Corpus completeness audit — what source documents are missing
Audited 2026-06-06. Scope: every collection under data/documents/ cross-referenced against the corpus’s own declared expectations — the PRR production tracker, the minutes manifest, the PRR-01 bundle index, and the provenance citations inside data/extracted/.
Method: (1) Substantive gaps read from the county’s own item-by-item response in
bosc-prr-production-2026-06-05.response-index.yaml— i.e. records requested and not produced, not inference. (2) Integrity gaps computed by script: minutes/agenda filename parsing + pairing overcommissioners/minutes/raw/, existence-checks of everyextracted/**, and the PRR-01 bundle’sderived_files. Ambiguous date-typo filenames were not auto-resolved (the heuristic produced impossible dates); they are listed for human verification. Minutes findings reflect the 2026-04-17 manifest snapshot, not today.
Integrity pass now automated. The mechanical half — existence + freshness across every catalogued dataset — is regenerated from the data catalog by
watermark catalog auditintodata/catalog/COMPLETENESS.mdand gated against drift bywatermark catalog check. This document keeps the substantive half (records an authority withheld), which is human judgment, not a script.
Headline
The archive itself is in good shape; what’s “missing” is mostly what the county hasn’t produced. Provenance is clean (94 distinct source PDFs cited across the extractions, 0 genuinely absent). The PRR production binaries are all present. The substantive holes are the records the county deferred or withheld — above all the entire county-wastewater engineering universe (PRR items 5–15) and the item-4 cost-benefit analysis — plus a modest set of corpus-hygiene issues in the minutes series (publication lag + ~12 malformed/misfiled filenames), none of which indicate lost evidence. A separate, newer thread (§2) tracks watershed/conservation grant docs worth digging up — the primary instruments behind a $650k Lost Creek (Maumee-headwater) ag-runoff grant whose SWCD summary is now in the corpus. A third thread (§4) tracked the OEPA air permit P0138965 — the secondhand-cited keystone behind the disclosed 313 MW power figure, and the expected source of the three-hall building footprint — which has now been ingested (2026-06-15): the 2026-05-28 final PTI (eDoc 4132514) confirms the genset count and three-hall emission-unit grouping on a primary footing, though the per-engine ekW behind 313 MW remains trade-secret-redacted (see §4).
Publishing vs. holding. What the corpus contains (this audit) is separate from what the public site serves. Every source document is default-deny on the public surface until a document publication review (epic #274; review checklist #281) clears its rel for the publish allowlist (#280) — a PII/legal-status pass, not byte redaction (chain of custody is immutable).
1. Substantively missing — records the county has NOT produced
Declared in the county’s first production tracker (cover letter 2026-06-05, Clerk of Board Brittany N. Woods). Of the 19-item request, these remain outstanding:
| Item(s) | Category | What’s owed | County’s stated posture |
|---|---|---|---|
| 5–15 | County wastewater works | BOSC pump-station/forcemain procurement; Shawnee II Phase 2; Shawnee Oaks / Hamlet of Hume sewering; permits DSWPTI-260294 & DSW-6756; 1996 CWA consent decree; Cridersville WWTP records | Produced 2026-06-12 (batch 2) — BOSC CMAR chain (#469-25→#137-26→#378-26), data-center-flows TO5 (#679-24), Shawnee Oaks (#113-26/#136-26), Hume (#135-26), Shawnee II Phase 2 (#220-24/#937-23). Outstanding = construction bids/award only (the CMAR is still at pre-construction). The other named items are already in-corpus: permits DSWPTI-260294 + DSW-6756 (data/extracted/permits/ 4074527/4074529/4074551, approved 2026-04-07, BOSC-1A private sanitary sewer); the 1996 CWA consent decree (regulatory/wastewater-enforcement-history.yaml — Civil Action 3:96 CV 7134, + source PDF); and Cridersville WWTP (NPDES OH0020222, 2.4 MGD → Little Ottawa, in the EPA ECHO inventory, now linked to the reroute economics in sanitary-economics.yaml). |
| 4 | Project BOSC | Cost-benefit analysis / projected tax-revenue impact / public-ROI inputs & assumptions | Withheld “being reviewed by our legal counsel for compliance with R.C. 149.43 and R.C. 9.66.” Not produced. |
