previously 26,835 files, 103 MB)
2026-10-02	Updated Harvard Dataverse: 354,800 files, 35.7 TB (previously 238,095 files, 25.0 TB)
2026-10-02	Added npm packages: 6,214,317 files, 0 B
2026-10-02	Updated Discourse forum posts: 5,167,761 files, 3.89 GB (previously 41,988 files, 38.8 MB)
2026-10-02	Updated UK Parliament Hansard debates: 994,124 files, 0 B (previously 17,452 files, 0 B)
2026-10-02	Updated PyPI packages: 23,792,575 files, 81.9 TB (previously 1,255,518 files, 2.39 TB)
2026-10-02	Updated UK Parliament Hansard debates: 1,370,783 files, 0 B (previously 994,124 files, 0 B)
2026-10-02	Updated PyPI packages: 21,693,722 files, 47.6 TB (previously 23,792,575 files, 81.9 TB)
2026-10-02	Added End of Term Web Archive: 337,638,097 files, 149 TB
2026-10-02	Added Common Crawl office files: 24,633,336 files, 19.3 TB
2026-10-02	Updated npm packages: 19,183,490 files, 0 B (previously 6,214,317 files, 0 B)
2026-10-02	Updated Discourse forum posts: 8,155,550 files, 6.18 GB (previously 5,167,761 files, 3.89 GB)
2026-10-03	Updated End of Term Web Archive: 5,000,000 files, 2.22 TB (previously 337,638,097 files, 149 TB)
2026-10-03	Updated Discourse forum posts: 440,917 files, 378 MB (previously 8,155,550 files, 6.18 GB)
2026-10-03	Updated npm packages: 51,735 files, 0 B (previously 19,183,490 files, 0 B)
2026-10-03	Updated End of Term Web Archive: 81,050,000 files, 35.8 TB (previously 5,000,000 files, 2.22 TB)
2026-10-03	Updated Discourse forum posts: 5,675,796 files, 4.33 GB (previously 440,917 files, 378 MB)
2026-10-03	Updated Common Crawl office files: 100,000 files, 77.8 GB (previously 24,633,336 files, 19.3 TB)
2026-10-03	Updated npm packages: 5,472,065 files, 0 B (previously 51,735 files, 0 B)
2026-10-03	Updated End of Term Web Archive: 88,300,000 files, 39.0 TB (previously 81,050,000 files, 35.8 TB)
2026-10-03	Added GovDocs1 files: 108,381 files, 62.5 GB
2026-10-03	Added NapierOne files: 350,556 files, 409 GB
2026-10-03	Updated Discourse forum posts: 8,958,648 files, 6.81 GB (previously 5,675,796 files, 4.33 GB)
2026-10-03	Updated Common Crawl office files: 9,600,000 files, 7.44 TB (previously 100,000 files, 77.8 GB)
2026-10-03	Updated npm packages: 13,970,278 files, 0 B (previously 5,472,065 files, 0 B)
2026-10-03	Updated GovDocs1 files: 976,399 files, 497 GB (previously 108,381 files, 62.5 GB)
2026-10-03	Updated NapierOne files: 400,566 files, 416 GB (previously 350,556 files, 409 GB)
2026-10-03	Updated Discourse forum posts: 8,998,807 files, 6.84 GB (previously 8,958,648 files, 6.81 GB)
2026-10-03	Updated Common Crawl office files: 18,000,000 files, 13.9 TB (previously 9,600,000 files, 7.44 TB)
2026-10-03	Updated npm packages: 22,370,107 files, 0 B (previously 13,970,278 files, 0 B)

License

Each file keeps the license of its source: see the license tiers and the license text of each source above, and the license columns of the catalog. This catalog itself (addresses, sizes, hashes and descriptive text) is released under CC0 1.0; that dedication does not extend to the files it points to. See docs/licenses.md.

The catalog itself contains no personal information. Some of the files it points to, such as email archives and Q&A posts, contain names and contact details that their authors published; follow each source's terms when you use them.

Questions and corrections are welcome in the Community tab.

Downloads last month: 624

Total file size:

79.2 GB