[MaxGhenis]

Stacked on #388 (itself on #387); merge after both. Adds packages/ostbelgien/kindergeld_2025, part of the Belgian transport port (ruling d048, B2e).

What it adds

35 facts from the article Kindergeld in Ostbelgien (update of 31.03.2026). The article's tables attribute them to the Ministerium der Deutschsprachigen Gemeinschaft ("Quelle: Ministerium der Deutschsprachigen Gemeinschaft").

Section	Facts	Coverage	Article wording
Children, payees, protected/new amounts (text)	6	2025-12-01 to 2025-12-31	"im Dezember 2025"
Protected/new amount shares (text, printed %)	4	2025-12-01 to 2025-12-31	"im Dezember 2025"
Age table	4	no dates; label "2025"	column "2025" only (ambiguity documented)
Supplements	5	2025-12-31	"zum Stand Ende Dezember 2025"
Birth and adoption premium payments (text)	2	2025-01-01 to 2025-12-31	"2025 wurde"
Expenditure table, EUR	14	2025-01-01 to 2025-12-31	"im Jahr 2025"

• Every fact has curated-CSV lineage plus two pointers into db/data/ostbelgien/kindergeld_2025/kindergeld-in-ostbelgien.article.html:
  • source_locator: the sentence, or the table, row and column.
  • source_evidence: a verbatim quote from the capture. That is the sentence, or the table row's own <tr> HTML, which occurs exactly once.
• The age-table total, the expenditure subtotals and the grand total are marked as totals. Supplement counts overlap and get no added total.
• The geography is ostbelgien:deutschsprachige_gemeinschaft (statistical_scope, nis_2025). It is named as the article prints it and defined by its nine NIS 2025 communes.
• The only shared-code change is one alias line in chronicle/source_package.py.

Choices for review

• Dates only where stated. Every fact keeps the coarse identity calendar_year:2025. period_coverage gives start and end dates only where the article states the span, and its notes quote that phrase.
  • The age table states no span, so it has the label and no dates. Its total matches the December count, but the article doesn't say whether it is an annual count or a December distribution.
  • The payee total comes from "Die 15.533 Kinder … verteilen sich auf 8.302 Empfänger", which refers back to the December child count.
• Entities.
  • Children and payees are person / benefit_recipient.
  • Premium payment counts ("700 … einmal ausgezahlt") are benefit_unit / premium_payment. That records the activity under the receiving unit, as scotgov/bus_coach_statistics_2024_25 counts passenger journeys under person / bus_passenger.
  • Spending is government / payer.
• Printed shares. These are kept as published facts, following the facts-only ADR and dwp/hbai_summary_fye2025. None is recomputed.
• Capture and reuse.
  • The site's Impressum states no data licence.
  • The capture is a 9.8 KB sanitized browser transcription of the article. Plain clients get a Cloudflare challenge, which was not bypassed.
  • It is committed as evidence, as other packages commit publisher HTML, for example db/data/kff/marketplace_effectuated_enrollment. Attribution is in the README and manifest.
• R2. Both files are stored in ledger-raw under their manifest keys and were verified by SHA-256 readback on 2026-10-24: the CSV, regenerated at 82487a9 with a new hash, and the capture.

Review history

• b658e93, GPT-6.1 Sol: REQUEST CHANGES. ledger-source-fidelity failed; ledger-boundary passed. All 35 values matched an independent parse. 82487a9 fixes the three findings:
  • exact dates the article doesn't state;
  • table locators that were not verbatim quotations;
  • an unsupported department name.

Stacking and rebase

Rebased from 200bf6d onto #388's head 099a4c1. git range-diff shows the two package commits unchanged apart from alias-registry context. Every file is byte-identical to 82487a9 except the registry, which differs by exactly this PR's alias line. #387, #388 and this PR merge as merge commits, in that order, each at its tested head. Until they merge, this PR's diff on GitHub also shows their commits.

Verification (at 82487a9)

• 7/7 tests in tests/test_source_ostbelgien_kindergeld_2025.py pass, run directly as functions in one process. A host reservation is holding back pytest on this machine.
• Five in-memory negative controls each fail the new checks:
  • a whitespace-altered row fragment;
  • another row's fragment;
  • a sentence missing the value;
  • a December fact dated 31 December;
  • an age fact given full-year dates.
• Regenerating with build_package.py reproduces every committed output byte for byte.
• validate-package ostbelgien-kindergeld-2025 --year 2025: 0 errors, 0 warnings.
• build-suite: valid, 35 facts.
• Ruff check and format pass.
• Bundle pins (8bd0af5, on top of Add SPF Finances municipal PIT surcharge rates, income year 2025 #387's and Add RSZ/ONSS employment by daily-wage class and annual wage key figures #388's): +35 facts, one statistical_scope geography (ostbelgien:deutschsprachige_gemeinschaft), calendar_year:2025 +35, entities benefit_unit 2, government 14 and person 19, one source, one source table and one built package, and no warning. Measured with Chronicle's bundle functions and confirmed against this PR's CI bundle on the old base. The full bundle test, replayed against Add Belgian exchange-rate and pension sources for Microcosm-BE transport #385's CI bundle plus Add ACS 2020–2024 county income source packages #386's skipped packages and Add SPF Finances municipal PIT surcharge rates, income year 2025 #387's, Add RSZ/ONSS employment by daily-wage class and annual wage key figures #388's and this package, passes; Add RSZ/ONSS employment by daily-wage class and annual wage key figures #388's test fails against it. CI runs the full bundle test on this head.

🤖 Generated with Claude Code

[MaxGhenis]

Independent review at 82487a9 (GPT-6.1 Sol, Subfleet run 20261024-011223-chronicle389-review-r2): APPROVE. ledger-source-fidelity: PASS. ledger-boundary: PASS.

The three findings from the b658e93 review are resolved:

• Dates: only spans the article states. December covers the month; the age table is undated.
• Evidence: 35/35 source_evidence strings occur verbatim in the capture. All 23 table fragments occur exactly once, with cells equal to [row label, printed value].
• Publisher: attribution follows the tables' "Quelle:" line.

Still to do before merge: the tests/test_chronicle_bundle.py pins, applied after #385, #387 and #388 land, and a review of that delta.

[MaxGhenis]

Delta review at 8bd0af5 (GPT-6.1 Sol, Subfleet run 20261024-042211-chronicle389-review-r3): APPROVE, no findings. ledger-source-fidelity: PASS. ledger-boundary: PASS.

• Rebase onto Add RSZ/ONSS employment by daily-wage class and annual wage key figures #388's head 099a4c1: only alias-hunk context changed. Every other file is byte-identical to the approved 82487a9, and the shared-code diff is exactly the Ostbelgien alias.
• Bundle pins (8bd0af5): derived independently, and every changed pin matches. +35 facts, calendar_year:2025 +35, benefit_unit +2, government +14, person +19, geography count 102,615 → 102,616 (statistical_scope:ostbelgien:deutschsprachige_gemeinschaft = 35), +1 source, +1 table and +1 package. Warning and duplicate counts are unchanged; the six ostbelgien.* dimensions can't conflict with Opgroeien's.
• Earlier reviews: b658e93 REQUEST CHANGES, fixed in 82487a9; 82487a9 APPROVE.

This merges as a merge commit after #387 and #388, at 8bd0af5, once CI passes.