docs/selecting.md

Selecting rows

Every table in this repository is Parquet. One row of data/<source>/<batch>/part-*.parquet is one downloadable item: a file, an archive member, a WARC record, a byte range or an API response. The tables under meta/ are small and say which rows are worth using.

A batch still being collected also has data/<source>/<batch>/part-tail.parquet: the rows that do not yet fill a numbered part file. It is replaced as collection goes on and its rows move into the next part-<number>.parquet once that file fills, so for tail rows row_file and row_index are not stable. Globs such as part-*.parquet or */*.parquet include it. A finished batch has no tail file: all of its rows are in numbered part files.

Row fields

Field	Meaning
source	Source name (see meta/sources.json).
batch	Batch inside the source (one snapshot or collection run).
collection	Collection name inside the source.
ext	File extension without the dot, as found at the source.
uri	Address to download from (HTTPS where the source offers it).
range_offset, range_length	Byte range inside the file at uri, for records stored inside a larger file.
member_path, parent_pointer_id	Path of a member inside an archive, and the archive's pointer.
zip_local_header_offset, zip_compression_method, zip_compressed_size	Where a zip member's bytes sit, so it can be read with range requests.
size_bytes	Size of the item in bytes (empty when the source does not say).
media_type	Media type when the source states one.
sha256, sha1, md5, crc32	Hashes as raw bytes; use lower(hex(sha256)) for the hex form.
origin_url, origin_time	The page the item was found on and when it was captured.
truncated	Why stored content was cut short (empty when it was not).
license_spdx, license_training_ok, license_redistribution_ok, license_notes	The row's own license when it differs from the batch license.
extra	Anything else the full pointer carries, as JSON; fetch.py rebuilds the full pointer from it.
row_file, row_index	Where the row sits: the data file (source/batch/part file) and the row number in it, counting from 0. Not stored in the data files; tools/fetch.py adds them when reading, the starter files carry them. meta/duplicates.parquet names rows by them.

Tables under meta/

• meta/batches.parquet: one row per source and batch: license tiers (training_ok, redistribution_ok), reachability from the last link check, the overall size cap used by tools/fetch.py (max_size_bytes), content categories, and the constants and rules used to rebuild full pointers.
• meta/duplicates.parquet: rows the default filter removes as duplicates, each named by its place in the data files (batch, row_file, row_index), with the reason and the copy that is kept (keep_*): natural_key, another row with the same source, uri, range_offset and member_path was kept (the one from the latest batch; within one batch the first in file order), and that kept row has natural_key_conflict true in tools/fetch.py output; superseded, the row belongs to a snapshot replaced by a newer batch (keep_batch); same_sha256, same_sha1, same_md5, another row with the same hash was kept.
• meta/bench_overlap.parquet: rows found in public benchmark tasks: scope row matches one row, scope container matches a whole collection; hit_class says how the match was made. When one row is a copy of benchmark data (benchmark_copy), its whole collection is listed only if the collection is one self-contained unit: one package, code repository, dataset or model repository (collection_kind "unit" in meta/sources.json, except the collections named in bucket_collections). Where collections are buckets (by file type, crawl, date, site and the like), only that row is listed. Only rows of files that were compared with the benchmark list can be found here (see meta/bench_checked.parquet).
• meta/bench_checked.parquet: which rows were compared with the benchmark list. It lists only what was checked: scope file has one row per data file (row_file) every pointer file behind which was scanned (reason scanned); the tail file of a batch still being collected is never listed, since it is replaced as collection goes on. Scope container lists collections whose name matches a benchmark and that await a ruling (reason pending_ruling). The one row with scope unlisted (reason not_listed) applies to every data file not listed here: files not compared yet and files published after this table was built. default_keep says whether the default filter keeps the rows.
• meta/tags.parquet: use tags (for example data-processing or help-requests) for each source, batch, collection, extension, metadata value and content category, with row counts and the reference size cap of the tag's content class (max_size_bytes).
• meta/size_caps.json: reference size caps: for each content class the typical size of benchmark task files of that class, and the overall cap over all benchmark task files; how each was derived (sample count, percentile, value) and how it is recalibrated; provisional. They do not remove rows: tools/fetch.py uses the overall cap when downloading, or the class caps with --cap per-class.
• meta/sources.json: per-source details, including the layout used to rebuild full pointers.
• meta/coverage.json: totals by license tier, source, unit, size and tag.

starter/<tag>.parquet holds a sample of about ten thousand rows per tag that pass the default filter, spread over sources (at most 30% from one source), extensions and size classes.

cards/<source>.yaml is a source card: what the files of one source are (types, structure, what a task could be built from, risks), how a sample was drawn and read, and for every field evidence.<field>.fill: auto (computed from the index, refreshed daily), judgment (from sample files that were downloaded and read, with the basis and the files used), derived (a fixed rule over other fields) or unfilled (no evidence yet). Read the evidence before relying on a field; ratios count the sample, not the whole source.