| 2 | Project BOSC | County ⇄ DoD / federal-contractor (GDIT, GDLS) comms re the American Twp facility / corridor | ”No records”; county narrowed the ask. Deemed fulfilled. |
| 16 | Website | CMS audit trail / edit history for the Sanitary Engineering pages, 2025-01-01→ | “No records… do not manage the county website.” Contested — WordPress revision history exists behind a 401 gate (see below). |
| 17 | Website | Internal comms re adding/removing BOSC / RFP / Bistrozzi / Shawnee II references on county sites | ”No responsive records.” |
| 18 | Address | Records indexed to 4110 N. Cole St., Lima not caught by entity/project name | ”No responsive records” (partial). |
| 19 | Engineer of record | County ⇄ EMH&T comms | None as to Commissioners; may supplement from Sanitary Eng. |
The Category-B deferral — once the largest hole — was produced 2026-06-12 (batch 2; item-by-item in prr-mandamus/bosc-prr-production-2026-06-12.response-index.yaml). The production also resolved a corpus discrepancy: resolutions #113-26 / #135-26 / #136-26 had been mislabeled “forcemain feasibility” — the primary records show #113-26 = Shawnee Oaks engineering (Access, $161k), #135-26 = Hamlet of Hume WPCLF ($2.0M), #136-26 = unincorporated Shawnee Twp / Shawnee Oaks WPCLF ($1.6M). These records tie to the OEPA Shawnee II permit 2PK00002 in data/extracted/oepa/ and the ECHO discharger inventory in data/reference/echo/. They also yield the data-center → WWTP keystone (Task Order #5, #679-24 — a “Data Center Flows Treatment Evaluation” of the campus blowdown stream).
Item 16 is “missing” by custody disclaimer, not non-existence. The response index’s own rebuttal shows the Sanitary pages’ WordPress modified dates fall inside the requested window and the /revisions endpoint returns HTTP 401 (gated, not 404) — the version history exists; the county simply disclaims holding it. Custody sits with the host (AhelioTech / CorpComm-built site) — see allen-county-web-vendor-audit.md and allen-county-level-sites.md.
Universe note: per the relator’s 2026-06-03 narrowing email, any request item not listed was treated as withdrawn — so the “owed” set is bounded by these items. Production is rolling (“every Friday”). Items 5–15 were produced in batch 2 (2026-06-12); item 4 (cost-benefit, held under §9.66(D)) remains the one to watch.
Allen SWCD production (2026-06-12) — site stormwater; two “no records” tensions
Added 2026-06-12. A separate agency — Allen Soil & Water Conservation District — answered the relator’s own 11-item (Parts A–E) request about the 4110 N. Cole St. site’s stormwater/erosion jurisdiction. Ingested as prr-mandamus/bosc-prr-production-2026-06-12-aswcd.response-index.yaml (+ analysis); binaries (4-pp letter + 54-pp records bundle) under prr-production-2026-06-12-aswcd/. It produced the site ESC inspections (2025-12 → 2026-06), the County Engineer’s mass-grading approval + SW1225 stormwater permit ($5,800), and the plan-review email chain — and withheld the plan sets under a new dual ground (R.C. 149.433 infrastructure records + R.C. 1333.61 trade secret — “water and wastewater usage for a data center”), now withholding-map layer 7.
What’s owed / unresolved from this track:
| Item(s) | What | SWCD posture |
|---|---|---|
| 1 | The NPDES CGP coverage number for the site | SWCD didn’t produce it (“TBD” on every inspection form; pointed to Ohio EPA) — but acquired direct from Ohio EPA 2026-06-16: Turner’s coverage 2GC08468*AG, effective 2025-11-10 (permits/lma1a-npdes-cgp-coverage). The “TBD” was a SWCD recording gap, not absent coverage. Resolves #143. |
| 3 | Wetland determinations (the 0.7-ac forested wetland, DSW401251760W) | “No records” — yet the produced inspections record “the existing wetland was mitigated.” A produced record contradicts the answer. (Corroborated by in-corpus permits/3788677 + /3796349.) |
| 4 | Tile/agricultural-drainage impact | ”No records” — yet the 2026-06-05 inspection documents an east farm-tile diversion swale failure (photo “East farm tile bypass”). A produced record contradicts the answer. |
| 6–11 | BOSC-1A forcemain, Shawnee II Phase 2, Hume/Shawnee forcemain (MGD design capacity), MS Consultants comms, Commissioner Beth Seibert comms | ”No records” → deferred to Ohio EPA / Allen County Sanitary Engineer / townships. |
The plan sets are the same documents the County withholds, now shielded a second way; the §149.433/trade-secret ground reaches even the plan-share links inside produced emails (redacted). Items 3 and 4 are the adequacy flags to carry forward.
