Harvested catalogues
Counts credit every catalogue that listed an entry, so a tool in three catalogues counts three times.Open items
- Nothing outstanding. Every step of the last run completed.
What changed
Steps of the last run
through the run logA run publishes only if every step exits 0, so a failed step means the public copy stays on the last good run rather than being overwritten with a partial harvest.
This list stops at the step that wrote it: the run log is recorded before this page is built, so the steps that follow — building this page and the API page, deploying, and committing the data — cannot appear on it. They are not hidden. That you are reading this page at all is the evidence the deploy step succeeded, since a failed run publishes nothing and you would be looking at the previous week's copy.
Surveyed and not harvested
Published with reasons, so nobody spends the same twenty minutes twice.302s to a Digital.gov policy page; api.code.gov returns the same HTML. The federal inventory that defined the code.json schema is gone. Its schema outlived it - Canada still uses it.
Operational as a SERVICE but empty as a catalogue. Tuleap Community Edition 16.12 (current), CORS-open REST API, 1,502 active public projects with real names (DigiLocker toolkits, eGov SmartCity, EPrabandhan) - but ZERO accessible source code. /api/projects/<id>/git returns an empty repositories array for 40 of 40 sampled projects; frs_packages is empty; the file-release pages say "empty"; and 0 of 25 git plugin pages carry a clonable URL. Not auth-gated - there is simply nothing published. Adding it would inject 1,502 entries with no repository URL, which is both the dedupe identity and the whole promise of the catalogue. Re-check periodically: if code lands, this becomes the largest single source here.
Live and substantial, but it is a national OSS PROMOTION portal, not a catalogue of government-produced software: developer contests, a contribution academy, licence verification, Open Up centre, news. Its /opensource/hub/<id> pages profile UPSTREAM projects (Node.js and the like) rather than Korean public-sector code, with no adoption data. Closest analogue here is SILL's recommendation axis, minus the government-use signal that makes SILL worth having. Nothing to ingest without changing what the catalogue means.
TLS certificate expired; /social/ 404s. Historically the most ambitious public software portal outside Europe - worth re-checking.
Centro de Transferencia de Tecnologia holds a real solutions directory, but every route is behind an F5/BIG-IP bot challenge (TSPD) that returns HTTP 200 with a CAPTCHA page instead of data - the HTML, the RSS feed and every /api/ path alike. Solving or bypassing a CAPTCHA is off the table, so this cannot be harvested as it stands. The legitimate route would be Spain asking to allowlist a harvester, or publishing the directory via datos.gob.es. Re-check rather than retry: the block is deliberate.
288,411 repositories with at least one government-email contribution across all 193 UN member states. A measurement dataset, not a curated catalogue - it answers 'who contributes' rather than 'what can we adopt'. Useful as discovery input for finding catalogues we have missed.
Checked properly rather than left open. The Software Heritage domain registry has collected 634 responding Cyprus government domains and NONE is a code platform or forge; there is no Cyprus entry in its github-gov-orgs.csv; and its candidate list names only parliament.cy. Keyword matches for git/code/repo across the subdomain list are false positives (digitalcoalition, reportdruginfo). Cyprus appears simply not to publish government source code centrally. EGDI rank 38 of 193, so this is a genuine gap rather than a discovery failure.
Reading the EU catalogue's SOURCE FACET NAMES from the page HTML resolved two of the three unknowns: hosting_platform:ogcio is Ireland's OGCIO, and hosting_platform:arte is Portugal's ARTE state technology agency - NOT the Franco-German broadcaster, which is what it was first read as here. Both are now ingested. Still unidentified: hosting_platform:city_of_ghent (a Belgian municipality, likely a handful of repos) and hosting_platform:dmrid_dits (github.com/DMRID is an individual user with one repo, not a registry).
English coverage dropped from 100% to 91% when OS2 Denmark and Bulgaria were added: 265 strings remain untranslated (171 Bulgarian, 82 Danish, 12 other). Recorded rather than papered over - every entry still carries description_lang so a consumer can tell, and meta.json reports the real figure. The Bulgarian set is mostly long EU-funding project titles rather than software summaries, so it needs judgement about what is worth translating at all.
The best discovery resource found so far, and it should have been checked first: a hand-curated, server-rendered directory of ~29 public-sector OSS catalogues with owner, language and geographic coverage. Unlike the EU OSS Catalogue on the same portal, this page renders fine. It confirmed 6 sources already ingested and surfaced these NOT yet evaluated in depth: OS2 (os2.eu, Danish municipal community - the Denmark source never found by guessing), dev.egov.bg (Bulgaria's e-government dev portal), ICT ReUse Belgium, Helsingborg City (SE municipal), Adullact (FR, runs gitlab.adullact.net), Forja redIRIS (ES academic) and OW2. publiccode.directory is a dead domain.
"Registry of Open Source Solutions" - a real national portal covering DPGs, open licences, approved git repositories and requirements for open source solutions. But it is WordPress with only stock post types (no custom solutions type in /wp-json/wp/v2/types) and the registry page contains zero repository links, so what is published looks like policy and guidance rather than a structured software list. Worth a closer read before writing off - it would be a new country.
Not a national catalogue but a CROSSWALK, and an open one: 780 entries in a single JSON with url_repository, wikidata, sill, cnll, framalibre and wikipedia ids on the same row. Would improve dedupe rather than add coverage - it maps SILL ids to Wikidata QIDs to repo URLs, which is exactly the identity resolution this catalogue does by hand.
NOT about open-source software. GlobE is UNODC's Global Operational Network of Anti-Corruption Law Enforcement Authorities, and "open-source registries" here means open-source INTELLIGENCE - publicly available official records such as company and beneficial-ownership registries. Its stated purpose: "Information from open-source registries is useful for anti-corruption law enforcement authorities... collected by investigators and prosecutors without the need of a formal mutual legal assistance request." Recorded because it surfaced in a software-catalogue search and reads plausibly relevant from the title alone.
A real register, and the one source here that is blocked only by paperwork: every read returns 401 without an x-api-key, and a key has not been requested. Not urgent - the Dutch CODE PLATFORM at code.overheid.nl is harvested and needs no key, and is arguably the better source anyway (first-hand repositories rather than a register of pointers). Recorded here because it was previously visible only as a '0 records' warning on the status page; when that page merged into this one the warning had nowhere to live, and a known gap with no home is a gap that gets forgotten.