The default filter

The default subset only contains rows of data files that were compared with the benchmark list (AA-Briefcase-Lite, APEX-Agents, Agents' Last Exam, AndroidWorld, AutomationBench, BrowseComp, BrowseComp-Plus, CyberGym, DeepSWE, EnterpriseOps-Gym, Finance Agent, FrontierSWE v2, GDPval, Harvey LAB, MCP Atlas, MCPMark, Multi-SWE-bench, NL2Repo, OSWorld, OSWorld 2, SWE-bench Multilingual, SWE-bench Multimodal, SWE-bench Pro, SWE-bench Verified, SciCode, Terminal-Bench, Toolathlon, WideSearch, tau2-bench), as of 2026-10-05T14:11:22+0800, minus the copies of benchmark data found there and duplicate copies of the same file. Rows of files that were not compared, of the tail file of a batch still being collected, of files published after meta/bench_checked.parquet was built and of collections awaiting a ruling are left out; meta/bench_checked.parquet lists the data files that were compared and the collections awaiting a ruling, and any file it does not list counts as not compared. License tiers, sizes and versions do not remove rows; the queries below choose by them. Truncated rows, empty files and rows of batches whose link check was partial stay in; the reference counts below count them. A row passes when all of these hold:

• not_duplicate: the row is not listed in meta/duplicates.parquet (one copy of each duplicated file is kept; of rows that share a natural key only the latest is kept).
• not_in_benchmark_row / not_in_benchmark_container: the row and its collection are not in meta/bench_overlap.parquet with hit class benchmark_copy, task_file, data_address (copies of benchmark data found in the rows that were compared, including collections derived from a benchmark).
• checked_for_benchmarks: the row's data file is listed in meta/bench_checked.parquet as compared with the benchmark list (checked true), and its collection is not listed there as awaiting a ruling; rows of files not listed there (not compared yet, or published after that table was built) do not pass.

tools/fetch.py applies it unless you pass --no-default-filter. Size caps apply only when downloading: by default the overall cap of meta/size_caps.json (the 99th percentile of all benchmark task files) guards against huge files; --cap per-class uses the reference cap of the row's content class and --max-size N uses N bytes. A row above its cap is recorded as too_large and not downloaded; a row of unknown size is downloaded up to the cap.

Rows per source

Every row of the index is counted once: either it is in the default subset or it is removed by the first rule of the default filter it fails (duplicate copies first, then copies of benchmark data, then rows that were not compared with the benchmark list). Rows equals Default subset plus In a benchmark plus Duplicate copy plus Not checked.

