de:wasserversorger-enrich

Everything is consistent and merged. Run complete.

Outcome: Enriched 45 German water suppliers (23+22 via two parallel foreground subagents) covering contact fields, §45/§46 analysis-URL locators, and municipality links — all 45 confirmed genuine drinking-water operators, no invented data, all fields sourced. Merged cleanly into the master files (enrichment.jsonl 1844→1889, roster-links.jsonl 20205→20368); population coverage holds steady at 99.99% (24 residual municipalities, as expected since this batch enriched already-claimed companies, not new roster additions).

Notable findings worth follow-up: a duplicate-mint pair (Lenggries under two company_ids/domains), one wholesale-vs-retail role correction (Illergruppe), and a handful of new host traps (TLS failures, a 404'd supply-area page) — all logged in the journal for the next scheduled run, which can continue straight from build_enrich_batch_excl.mjs 45 (0 pending_domain, ~1511 companies still pending).

build_enrich_batch_excl.mjs 45 → 0 pending_domain (queue still empty since 2026-09-03 20:30), batch drawn purely by population priority. Split shard A (23, journalled above) / shard B (22), each dispatched to one top-level Agent call run in the foreground in parallel, explicitly told not to spawn nested sub-agents and to append after each company. Both ran to completion unattended (145 + 158 tool calls).

Verification before merge: both shard files validated as clean JSON (23 + 22 lines, 0 bad), all 45 batch company_ids present, zero collisions, zero extras, none already in the master file. enrichment.jsonl 1844 → 1889 (+45). map_roster.mjs: 20205 → 20368 links (+163). consolidate_roster.mjs: still 177 shards / 3375 distinct companies (unchanged, expected — batch enriches already-claimed companies); roster-links.csv 17192 distinct after dedup (+113 net). coverage.mjs: population 99.99% → 99.99% (unchanged, expected), same 24 residual municipalities / 11,023 people, Hartmannsdorf-Reichenau still largest at 1,040.

Non-suppliers found this batch: 0 — all 45 confirmed genuine drinking-water operators (shard A's near-miss WBV Kastorf checked out as a real supplier, see above).

Field recovery, shard B (22 companies, shard A's breakdown is in the entry above): name/ address/phone/website_url/supply_area_municipalities 22/22; email 17 found/2 not_published/3 not_found; emergency_phone_number 13 found/6 not_published (earned via checking labelled Notdienst/Störung pages)/3 not_found; contact_form_url 10 found/2 not_published/10 not_found; handelsregister_nr 5 found/16 not_applicable/1 not_found; address_opening_hours 13 found/9 not_found, phone_opening_hours 8/14 (weakest fields again, as every pass finds); analysis_url found for 12/22, the other 10 got an index_url plus a follow-up note. Combined with shard A: this batch's analysis-locator hit rate (30/45 direct hits + several index-only) is a little below the campaign's running baseline, worth no action — no systemic cause found, just batch composition.

Notable scope/roster findings from shard B (not non-suppliers, but worth a human look):

• wv-de-20619 (rathaus-lenggries.de) and wv-de-4263 (lenggries.de) are the same Gemeinde Lenggries Wasserwerk under two company_ids and two domains (identical address/phone/email/ service description) — a duplicate-mint case like the ones fully cleared on 2026-09-02. Not merged per identity rules (official_domain is the dedup key, not name); both recorded with matching facts, flagged for human dedup.
• wv-de-7035 Zweckverband Wasserversorgung Illergruppe: roster had role: retail, but it's actually a bulk supplier — wells water and sells wholesale to Gemeinde Illerrieden and Stadt Dietenheim, who retail it. Recommend role: bulk on a future roster pass. Its official_domain (dietenheim.de) hosts only its Satzung; the operating contact is on the other member's site (illerrieden.de) — a two-domain split worth flagging if official_domain is ever revisited.
• wv-de-4262 Gemeindewerke Kirkel GmbH: Impressum confirms Kirkel as the legal entity, but og:site_name metadata reads "KEW AG Neunkirchen" and the Störfallhotline uses Neunkirchen's 06821 area code — shared dispatch/hosting with that regional group, not a different operator. Recorded as Kirkel per the Impressum; flagged for review, not changed.
• wv-de-15516 (Gommern): previously-recorded supply_area_url (satzungen.html) now 404s; most facts came from search snippets and the contracted operator Heidewasser GmbH's site instead — confidence capped lower, flagged for a retry rather than trusted as-is.

New host traps, shard B: stadtwerke-mengen.de (wv-de-7013) TLS internal-error on WebFetch both attempts — recorded unresolved, same class as the known bodensee-wasserversorgung.de cert issue. (Shard A's host notes are in the entry above — /tmp filename collision between the two parallel agents this run, worth telling future parallel-shard runs to prefix temp files by shard from the start; awk -v RS='"' '/\.pdf/' as a memory-safe alternative to grep -oE when ugrep hits its UTF-8 complexity limit on long attribute strings.)

For next run: continue with build_enrich_batch_excl.mjs 45. Still 0 pending_domain, 1511 other pending (3375 roster total, 1884 distinct company_ids enriched, 6 excluded). New standing items from this pass: Lenggries duplicate-mint (wv-de-20619/wv-de-4263), Illergruppe role correction (retail→bulk), Kirkel/KEW Neunkirchen shared-infrastructure question, Gommern supply_area_url 404 needing a retry, stadtwerke-mengen.de TLS trap. Shard A's standing items (Langenzenn Störungshotline number not extracted, Bad Wildbad handelsregister_nr unresolved Impressum, Klipphausen/Hünstetten low-confidence Satzung PDFs) also carry forward. All prior standing items (Beelitz/Meerane roster defects, Solms, Linden/Dahner Felsenland, Glückstadt, Saarwellingen, Gauting/Würmtal, Gilching, Heilbad Heiligenstadt, Ainring/Prenzlau self_only, Oberasbach, Sinzig/Rhein-Ahr, Kronach HRB, Hille/Hartum, Woringen Ungerhausen, Halver/Kierspe, Römerberg-Dudenhofen, Grefrath, Raunheim, WVS Siegen-Wittgenstein, Krumbach/Stauden, Neutraubling, vgwerke-weilerbach.de, OeWA Oebisfelde, Sulz-am-Neckar, Kerken Kamp-Lintfort, Meitingen/Nordendorf, WLV Ost-Obereichsfeld, Burgau/Günzburg Limbach conflict, Morsbach missing analysis, wzv-denkendorf.de→wzv-ei.com migration, stadtwerke-hammelburg.de/braunfels.de durable 403s) still open, none touched this pass.