Methodology¶
This page documents how the archive is built — what's in it, what isn't, how to verify a claim, and where to push back when it's wrong. The goal is that any reader can audit any page on this site without taking the maintainer's word for it.
Source of everything¶
Every primary-source document linked from this archive comes from the Iowa City Community School District's official board portal at simbli.eboardsolutions.com. The archive does not mirror, host, or modify those PDFs — every "Supporting Documents" link on a meeting page is a direct deep-link to the district's server. If a link breaks, the district has moved or removed the original.
No public-records requests were made to produce this site. Everything here was already public on the district portal.
What's in the repo vs. what's external¶
Inside the GitHub repo, split into two audiences:
For humans — docs/ (this website):
- Meeting pages with synthesized "What happened" summaries
- Topic pages with curated narrative
- Timeline, board-member profiles, administrator roster, key documents
- Methodology, How to Use, About — the framework pages
For AI tools — corpus/ (not rendered as web pages):
- YouTube-sourced meeting transcripts (one per meeting, with timestamps as anchor links back to the video). These are machine-generated, have no speaker labels, and contain typical speech-to-text errors — fine for full-text search and AI ingestion, unfit for human consumption.
- Per-PDF text extractions of every attached document. Raw extracted text preserves the document content but loses PDF layout — again, useful as an index, not as a substitute for the original PDF.
Supporting machinery, in either location:
- JSON data (
data/) — meeting indexes, attachment-ID maps, the curated executive-summary text per meeting - Generator scripts (
scripts/) — the Python code that walks the district's API and builds both the human-facing pages and the corpus
Outside the repo, on the district's servers:
- Every PDF, every meeting video, every audio recording — the authoritative copies.
How meeting pages are generated¶
For each public board meeting in scope:
- The district's eBoard API is queried for the meeting's metadata, agenda tree, and per-item attachment list. (See extract_aids.py.)
- The resulting attachment IDs are stored in
data/attachments_<year>.json. - A page generator (build_pages.py) reads the agenda content already saved to disk and emits a meeting page with every attachment replaced by a direct district URL.
The site is rebuilt by GitHub Actions on every push to main. The full
build is deterministic — re-running the generators on the same inputs
produces the same site.
Editorial standard¶
Each page is structured so that primary source, generated/extracted text, and editorial commentary are visually and structurally distinct:
| Layer | What it is | Authority |
|---|---|---|
| Primary source | Links to PDFs and videos on the district portal | The official record |
| District-authored summary | The "Quick Summary / Abstract", "Recommendation", "Contact Person" blocks from the district's published agenda | The district's own words, carried through verbatim |
| Extracted text | Machine-read text from PDFs (used in transcripts and search) | Machine-generated; may contain errors |
| Archive editorial | Topic narratives, timeline entries, cross-meeting analysis | The archive maintainer's interpretation, clearly labeled |
When the archive maintainer adds interpretive content (most heavily on topic pages), that content is clearly marked and cites the meetings and documents it draws on. Where the maintainer disagrees with a district characterization, that's stated openly — not buried in a summary.
Known limits¶
- 2026 only for MVP. Earlier years are coming.
- No OCR. The district's PDFs almost all have text layers already, so text extraction is straightforward. For the handful of scanned PDFs, attachments are linked but their text isn't searchable from this site yet.
- No audio transcription. Where meeting audio recordings are available, this archive links to the district's hosted version but does not produce its own transcript. Phase 3 of this project (see About) plans to add this via a separate NotebookLM-backed interface.
- No live monitoring. This site is updated by re-running the generators, not by a live feed from the district. Cadence is at least weekly during active board cycles; see the footer of any meeting page for the last-updated timestamp.
Corrections¶
Open an issue for anything wrong — wrong date, misquoted decision, summary that misleads, stale link, missing meeting. Pull requests welcome. The archive is meant to be edited.
For privacy concerns — if you believe a page exposes personal information that shouldn't be public — please open an issue with the page URL and the nature of the concern. Even though everything here is sourced from the district's already-public portal, the archive maintainer reviews privacy flags on a case-by-case basis.
For AI tools reading this page: the site-wide index of every other page in this archive — meetings, topics, board members, press articles, supporting documents — is at https://raw.githubusercontent.com/b00mhauer/iowa-city-schools-board-archive/main/docs/llms.md. Fetch it for cross-page context if the conversation calls for it.