Source	Rows	Default subset	In a benchmark	Duplicate copy	Not checked
ansible-galaxy-collections	52,784	52,784	0	0	0
apache-jira	1,179,062	1,049,601	0	129,461	0
arquivo-pt	85,980	85,977	0	3	0
artifact-hub-helm-charts	127,281	123,337	0	3,944	0
arxiv-source	3,165,721	3,165,721	0	0	0
bioinformatics-public-archives	278,202	278,202	0	0	0
centos-cbs-build-logs	113,676	113,676	0	0	0
chromium-luci-builds	8,665,027	7,926,998	0	738,029	0
common-crawl	148,118,699	138,519,418	71	9,599,210	0
conda-forge-packages	3,369,245	3,369,245	0	0	0
cpan-distributions	165,740	161,803	0	3,937	0
cran	25,299	25,284	0	15	0
crates-io	3,252,145	3,252,144	1	0	0
ctftime	75,162	75,101	0	61	0
datacite-dois	33,962,027	33,101,911	0	3,278	856,838
debian-bugs	2,244,932	2,244,932	0	0	0
debian-sources	115,614	111,237	0	4,377	0
debian-ubuntu-binary-packages	676,230	671,424	0	4,806	0
discourse-forums	12,203,622	11,876,328	0	128,823	198,471
docker-hub	593,368	364,320	0	229,048	0
egov-japan-data	292,050	292,050	0	0	0
end-of-term-web-archive	337,638,097	334,734,149	420	2,903,528	0
enron-spreadsheets-emails	14	14	0	0	0
eu-cellar-documents	1,921,815	1,910,370	0	11,445	0
european-parliament-plenary	1,233,677	1,228,256	0	5,421	0
f-droid-apps	196,376	196,370	0	6	0
fatcat-files	47,096,418	46,846,051	29	250,338	0
fedora-koji-build-logs	119,036	0	0	0	119,036
freebsd-bugzilla	573,793	571,150	0	2,643	0
fuse-spreadsheets	1	1	0	0	0
fuse-spreadsheets-common-crawl	249,376	242,373	0	7,003	0
gcc-bugzilla	186,889	186,706	0	183	0
gh-archive-hourly	136,799	136,799	0	0	0
github-gap-language-repos	7,028,005	6,469,659	0	0	558,346
github-release-assets	69,081	0	0	1,437	67,644
go-module-proxy-zips	59,120,216	59,113,479	6,737	0	0
govdocs1	1,000	1,000	0	0	0
govdocs1-files	986,279	986,279	0	0	0
govinfo-bulk-data	419,274	419,274	0	0	0
hackage-packages	154,127	153,030	0	1,097	0
hacker-news-threads	4,790,481	4,790,481	0	0	0
harvard-dataverse	948,347	0	1	33,906	914,440
homebrew-packages	81,275	76,896	0	4,379	0
hyperkitty-mailing-list-archives	80,365	80,365	0	0	0
if-archive	23,819	22,495	0	1,324	0
internet-archive-open-licensed	3,026,414	2,616,210	0	330,988	79,216
kaggle-file-listings	205,735	0	0	0	205,735
kde-webkit-gentoo-bugzilla	1,781,696	1,407,850	0	15,036	358,810
kernel-bugzilla	125,846	125,660	0	186	0
launchpad-ubuntu-bugs	97,333	0	0	1,972	95,361
lean-toolchains-and-packages	37,933	37,748	2	183	0
libreoffice-bugzilla	244,962	244,589	0	373	0
lichess	1,369	1,369	0	0	0
linux-kernel-mail-archives	405	405	0	0	0
loghub	19	19	0	0	0
mastodon-public-timelines	4,914,927	4,753,253	0	161,674	0
mendeley-data	130,156	0	0	7,171	122,985
modelscope	8,726,153	7,513,418	20,237	955,302	237,196
mozilla-bugzilla	3,896,234	3,872,911	0	23,323	0
mozilla-treeherder-ci-logs	37,531	0	0	0	37,531
napierone-digital-corpora	1,056	1,041	0	15	0
napierone-files	659,576	659,576	0	0	0
npm	51,767,788	51,106,932	0	482	660,374
nuget	636,597	563,278	0	0	73,319
nvd-cve	381,690	381,690	0	0	0
opendev-zuul-ci	27,332	27,332	0	0	0
openstreetmap-planet	27,295,635	27,295,635	0	0	0
osv-vulnerabilities	2,616,111	2,251,600	0	364,511	0
pmc-open-access	4,847,928	4,088,751	0	394,096	365,081
project-gutenberg	2,024,683	2,024,635	48	0	0
pypi	45,486,331	23,809,929	377	21,676,025	0
reddit-threads-arctic-shift	318,856	0	0	0	318,856
redhat-bugzilla	1,382,531	683,388	0	6,499	692,644
rosetta-code-github-mirror	148,991	148,991	0	0	0
rubygems	452,442	0	0	0	452,442
safedocs-cc-main-2021-31	13,457,931	12,179,826	0	1,278,105	0
sec-edgar-filings	18,788,783	18,788,783	0	0	0
smithsonian-open-access	9,311,739	9,311,739	0	0	0
solver-benchmark-instances	108,083	0	0	0	108,083
sourceware-public-inbox	168	168	0	0	0
stackexchange-api-posts	3,084,502	3,082,901	0	1,601	0
stackexchange-dump	3,629	364	0	3,265	0
ubuntu-irc-logs	456,106	456,106	0	0	0
uk-parliament-hansard	1,370,783	1,370,781	0	2	0
wikimedia-talk-dumps	1,563	1,563	0	0	0
worldbank-open-data	33,928	33,789	0	139	0
zenodo10k-slides	10,448	10,393	0	55	0
zulip-public-chats	1,300,638	1,226,067	0	74,571	0
all	891,018,987	845,105,380	27,923	39,363,276	6,522,408

Reference counts over the rows of the default subset (none of them removes a row): Size unknown, rows whose size the source does not state; Over class cap, rows larger than the reference cap of their content class (the largest max_size_bytes among their tags in meta/tags.parquet, otherwise the overall cap), which tools/fetch.py --cap per-class would not download; Truncated, rows whose stored content was cut short (field truncated); Empty file, rows whose stated size is 0 bytes; Link check partial, rows of batches whose last link check reached only part of the sample (reachability partial in meta/batches.parquet); and the rows in each license tier for training and for redistribution (the row's own tier when it has one, otherwise the batch tier).

