OregonAI.github.io

The organization landing page for the Civic Corpus Platform, served at https://oregonai.github.io/.

This repository contains no corpus data. It renders one page describing the platform and the corpora, and it reads every corpus fact live rather than storing it.

How the corpus data gets here

Each corpus publishes corpus-index.json to its own GitHub Pages site as part of its deploy. build_site.py reads the header of that file for each corpus in its registry:

• index found → the corpus renders as Active with its live document count and contract version, and links to its site.
• no index → the corpus renders with the status declared in the registry (In progress / Planned) and no counts. The page degrades to the registry rather than showing a stale or invented number.

The build prints exactly which corpora were probed live and which fell back, so a corpus silently dropping off the page is visible in the Actions log rather than inferred from a missing card.

corpus-index.json is ~8 MB, and only three scalars at the front of it are needed, so the probe issues a Range: bytes=0-2047 request. A full fetch is the fallback for a host that ignores Range.

Adding a corpus

Add an entry to CORPORA in build_site.py. Only the descriptive fields are hand-maintained — repo, name, scope, archetype, a fallback status, an optional mcp URL, and the blurb. Counts and live/active state come from the corpus itself.

Building

python3 build_site.py            # probes each corpus, writes site/index.html
python3 build_site.py --offline  # registry-only render, no network (for local dev)

Stdlib only — no dependencies to install.

site/ is generated and gitignored; it is built in CI on every push to main, weekly (so a sibling corpus going live appears without a commit here), and on manual dispatch.

---

Unofficial and non-authoritative. Not affiliated with the State of Oregon. Every document in every corpus links to its authoritative source; always verify there.