The item-1 “TBD” feeds a permit-vs-disturbance ordering reconstruction — prr-mandamus/bosc-site-permit-sequence.yaml (+ narrative), #150: active clearing/mass-grading is documented from 2025-12-08 on a 195-ac footprint with the NPDES CGP number “TBD” through 2026-06-05, and the broader Level-2 wetland fill (DSW401252260W) was filed 2025-12-09 — the day after clearing — then found incomplete 2025-12-23. Both are framed as open questions; the dispositive CGP coverage record is pursued under #143.
NPDES coverage thread RESOLVED (2026-06-16). Three pieces were ingested 2026-06-16. (1) The governing statewide general permit OHC000006 (+ its Response to Comments) is committed at ../regulatory/ohc000006-construction-stormwater-gp.yaml — the standard is now primary-source: a complete NOI is due “at least 21 days … prior to the commencement of construction activity” and “Coverage under this permit is not effective until an approval letter … is received.” (2) The campus coverage record itself (../permits/lma1a-npdes-cgp-coverage.epa.yaml, Ohio EPA facility file): Turner’s 2GC08468*AG effective 2025-11-10 (NOI 2025-10-29), Igel added as co-permittee (*BG, 2025-11-12), modified to 309.2 ac 2026-06-10 — coverage attached ~4 weeks BEFORE the 2025-12-08 documented disturbance, so the SWCD’s “TBD” was a recording gap, not absent coverage (closes #143 + #154). (3) The BOSC Storm Outfall’s own coverage is end-to-end on the record too — eDoc 4091289, Facility Permit Number 2GC08747*AG (effective 2026-04-22; Igel/WSP). Note the per-site number format is 2GC…*AG, not an OHC… number (OHC000006 is the general permit).
Cross-production referral map (#151)
With three producing bodies now on the record — Commissioners (batch 1), Sanitary Engineer instruments (batch 2), and the ASWCD — the “no records” answers form a referral structure: each body disclaims a slice and points to the next. Mapped item-by-item (who disclaimed, who they pointed to, whether anyone produced it) in cross-production-referral-map.yaml.
| Item(s) | Disclaiming body | Pointed to | Producer in corpus? |
|---|---|---|---|
| County 5–15 | Commissioners | Sanitary Engineer | Yes — produced in batch 2 (the referral that didn’t dead-end) |
| County 16 (CMS edit history) | Commissioners | third-party web host | No — records exist at the host (WordPress; REST API 401, not 404) |
| County 19 (EMH&T comms) | Commissioners | Sanitary Engineer | No County↔EMH&T comms produced (ASWCD produced its own EMH&T emails) |
| ASWCD 6 (BOSC-1A forcemain NPDES/SWPPP/ESC) | ASWCD | Ohio EPA / Sanitary Eng | No — batch 2 gave procurement, not the environmental-stormwater records |
| ASWCD 8 (Shawnee II Ph2 stormwater/ESC) | ASWCD | Ohio EPA / Sanitary Eng | No — batch 2 gave the upgrade resolutions, not the ESC/NPDES records |
| ASWCD 9 (Hume/Shawnee forcemain MGD capacity) | ASWCD | Ohio EPA / Sanitary Eng | No — batch 2 gave financing + the engineering contract, not the MGD figure |
| ASWCD 7 (forcemain drainage/tile/wetland) | ASWCD | Sanitary Eng / townships | No — townships not yet a requested custodian |
Headline: the corridor-level environmental records (NPDES construction-stormwater / SWPPP / ESC for the forcemain corridors and Shawnee II Phase 2, plus the item-9 MGD design capacity) are owned by no county body — each points to Ohio EPA or the townships. The County’s Sanitary-Engineer batch produced the procurement / financing / resolution layer of those same projects but not their environmental-permit layer. A routing map, not an accusation — but it isolates exactly where the relator must go next (Ohio EPA; the townships) and which referral already resolved (Commissioners→Sanitary, items 5–15).