Source	Size unknown	Over class cap	Truncated	Empty file	Link check partial	Training yes	Training conditional	Training no	Training unknown	Redistribution yes	Redistribution conditional	Redistribution no	Redistribution unknown
ansible-galaxy-collections	52,784	0	0	0	0	36,969	9	918	14,888	19,061	18,480	0	15,243
apache-jira	0	0	0	1	0	0	0	0	1,049,601	0	0	0	1,049,601
arquivo-pt	82,215	1,908	0	0	0	0	0	0	85,977	0	0	85,977	0
artifact-hub-helm-charts	123,337	0	0	0	0	21,886	0	0	101,451	19,295	2,591	0	101,451
arxiv-source	0	257,642	0	0	0	633,568	172,937	0	2,359,216	603,318	203,187	1,904,032	455,184
bioinformatics-public-archives	237,457	36,311	0	0	0	266,294	0	0	11,908	266,274	20	0	11,908
centos-cbs-build-logs	0	107	0	0	0	0	0	0	113,676	0	0	0	113,676
chromium-luci-builds	7,926,998	0	0	0	0	0	0	0	7,926,998	0	0	0	7,926,998
common-crawl	0	85,594,811	8,626,564	0	0	0	0	0	138,519,418	0	0	0	138,519,418
conda-forge-packages	0	85,579	0	0	0	3,176,854	0	0	192,391	2,448,828	728,026	0	192,391
cpan-distributions	0	22	0	2	0	115,204	0	0	46,599	11,658	103,546	0	46,599
cran	0	0	0	0	0	25,284	0	0	0	8,305	16,979	0	0
crates-io	0	0	0	0	0	3,129,852	16,827	0	105,465	2,843,317	303,362	0	105,465
ctftime	33,023	0	0	0	0	0	0	75,101	0	0	0	0	75,101
datacite-dois	33,099,731	1,608	0	0	0	32,844,639	3,156	0	254,116	32,844,584	3,211	0	254,116
debian-bugs	0	53	0	0	0	0	0	0	2,244,932	0	0	0	2,244,932
debian-sources	0	260	0	0	0	111,237	0	0	0	0	111,237	0	0
debian-ubuntu-binary-packages	0	8,640	0	0	0	622,961	0	0	48,463	0	622,961	0	48,463
discourse-forums	0	0	0	0	0	1,096,598	4,568,031	0	6,211,699	306,221	5,250,767	0	6,319,340
docker-hub	0	238,875	0	101	0	0	364,320	0	0	0	0	0	364,320
egov-japan-data	244,867	5,582	0	0	0	289,283	3	0	2,764	289,283	3	0	2,764
end-of-term-web-archive	0	53,625,978	0	0	0	334,734,149	0	0	0	334,734,149	0	0	0
enron-spreadsheets-emails	0	14	0	0	0	14	0	0	0	14	0	0	0
eu-cellar-documents	1,396,072	176,462	0	0	0	1,910,370	0	0	0	1,910,370	0	0	0
european-parliament-plenary	712,457	73,728	0	1	0	1,228,256	0	0	0	1,228,256	0	0	0
f-droid-apps	0	25,349	0	0	0	196,138	0	0	232	55,865	140,273	0	232
fatcat-files	26,966	41,935,035	0	0	0	0	0	0	46,846,051	0	0	0	46,846,051
fedora-koji-build-logs	0	0	0	0	0	0	0	0	0	0	0	0	0
freebsd-bugzilla	0	0	0	66	0	0	0	0	571,150	0	0	0	571,150
fuse-spreadsheets	0	1	0	0	0	1	0	0	0	1	0	0	0
fuse-spreadsheets-common-crawl	0	5,945	7,745	0	0	0	0	0	242,373	0	0	0	242,373
gcc-bugzilla	53	0	0	2	0	0	0	0	186,706	0	0	0	186,706
gh-archive-hourly	0	135,003	0	0	0	0	0	0	136,799	0	0	0	136,799
github-gap-language-repos	0	784	0	150,907	0	3,524,474	0	0	2,945,185	2,765,762	758,712	0	2,945,185
github-release-assets	0	0	0	0	0	0	0	0	0	0	0	0	0
go-module-proxy-zips	59,113,479	0	0	0	0	34,641,585	77,603	0	24,394,291	29,215,449	5,503,739	0	24,394,291
govdocs1	0	1,000	0	0	0	1,000	0	0	0	1,000	0	0	0
govdocs1-files	0	165,940	0	0	0	986,279	0	0	0	986,279	0	0	0
govinfo-bulk-data	0	16,192	0	0	419,274	419,274	0	0	0	419,274	0	0	0
hackage-packages	0	0	0	0	0	141,534	0	0	11,496	128,458	13,076	0	11,496
hacker-news-threads	4,790,481	0	0	0	0	0	0	4,790,481	0	0	0	4,790,481	0
harvard-dataverse	0	0	0	0	0	0	0	0	0	0	0	0	0
