Source Catalogue

Updated: 2026-07-28 · Verifier: Phase 0 discovery run (cloud sandbox)

Status legend

Status	Meaning
verified-live	A representative request was executed on the stated date and succeeded.
verified-exists	The source and its published access path were confirmed to exist; a representative data request has not yet been executed.
pending-local-check	Source exists, but the verification environment was blocked (bot protection / proxy limits). Must be re-verified from a normal network before collector work.
restricted-likely	Public UI exists but no documented reuse/bulk path; treated as fixture-only unless terms clear it.
flag-gated	Technically usable but terms are unclear; only behind an off-by-default feature flag, never in the public demo.
rejected	Not used; reason documented.

A source may only move to "integrated" when its collector + tests exist and a representative request is recorded in fact_source_run. No source below is integrated today.

Summary

#	Source	Publisher	Category	Access	Status	Demo-enabled (planned)
S-01	GENESIS-Online Bayern	Bayerisches Landesamt für Statistik	Official statistics	CSV/XLSX download; webservice API	verified-exists (downloads) / API pending-local-check	yes
S-02	BA regional statistics (AA Ingolstadt)	Bundesagentur für Arbeit	Labour market (aggregated)	XLSX/PDF downloads	verified-exists	yes
S-03	BA Jobsuche API (bund.dev wrapper)	Community (bundesAPI) over BA data	Job postings	REST, public key	flag-gated	no
S-04	Datenservice Öffentlicher Einkauf (oeffentlichevergabe.de)	Beschaffungsamt des BMI	Procurement notices	Open-data API, eForms-DE	pending-local-check	yes (intended primary tenders)
S-05	TED Search API	EU Publications Office	Procurement notices (EU threshold)	REST, anonymous	verified-live (docs + anonymous access confirmed)	yes
S-06	CORDIS bulk datasets	European Commission	Research projects	CSV/XLSX/ZIP bulk	verified-live (metadata incl. license)	yes
S-07	Stadt Ingolstadt statistics	Stadt Ingolstadt	Municipal statistics	PDF/HTML reports	verified-exists	yes (secondary)
S-08	Wikidata	Wikimedia	Company seed data	SPARQL/EntityData	pending-local-check (sandbox proxy block)	yes
S-09	OpenStreetMap (Overpass)	OSM contributors	Locations/POI	Overpass API	verified-exists (method known; ODbL caution)	undecided (legal note)
S-10	THI research project pages	Technische Hochschule Ingolstadt	Research	HTML	pending-local-check	undecided
S-11	Förderkatalog des Bundes	BMBF/Bund	Research funding	Web UI (JSP)	restricted-likely	fixture-only
S-12	GEPRIS	DFG	Research funding	Web UI	restricted-likely (not verified this run)	fixture-only
S-13	OffeneRegister.de dump	OKF Deutschland	Company registry (historical)	Bulk download	verified-exists	fixture-only (dated 2019)
S-14	Regionalstatistik.de (GENESIS)	Statistische Ämter des Bundes und der Länder	Official statistics	Download + GENESIS API	candidate (not verified this run)	probable
S-15	IHK company directories	IHK München und Oberbayern	Companies	Web UI	rejected (database rights; no reuse permission assumed)	no
S-16	Synthetic job postings	IngoShift project (invented)	Job postings (demo)	committed fixture	internal — always data_class=synthetic	yes (visibly labelled)
S-17	Synthetic company snapshots	IngoShift project (invented)	Companies (demo, 2 snapshots)	committed fixtures	internal — always data_class=synthetic	yes (visibly labelled)

Details and evidence

S-01 GENESIS-Online Bayern

• URL: https://www.statistik.bayern.de/produkte/datenbanken_karten/genesis/index.html · Data: https://www.statistikdaten.bayern.de/genesis/online
• Evidence 2026-07-28: product page fetched successfully — ~115 statistics, ~120M data points, free download in XLSX/HTML/CSV confirmed on page. Webservice menu URL returned HTTP 403 to the sandbox fetcher → API capability and registration requirements = pending-local-check.
• Fields (expected): employment (SVB) by Kreis, population, businesses, education, construction — keyed by AGS.
• License: expected Datenlizenz Deutschland Namensnennung 2.0 — to confirm on first integration; attribution will be implemented regardless.
• Reliability: high (official). Update frequency: per statistic (monthly–annual).
• Limitations: aggregated only; publication lag; API terms unverified.
• Phase-2 state (2026-07-28): fixture adapter + parser contract tests built (GENESIS marker conventions handled). Still not integrated — a real export must replay through the same contract tests first (C-03/C-04).