2. Watershed / conservation grant docs — to dig up
A captured Allen SWCD project page entered the corpus 2026-06-06 as
watershed/osu-lima-ag-runoff.allenswcd.2026-06-06.pdf
(extraction: watershed/osu-lima-ag-runoff-treatment.capture.yaml).
It documents a $650,000 ag-runoff retention/treatment grant (Project
ALLENSWCD-FDFARM22) on Lost Creek HUC12 041000070305 — a Maumee headwater
2.7 mi east of I-75 — i.e. baseline nutrient-burden context for the hydrology
axis. The page is a secondary SWCD summary.
The project’s Lead PI, Dr. Vinayak S. Shedekar (OSU FABE), was identified from
the page; his CV (shedekar-cv.2025-09-28.pdf,
provided 2026-06-06; bibliography at
../watershed/shedekar-vinayak.bibliography.md)
resolved the funding source: Great Lakes Restoration Initiative (GLRI), subaward
through Ohio EPA, OSU portion $327,450, term 2023–2025 (CV grant #13) —
closing the prior [open]. The primary instruments still owed:
| Item | What’s owed | Where to get it |
|---|---|---|
| Grant agreement / award instrument | The signed GLRI / Ohio EPA subaward itself — deliverables, match, reporting (program now known; instrument not in hand) | Ohio EPA GLRI subaward files; Allen SWCD; public-records request |
| OSU application & conceptual design | Engineering basis for the BMPs (berm/wetland/forebay/pump-vault; saturated buffer) | OSU FABE (Dr. V. Shedekar); Allen SWCD project file |
| Load-reduction estimate table | Referenced on the captured page (“provided in the table below”) but not transcribed in the capture | Re-capture full page / SWCD project file |
| OSU monitoring data | Continuous flow + water-quality (3 sites, ISCO6712 samplers) — quantifies actual nutrient/flow reduction | OSU Overholt Drainage program |
These quantify the existing Maumee-headwater load and remediation spend the
corpus reasons against; the funding source is now verified, the rest not yet in
hand. Cross-ref the ECHO Maumee discharger inventory
(../../reference/echo/) and
../../../docs/HYDROLOGY.md.
3. Corpus-integrity gaps (minor)
Commissioners minutes series — broadly complete
934 files in commissioners/minutes/raw/ spanning 2023–2026: 475 agendas (A…) / 453 minutes (M…), 468 distinct meeting-dates parsed. The 22-file agenda/minutes asymmetry is mostly not real absence:
- Recent publication lag (not a gap): 12 dates 2026-03-03 → 2026-04-20 have an agenda but no minutes — these are the most recent meetings relative to the 2026-04-17 manifest snapshot; minutes simply weren’t posted yet.
- Partial 2023 backfill (scope, not loss): an 8-date cluster in Dec 2023 (agendas, no minutes) sits in the partially-backfilled 2023 range (the corpus is pinned to 2024–2026).
- Typo’d agendas masquerading as gaps: several early-2024 “minutes, no agenda” dates do have an agenda under a malformed name — e.g.
A20524-Special.pdf(2024-02-05),A22124-Special-1.pdf(2024-02-21),A22824-Special.pdf(2024-02-28).