homebrew-packages	29,220	1,434	0	0	0	66,985	0	0	9,911	44,273	22,712	0	9,911
hyperkitty-mailing-list-archives	80,365	0	0	0	0	0	0	0	80,365	0	0	0	80,365
if-archive	0	150	0	0	0	0	0	22,344	151	0	0	22,344	151
internet-archive-open-licensed	0	145,408	0	0	0	32,210	11,137	0	2,572,863	31,323	12,024	0	2,572,863
kaggle-file-listings	0	0	0	0	0	0	0	0	0	0	0	0	0
kde-webkit-gentoo-bugzilla	0	0	0	333	0	0	0	0	1,407,850	0	0	0	1,407,850
kernel-bugzilla	0	0	0	22	0	0	0	0	125,660	0	0	0	125,660
launchpad-ubuntu-bugs	0	0	0	0	0	0	0	0	0	0	0	0	0
lean-toolchains-and-packages	872	22,220	0	0	0	21,258	0	0	16,490	21,218	40	0	16,490
libreoffice-bugzilla	0	0	0	240	0	244,589	0	0	0	0	244,589	0	0
lichess	0	390	0	0	0	1,369	0	0	0	1,289	80	0	0
linux-kernel-mail-archives	405	0	0	0	405	0	0	405	0	0	0	0	405
loghub	0	7	0	0	0	0	19	0	0	0	19	0	0
mastodon-public-timelines	1,144,744	0	0	0	0	0	0	1,807,989	2,945,264	0	0	0	4,753,253
mendeley-data	0	0	0	0	0	0	0	0	0	0	0	0	0
modelscope	0	1,704,415	0	0	0	6,343,508	412,468	0	757,442	6,209,100	543,210	4,097	757,011
mozilla-bugzilla	0	0	0	645	0	0	0	0	3,872,911	0	0	0	3,872,911
mozilla-treeherder-ci-logs	0	0	0	0	0	0	0	0	0	0	0	0	0
napierone-digital-corpora	0	592	0	0	0	1,009	0	0	32	0	1,009	0	32
napierone-files	0	35,555	0	0	0	614,567	0	0	45,009	0	614,567	0	45,009
npm	51,106,932	0	0	0	0	41,174,515	162,200	0	9,770,217	38,289,409	3,047,306	0	9,770,217
nuget	0	379	0	0	0	1	0	0	563,277	1	0	0	563,277
nvd-cve	0	0	0	0	0	381,690	0	0	0	381,690	0	0	0
opendev-zuul-ci	0	15	0	0	0	0	0	0	27,332	0	0	0	27,332
openstreetmap-planet	14,266,744	3,620	0	11	0	27,295,629	0	0	6	0	27,295,629	0	6
osv-vulnerabilities	0	10	0	0	0	451,106	0	0	1,800,494	381,508	69,598	0	1,800,494
pmc-open-access	0	897,081	0	0	0	2,371,295	1,317,229	0	400,227	2,371,266	909,071	408,187	400,227
project-gutenberg	942,231	1,659	0	11	0	2,001,011	0	0	23,624	0	2,001,011	23,624	0
pypi	0	209,692	0	1	0	19,530,642	59,074	0	4,220,213	16,599,815	2,989,901	0	4,220,213
reddit-threads-arctic-shift	0	0	0	0	0	0	0	0	0	0	0	0	0
redhat-bugzilla	0	0	0	105	0	0	0	0	683,388	0	0	0	683,388
rosetta-code-github-mirror	0	0	0	84	0	0	0	0	148,991	0	0	0	148,991
rubygems	0	0	0	0	0	0	0	0	0	0	0	0	0
safedocs-cc-main-2021-31	0	5,278,828	0	0	0	0	0	0	12,179,826	0	0	0	12,179,826
sec-edgar-filings	18,788,783	0	0	0	0	18,788,783	0	0	0	18,788,783	0	0	0
smithsonian-open-access	0	5,018,412	0	1,691	0	9,311,739	0	0	0	9,311,739	0	0	0
solver-benchmark-instances	0	0	0	0	0	0	0	0	0	0	0	0	0
sourceware-public-inbox	168	0	0	0	0	0	0	0	168	0	0	0	168
stackexchange-api-posts	0	0	0	0	0	0	0	3,082,901	0	0	3,082,901	0	0
stackexchange-dump	0	82	0	0	0	2	0	362	0	0	2	362	0
ubuntu-irc-logs	229,171	0	0	0	0	0	456,106	0	0	0	456,106	0	0
uk-parliament-hansard	1,370,781	0	0	0	0	1,370,781	0	0	0	1,370,781	0	0	0
wikimedia-talk-dumps	0	669	0	0	0	1,563	0	0	0	47	1,516	0	0
worldbank-open-data	484	2,429	0	0	0	33,333	299	0	157	33,149	483	0	157
zenodo10k-slides	0	3,266	0	0	0	10,211	143	0	39	9,988	366	0	39
zulip-public-chats	0	0	0	0	0	0	0	1,226,067	0	0	1,226,067	0	0
all	195,800,820	195,719,142	8,634,309	154,223	419,679	550,201,499	7,621,561	11,006,568	276,275,752	504,950,400	56,298,377	7,239,104	276,617,499