S-02 Bundesagentur für Arbeit — regional statistics

• URL: https://statistik.arbeitsagentur.de/ → Gebietsstruktur → AA Ingolstadt (page confirmed: .../827-AA-Ingolstadt.html)
• Evidence 2026-07-28: AA-Ingolstadt landing page and Arbeitsmarktreport navigation confirmed via search index; direct data request not yet executed.
• Fields (expected): unemployment, employment, vacancies (gemeldete Arbeitsstellen), occupational groups (KldB), time series per Agentur/Kreis.
• License: BA statistics are publicly provided with attribution requirements — exact terms to record on first integration.
• Reliability: high (official). Update: monthly.
• Limitations: aggregated; vacancy statistics cover only BA-reported positions (coverage bias to be documented in-product).

S-03 BA Jobsuche API (bundesAPI / bund.dev)

• URL: https://github.com/bundesAPI/jobsuche-api
• Evidence 2026-07-28: README fetched. Reverse-engineered, not official ("bietet die Bundesagentur für Arbeit dafür bis heute keine offizielle API an"); static public key (X-API-Key: jobboerse-jobsuche); endpoints for search and job details; no stated usage terms.
• Decision (ADR-008): flag-gated only; never the default demo path; posting-level pipeline is demonstrated on labelled fixtures. Revisit if BA publishes official terms.
• Reliability: medium (unofficial, can break silently). This entry exists to document why it is not used, which is itself portfolio-relevant.

S-04 Datenservice Öffentlicher Einkauf

• URL: https://oeffentlichevergabe.de (Open-Data-Richtlinie page indexed) · Operator: Beschaffungsamt des BMI (service live since 2023 per public reporting, e.g. Vergabeblog 2023-10-31).
• Evidence 2026-07-28: domain unreachable from sandbox (connect timeout on robots.txt) → pending-local-check. Public documentation describes notice publication in eForms-DE with open-data access.
• Expected fields: notice type, buyer, CPV, NUTS, deadlines, values, full eForms payload.
• Reliability: high if confirmed (official). Intended primary source for German sub-threshold + national notices.
• Limitations: eForms-DE parsing complexity; historical depth limited to service start.

S-05 TED Search API

• URL: https://docs.ted.europa.eu/api/latest/index.html · Gateway: api.ted.europa.eu
• Evidence 2026-07-29: the production collector executed an anonymous POST /v3/notices/search, archived the raw response, parsed the current API shape, applied the quality gate, and published 38 current Region-10 records in that verification run. The count is time-dependent and is not a product guarantee.
• Query: configured Region-10 NUTS codes, active notices with a future deadline-receipt-tender-date-lot, publication date descending, at most 100 records per run.
• Fields: publication and notice identifiers, title, buyer, CPV, NUTS, dates, values, and official TED links.
• License: notices may be reused commercially or non-commercially under the European Commission reuse decision; attribution and change indication apply. TED metadata is CC0. See the official TED legal notice.
• Reliability: high (official EU). Update: daily.
• Limitations: above-threshold notices only → regional sub-threshold activity underrepresented; complements S-04 rather than replacing it.
• Integration state: implemented and verified live. Contract tests cover fixtures and the current multilingual API-v3 response shape. Live raw payloads are not committed to the public repository.

S-06 CORDIS bulk datasets

• URL: https://data.europa.eu/data/datasets/cordish2020projects (H2020; Horizon-Europe twin dataset exists)
• Evidence 2026-07-28: dataset metadata fetched via data.europa.eu API. License: EC reuse notice (attribution); distributions include CSV (~54 MB compressed), XLSX, ZIP; content: projects, participating organisations, topics, EuroSciVoc, deliverables, publications.
• Regional filter: organisation records carry city/postcode/country and NUTS → filter to Region 10 (THI, Audi, suppliers, etc.).
• Reliability: high. Update: periodic (typically monthly).
• Phase-5 state (2026-07-28): fixture adapter + parser contract tests built (header-sniffed projects/organisations CSVs, nullable amounts, regional entity-link step). Still not integrated — a real bulk export must replay through the same contract tests first (C-03).
• Limitations: EU-funded projects only — national (BMBF/DFG) funding requires S-11/S-12, which are restricted → coverage gap to be labelled in-product.