~12 malformed / misfiled filenames (break automated indexing)
| Filename | Issue |
|---|---|
Govt.-Structure-Org-Chart-1.pdf | Not a meeting record — reference handout misfiled in minutes/raw/ |
Mayors-and-Zoning-Persons.pdf | Not a meeting record — contact list misfiled |
Township-Trustees-Fiscal-Officers.pdf | Not a meeting record — contact list misfiled |
CANCELLED-MEETING.pdf | Cancelled-meeting placeholder (no date) |
Board-of-Commissioners-General-Session-September-9-2025-Meeting-Minutes.pdf | Long-form name for the 2025-09-09 minutes (should be M090925.pdf) |
ACC-M102423.pdf | Stray ACC- prefix (minutes 2023-10-24) |
A0101024.pdf, A0404024.pdf, M0115226.pdf | 7-digit date typos — ambiguous; verify against the source before renaming |
A20524-Special.pdf, A22124-Special-1.pdf, A22824-Special.pdf | 5-digit dates (leading zero dropped) |
PRR-01 bundle derived files — not committed (low concern)
The bundle index references seven derived_files (Allen_County_PRR_searchable.pdf, ..._full_text.txt, ..._exhibit_index.txt, Allen_County_Project_Master_Table.md, WATERMARK_Tetra_Tech_OPC.yaml, WATERMARK_OPC_Detailed_Line_Items.yaml). None are in data/documents/aedg/ — but they are regenerable derivatives of PRR-01-bundle.ocr.pdf, which is present. Not lost source material.
Sanitary as-built present but not yet extracted
Added 2026-06-11 (issue #41). sanitary/indianbrook-ps-asbuilt-2007.pdf — the 2007 Indian Brook pump-station as-built (4-page scan, no text layer) — is present in the corpus but not yet grounded: data/extracted/sanitary/ carries no structured extraction, so the 8” forcemain / three-phase upgrade it documents is cited only secondhand (data/reference/periplus/watch-items.geojson “2007 as-built”; ../../../docs/legal/mandamus-analysis.md §503). The discipline-agnostic kind=engineering / kind=sanitary extractor (watermark.pipeline.extract.extract_engineering) now exists; the structured .sanitary.yaml awaits a keyed vision pass (ANTHROPIC_API_KEY, tracked in #124). This is the one piece of sanitary as-built evidence that is in hand against the Category-B wastewater hole above (§1, items 5–15).
Provenance — clean
Every *.pdf cited across extracted/** resolves to a real file under data/documents/. (Of 94 distinct cited names, the only 2 “unresolved” are false positives: a Google-Slides export title and a prose fragment, not file references.) The 5 PRR-production binaries named in the response index are all present in prr-production-2026-06-05/ — the index’s “binaries to be added” note is stale.
4. Air permit (PTI P0138965) — keystone power figure — INGESTED 2026-06-15
Added 2026-06-09 (compute-capacity axis); resolved 2026-06-15. The Ohio EPA Air Permit-to-Install P0138965 (Facility 0302022054) is the keystone behind the campus’s disclosed electrical scale — 114 emergency generators × 2,750 ekW ≈ 313 MW backup → ~250–300 MW IT (N+1) — anchoring both the cooling-water balance (../../../docs/HYDROLOGY.md; watermark.hydrology.cooling) and the compute / AI-capacity derivation (../../../docs/COMPUTE.md; watermark.facility).
Now ingested. The 2026-05-28 final PTI (Ohio EPA eDocument 4132514, 66 pp incl. a 64-item Response to Comments) is committed at ../../documents/permits/bistrozzi-permits/4132514.pdf → ../permits/4132514.epa.yaml. It joins the already-committed 2025-12-10 draft of the same permit (3987141 / 3987144) — which it supersedes and whose flagged 114-vs-115 generator discrepancy it resolves.
What it puts on a primary footing:
- Genset count + the three-hall grouping. 115 emissions units P001–P115 — P001–P114 are identical data-hall gensets in three groups of 38 (GEN 1/2/3), P115 is a separate, smaller HUBGEN; the 36 cooling towers are three groups of 12 (TWR 1/2/3). The three-group emission-unit structure corroborates the anticipated ≈ three data halls (38 gensets + 12 towers each).
- Synthetic-minor caps: NOx 235.62 tpy + CO 96.06 tpy (rolling 12-month, P001–P115 combined) — the federally enforceable limits keeping the facility under major-source NSR (it is major for Title V); Tier 2 CI engines under 40 CFR 60 Subpart IIII, fueled ULSD/HVO ≤ 15 ppm S.