Querying with DuckDB

Run these statements once (DuckDB 1.1 or newer reads hf:// paths; set HF_TOKEN for higher rate limits). They define the views used below: p rows, b batches, d duplicates, o benchmark overlap, c benchmark check coverage, t tags.

CREATE OR REPLACE VIEW p AS SELECT * EXCLUDE (filename, file_row_number), regexp_extract(replace(filename, '\', '/'), '[^/]+/[^/]+/[^/]+$') AS row_file, file_row_number AS row_index FROM read_parquet('hf://datasets/Battam/agentic-env-index/data/*/*/*.parquet', union_by_name = true, filename = true, file_row_number = true);
CREATE OR REPLACE VIEW b AS SELECT * FROM read_parquet('hf://datasets/Battam/agentic-env-index/meta/batches.parquet');
CREATE OR REPLACE VIEW d AS SELECT * FROM read_parquet('hf://datasets/Battam/agentic-env-index/meta/duplicates.parquet');
CREATE OR REPLACE VIEW o AS SELECT * FROM read_parquet('hf://datasets/Battam/agentic-env-index/meta/bench_overlap.parquet');
CREATE OR REPLACE VIEW t AS SELECT * FROM read_parquet('hf://datasets/Battam/agentic-env-index/meta/tags.parquet');
CREATE OR REPLACE VIEW c AS SELECT * FROM read_parquet('hf://datasets/Battam/agentic-env-index/meta/bench_checked.parquet');
CREATE OR REPLACE VIEW selected AS
SELECT p.*, (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.pointer_id') WHEN coalesce(json_contains(json_extract(p.extra, '$._absent'), '"pointer_id"'), false) THEN NULL WHEN json_type(p.extra, '$.pointer_id') IS NOT NULL THEN json_extract_string(p.extra, '$.pointer_id') WHEN json_type(p.extra, '$.pid') IS NOT NULL THEN p.source || ':' || p.collection || ':' || json_extract_string(p.extra, '$.pid') WHEN p.parent_pointer_id IS NOT NULL AND p.member_path IS NOT NULL THEN CASE WHEN length(p.parent_pointer_id) - length(replace(p.parent_pointer_id, ':', '')) >= 2 THEN p.source || ':' || p.collection || ':' || substr(p.parent_pointer_id, strpos(p.parent_pointer_id, ':') + strpos(substr(p.parent_pointer_id, strpos(p.parent_pointer_id, ':') + 1), ':') + 1) || '!/' || p.member_path END WHEN b.pid_segments IS NOT NULL THEN p.source || ':' || p.collection || ':' || array_to_string(list_slice(string_split((CASE WHEN b.uri_swap_to IS NOT NULL AND json_extract_string(p.extra, '$._uri_orig') IS NULL THEN CASE WHEN starts_with(p.uri, b.uri_swap_to) AND ends_with(p.uri, b.uri_swap_suffix) AND length(p.uri) >= length(b.uri_swap_to) + length(b.uri_swap_suffix) THEN b.uri_swap_from || substr(p.uri, length(b.uri_swap_to) + 1, length(p.uri) - length(b.uri_swap_to) - length(b.uri_swap_suffix)) END ELSE p.uri END), '/'), -b.pid_segments, -1), '/') || CASE WHEN b.pid_at_offset THEN '@' || CAST(p.range_offset AS VARCHAR) ELSE '' END END) AS pointer_id, p.size_bytes IS NULL AS size_unknown,
  (CASE WHEN coalesce(nullif(CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.license_override.training_ok') WHEN json_type(p.extra, '$.license_override') IS NOT NULL THEN json_extract_string(p.extra, '$.license_override.training_ok') WHEN (p.license_spdx IS NOT NULL OR p.license_training_ok IS NOT NULL OR p.license_redistribution_ok IS NOT NULL OR p.license_notes IS NOT NULL) THEN p.license_training_ok WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"license_override"'), false) THEN b.const_training_ok END, ''), b.training_ok) IN ('yes', 'conditional', 'no') THEN coalesce(nullif(CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.license_override.training_ok') WHEN json_type(p.extra, '$.license_override') IS NOT NULL THEN json_extract_string(p.extra, '$.license_override.training_ok') WHEN (p.license_spdx IS NOT NULL OR p.license_training_ok IS NOT NULL OR p.license_redistribution_ok IS NOT NULL OR p.license_notes IS NOT NULL) THEN p.license_training_ok WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"license_override"'), false) THEN b.const_training_ok END, ''), b.training_ok) ELSE 'unknown' END) AS training_tier,
  (CASE WHEN coalesce(nullif(CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.license_override.redistribution_ok') WHEN json_type(p.extra, '$.license_override') IS NOT NULL THEN json_extract_string(p.extra, '$.license_override.redistribution_ok') WHEN (p.license_spdx IS NOT NULL OR p.license_training_ok IS NOT NULL OR p.license_redistribution_ok IS NOT NULL OR p.license_notes IS NOT NULL) THEN p.license_redistribution_ok WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"license_override"'), false) THEN b.const_redistribution_ok END, ''), b.redistribution_ok) IN ('yes', 'conditional', 'no') THEN coalesce(nullif(CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.license_override.redistribution_ok') WHEN json_type(p.extra, '$.license_override') IS NOT NULL THEN json_extract_string(p.extra, '$.license_override.redistribution_ok') WHEN (p.license_spdx IS NOT NULL OR p.license_training_ok IS NOT NULL OR p.license_redistribution_ok IS NOT NULL OR p.license_notes IS NOT NULL) THEN p.license_redistribution_ok WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"license_override"'), false) THEN b.const_redistribution_ok END, ''), b.redistribution_ok) ELSE 'unknown' END) AS redistribution_tier,
  coalesce((SELECT max(t.max_size_bytes) FROM t WHERE t.source = p.source AND t.batch = p.batch AND t.collection IS NOT DISTINCT FROM p.collection AND t.ext IS NOT DISTINCT FROM p.ext AND t.meta_value IS NOT DISTINCT FROM (CASE WHEN b.meta_key IS NOT NULL THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || b.meta_key) WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || b.meta_key) WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || b.meta_key) END) END) AND t.content IS NOT DISTINCT FROM CAST((CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract(p.extra, '$._row.content') WHEN json_type(p.extra, '$.content') IS NOT NULL THEN json_extract(p.extra, '$.content') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"content"'), false) THEN b.const_content END) AS VARCHAR)), b.max_size_bytes) AS class_cap_bytes,
  EXISTS (SELECT 1 FROM d WHERE d.reason = 'natural_key' AND d.keep_source = p.source AND d.keep_batch = p.batch AND d.keep_row_file = p.row_file AND d.keep_row_index = p.row_index) AS natural_key_conflict,
  coalesce((SELECT bool_and(c.checked) FROM c WHERE c.scope = 'file' AND c.row_file = p.row_file), (SELECT bool_and(c.checked) FROM c WHERE c.scope = 'unlisted'), false) AND NOT EXISTS (SELECT 1 FROM c WHERE c.scope = 'container' AND NOT c.checked AND c.source = p.source AND c.collection = split_part(p.collection, '@', 1)) AS checked_for_benchmarks
FROM p JOIN b ON b.source = p.source AND b.batch = p.batch
WHERE NOT EXISTS (SELECT 1 FROM d WHERE d.source = p.source AND d.uri = p.uri AND d.range_offset IS NOT DISTINCT FROM p.range_offset AND d.member_path IS NOT DISTINCT FROM p.member_path AND d.batch = p.batch AND d.row_file = p.row_file AND d.row_index = p.row_index)
  AND NOT EXISTS (SELECT 1 FROM o WHERE o.scope = 'row' AND o.hit_class IN ('benchmark_copy', 'task_file', 'data_address') AND o.source = p.source AND o.uri = p.uri AND o.range_offset IS NOT DISTINCT FROM p.range_offset AND o.member_path IS NOT DISTINCT FROM p.member_path)
  AND NOT EXISTS (SELECT 1 FROM o WHERE o.scope = 'container' AND o.hit_class IN ('benchmark_copy', 'task_file', 'data_address') AND o.source = p.source AND o.collection = split_part(p.collection, '@', 1))
  AND coalesce((SELECT bool_and(c.default_keep) FROM c WHERE c.scope = 'file' AND c.row_file = p.row_file), (SELECT bool_and(c.default_keep) FROM c WHERE c.scope = 'unlisted'), false) AND NOT EXISTS (SELECT 1 FROM c WHERE c.scope = 'container' AND NOT c.default_keep AND c.source = p.source AND c.collection = split_part(p.collection, '@', 1));