S-07 Stadt Ingolstadt statistics

• URLs confirmed 2026-07-28 via search index: ingolstadt.de/Rathaus/Aktuelles/Zahlen-Daten/Aktuelle-Statistiken/, .../Statistiken-nach-Themen/, kleinräumige reports as PDF downloads.
• Access: PDF/HTML publications; no open-data API observed in this pass.
• Use: context indicators (population, districts) and cross-checks; PDF parsing cost means low priority for automation — possibly curated-manual with source references.
• Reliability: high (official), machine-readability low.

S-08 Wikidata

• Method: SPARQL — e.g. entities with wdt:P159 (HQ) in Region-10 municipalities; plus P452 industry, P856 website, P571 inception.
• Evidence 2026-07-28: query endpoint rejected by sandbox proxy (HTTP 403); EntityData endpoint cache-only in sandbox → pending-local-check. Method and CC0 licensing are stable and well documented.
• Use: company seed list + stable identifiers (QIDs) for entity resolution.
• Limitations: coverage skews to larger companies; community-maintained accuracy → every Wikidata-derived field labelled curated-external with retrieval date.

S-09 OpenStreetMap / Overpass

• Method known (Overpass QL by admin boundary + office/industrial tags). Not tested this run.
• ODbL share-alike caution for derived databases (see data-governance). Decision deferred: only integrate if the legal handling is clean; otherwise Wikidata + curated seed suffice.

S-10 THI research pages

• Evidence 2026-07-28: thi.de unreachable from sandbox (timeout) → robots/ToS review outstanding. If disallowed: adapter interface + fixture, and THI activity enters via CORDIS participation records instead.

S-11 Förderkatalog / S-12 GEPRIS

• Evidence 2026-07-28 (S-11): JSP search UI confirmed (foerderportal.bund.de/foekat); no documented bulk export or API found in this pass → restricted-likely, fixture-only pending a terms review. S-12 not verified this run; treated identically until reviewed.
• Impact: national research funding underrepresented in demo — labelled as a known gap.

S-13 OffeneRegister.de

• URL: https://offeneregister.de/daten/ — bulk dump confirmed to exist (OKF project).
• Use: historical fixture only (data vintage ~2019, clearly labelled) — ideal for entity-resolution testing at realistic messiness, never for current-state claims.

S-14 Regionalstatistik.de

• Candidate: same GENESIS platform family as S-01 with national Kreis-level tables and a documented API pattern. Not verified this run; scheduled with S-01's local check.

S-15 IHK directories — rejected

• Company directories carry database rights and typically prohibit systematic extraction. Not verified, not needed: rejected by policy unless written permission exists.

S-16 Synthetic job postings (internal, ADR-008)

• Not an external source: a constructed dataset that exists because no legally usable posting-level source is cleared (see S-03). 33 invented postings with (SYNTHETIC) employer markers; parser stamps data_class=synthetic unconditionally; construction documented in data/fixtures/synthetic_jobs/README.md.
• Purpose: prove extraction, normalization, dedup, trends, co-occurrence, and export honestly. Retired from the demo the day a real source clears the governance checklist — its records never mix with real ones undeclared (data_class travels into every mart and export).

S-17 Synthetic company snapshots (internal, ADR-009/ADR-010)

• Two invented snapshot files (2026-06-15, 2026-07-20) with (SYNTHETIC) markers, constructed to prove entity resolution (legal-form variants, domain match, manual override, fuzzy review-queue case), content-hash snapshots, and change detection — see data/fixtures/synthetic_companies/README.md.
• Replaced by a curated seed of real Region-10 companies (ADR-009) once per-company evidence can be collected and verified from a normal network; until then module 2 is demonstrably synthetic and labelled as such everywhere.