What it does NOT resolve (carry forward):
| Still owed | Why | Where to get it |
|---|---|---|
| Per-engine ekW / engine make-model — CBI-LOCKED, confirmed 2026-06-16 (#159) | The DAPC application file was acquired: Ohio EPA granted trade-secret protection for the “size/rating of emergency generators and fire pumps” (OAC 3745-49-03, grant 2025-10-08; permits/3859883 + justification 3859888). The exact ekW lives only in the confidential version of A0080278; the 2,750 ekW × 114 ≈ 313 MW basis stays the draft public-notice figure and is not obtainable absent a legal challenge to the designation. Surviving public brackets: main gensets ≥ 560 kW (Tier 2). | Closed — only a trade-secret challenge would unlock it |
| Emission-unit plot plan / building footprint — also trade-secret-shielded (#160) | The same trade-secret grant covers “internal layout details” — so the footprint is shielded at the DAPC level too (atop the County/ASWCD R.C. 149.433 + 1333.61 withholding). Method 3 stays the land-area envelope. | Township building-permit filings; a trade-secret/PRR challenge |
| Architectural site-plan sheets (CI Design / WSP) | The committed plan set is a single grading & storm sheet (1A-C-3104) showing only ancillary SSS/GPS buildings on piers — the data-hall footprints sit on architectural sheets not in hand (see #160). | EMH&T / CI Design plan set; PRR follow-up |
Net: the genset count, three-hall emission-unit grouping, and synthetic-minor caps are primary-source; the 313 MW per-unit ekW is now confirmed trade-secret-locked (#159 closed — the application file withholds it under an Ohio EPA OAC 3745-49-03 grant), so it permanently rests on the draft public notice absent a legal challenge; and Method 3 stays the land-area envelope — the floor area / internal layout is itself trade-secret-claimed (#160).
Genuine-absence shortlist — VERIFIED 2026-06-12 (#46)
All 8 low-confidence dates were checked against commissioners.allencountyohio.com
(year-specific minutes + agenda archives, raw-link inspection). None is a corpus
capture lag — our holdings already mirror what the county publishes. Outcomes are
recorded per-date in ../commissioners/minutes/filename-map.yaml
under genuine_absence_verified:. Summary:
- Minutes present, no agenda → special-session structure (no separate agenda is ever published):
2023-03-22, 2023-06-21, 2025-08-13, 2025-11-14 — all confirmed “-Special Session”
upstream. (Correction: 2024-06-20, also on the old shortlist, is not special —
the held
M062024.pdfis a regular Thursday session; its agenda is genuinely absent upstream, the county never posted anA062024.) - Agenda present, minutes never captured → genuine upstream absence (county never posted minutes):
2024-09-25 (Wed between regular Tue/Thu sessions), 2024-12-30 (year-end special
session), 2025-12-30 (the county page’s “December 30, 2025” minutes link is
mislabeled — its href points to the Dec 23 file
M122325.pdf; no realM123025.pdfexists upstream).
No corpus action required for any of the 8.
Full civic cutover (2026-06-12, #133 follow-on): the commissioners’ entire meeting
record (Jan 2023–, 991 files) is now connector-sourced under
data/documents/commissioners/meetings/.
The legacy hand-assembled minutes/raw/ tree (930 PDFs) was retired after every
file was verified byte-identical to its connector copy — the per-file record is
cutover-reconciliation.yaml
(930/930 matched, 0 retained). The download manifest + meeting index sit alongside it.
OCR pass complete (2026-06-12, #135): all 991 files text-extracted, 969/991 dates
content-verified and 270 meetings flagged for corridor topics (up from 497/91 when only
the agendas had a text layer). The “934 files in minutes/raw/” figures above describe the
pre-cutover 2026-04-17 snapshot.
Cross-refs
bosc-prr-production-2026-06-05.response-index.yaml— the production tracker §1 is read frombosc-prr-production-2026-06-05.analysis.md·../../../docs/legal/mandamus-analysis.mdallen-county-web-vendor-audit.md·allen-county-level-sites.md— item-16 custody../oepa/(Shawnee II permit2PK00002) ·../../reference/echo/(Maumee discharger inventory) — Category-B wastewater ties../../../docs/COMPUTE.md·../../reference/compute/README.md— the compute / AI-capacity derivation whose keystone (air permit P0138965) §4 tracks