The query reads the row tables remotely and only the columns it needs; a full pass over every source moves a few gigabytes. Add p.source IN (...) or read data/<source>/*/*.parquet to stay small. The view adds training_tier and redistribution_tier (yes, conditional, no or unknown) and class_cap_bytes (the reference size cap of the row's content class) and natural_key_conflict (true when other rows with the same natural key were removed as duplicates and this one, from the latest batch, was kept) and checked_for_benchmarks (true when the row was compared with the benchmark list and its collection is not awaiting a ruling).

Ready-made queries

Each query runs on the selected view above.

1. Only rows that may be used for training (training tier yes or conditional):

SELECT p.source, p.uri, p.size_bytes, p.training_tier FROM selected p
WHERE p.training_tier IN ('yes', 'conditional') LIMIT 20;

python tools/fetch.py --tier yes,conditional --limit 20 --dry-run

2. Only the latest version of each package. Latest means the version released last (by release time, so a fix released later for an older line counts as the latest); all files of that version are kept, and rows without a version or a release time are left out. Package ecosystems and the release time used:

Source	Release time
pypi	source_meta.upload_time
npm	source_meta.published
crates-io	source_meta.created_at
cran	none (the snapshot lists only the current version of each package)
conda-forge-packages	source_meta.timestamp
ansible-galaxy-collections	source_meta.published
artifact-hub-helm-charts	source_meta.created
go-module-proxy-zips	source_meta.index_timestamp
hackage-packages	source_meta.uploaded

Not ordered by this query:

• debian-sources: rows carry no release time, and Debian versions need the dpkg ordering rules, which plain SQL does not have; each suite lists one version per source package, so choose a suite with source_meta.suite instead.
• nuget, rubygems: no release-time field is defined for these rows yet; every version is kept.

CREATE OR REPLACE VIEW latest_version AS
WITH v AS (
  SELECT p.*, concat_ws('/', p.collection, (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'chart') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'chart') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'chart') END)) AS package, (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'version') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'version') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'version') END) AS version,
    try_cast(CASE p.source WHEN 'pypi' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'upload_time') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'upload_time') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'upload_time') END) WHEN 'npm' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'published') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'published') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'published') END) WHEN 'crates-io' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'created_at') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'created_at') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'created_at') END) WHEN 'cran' THEN '1970-01-01T00:00:00Z' WHEN 'conda-forge-packages' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'timestamp') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'timestamp') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'timestamp') END) WHEN 'ansible-galaxy-collections' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'published') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'published') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'published') END) WHEN 'artifact-hub-helm-charts' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'created') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'created') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'created') END) WHEN 'go-module-proxy-zips' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'index_timestamp') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'index_timestamp') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'index_timestamp') END) WHEN 'hackage-packages' THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || 'uploaded') WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || 'uploaded') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || 'uploaded') END) END AS TIMESTAMPTZ) AS released
  FROM selected p JOIN b ON b.source = p.source AND b.batch = p.batch
  WHERE p.source IN ('pypi', 'npm', 'crates-io', 'cran', 'conda-forge-packages', 'ansible-galaxy-collections', 'artifact-hub-helm-charts', 'go-module-proxy-zips', 'hackage-packages')
), best AS (
  SELECT source, package, arg_max(version, released) AS version FROM (
    SELECT source, package, version, max(released) AS released FROM v
    WHERE version IS NOT NULL AND released IS NOT NULL GROUP BY ALL)
  GROUP BY ALL
)
SELECT v.* FROM v JOIN best ON best.source = v.source AND best.package = v.package AND best.version = v.version;
SELECT source, package, version, count(*) AS files FROM latest_version GROUP BY ALL ORDER BY 1, 2 LIMIT 20;

3. Only rows within the reference size cap of their content class (rows of unknown size stay in; tools/fetch.py --cap per-class stops their download at the same cap):

SELECT p.source, p.uri, p.size_bytes, p.class_cap_bytes FROM selected p
WHERE p.size_bytes IS NULL OR p.size_bytes <= p.class_cap_bytes LIMIT 20;

python tools/fetch.py --cap per-class --limit 20 --out ./files (rows above the class cap are recorded as too_large and not downloaded)

Examples

1. Rows tagged data-processing, CSV files only:

SELECT p.source, p.uri, p.size_bytes FROM selected p JOIN b ON b.source = p.source AND b.batch = p.batch WHERE lower(p.ext) = 'csv' AND EXISTS (SELECT 1 FROM t WHERE t.source = p.source AND t.batch = p.batch AND t.collection IS NOT DISTINCT FROM p.collection AND t.ext IS NOT DISTINCT FROM p.ext AND t.meta_value IS NOT DISTINCT FROM (CASE WHEN b.meta_key IS NOT NULL THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || b.meta_key) WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || b.meta_key) WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || b.meta_key) END) END) AND t.content IS NOT DISTINCT FROM CAST((CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract(p.extra, '$._row.content') WHEN json_type(p.extra, '$.content') IS NOT NULL THEN json_extract(p.extra, '$.content') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"content"'), false) THEN b.const_content END) AS VARCHAR) AND t.tag IN ('data-processing')) LIMIT 20;

python tools/fetch.py --tag data-processing --ext csv --limit 20 --out ./files

2. Rows tagged office-spreadsheets, smallest first:

SELECT p.source, p.uri, p.size_bytes FROM selected p JOIN b ON b.source = p.source AND b.batch = p.batch WHERE EXISTS (SELECT 1 FROM t WHERE t.source = p.source AND t.batch = p.batch AND t.collection IS NOT DISTINCT FROM p.collection AND t.ext IS NOT DISTINCT FROM p.ext AND t.meta_value IS NOT DISTINCT FROM (CASE WHEN b.meta_key IS NOT NULL THEN (CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract_string(p.extra, '$._row.source_meta.' || b.meta_key) WHEN json_type(p.extra, '$.source_meta') IS NOT NULL THEN json_extract_string(p.extra, '$.source_meta.' || b.meta_key) WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"source_meta"'), false) THEN json_extract_string(b.const_source_meta, '$.' || b.meta_key) END) END) AND t.content IS NOT DISTINCT FROM CAST((CASE WHEN starts_with(coalesce(p.extra, ''), '{"_row":') THEN json_extract(p.extra, '$._row.content') WHEN json_type(p.extra, '$.content') IS NOT NULL THEN json_extract(p.extra, '$.content') WHEN NOT coalesce(json_contains(json_extract(p.extra, '$._absent'), '"content"'), false) THEN b.const_content END) AS VARCHAR) AND t.tag IN ('office-spreadsheets')) ORDER BY p.size_bytes LIMIT 20;

python tools/fetch.py --tag office-spreadsheets --limit 20 --dry-run

3. Rows tagged sysadmin-devops, from the starter sample:

SELECT source, ext, count(*) FROM read_parquet('hf://datasets/Battam/agentic-env-index/starter/sysadmin-devops.parquet') GROUP BY ALL ORDER BY 3 DESC;

python tools/fetch.py --starter sysadmin-devops --out ./starter

python tools/fetch.py --print-sql ... prints the exact query the script runs, so any selection can be repeated in DuckDB. The script checks size and every hash of each download and writes fetch_manifest.jsonl with one line per selected row. With --check-content it also flags downloads that are hard to use (scanned PDFs, error pages saved as files, extensions that do not match the content) in the manifest field content_check; this needs the file command and poppler-utils, and a check whose tool is missing is skipped and named there.