Handbook snapshot: 2020–2026 (May 2026)

Full handbook corpus exported from production Postgres, covering seven academic years.

Contents (tar -xzf to repo root → data/):

• data/raw/2020/{units,courses,aos}/*.json
• data/raw/2021/{units,courses,aos}/*.json
• data/raw/2022/{units,courses,aos}/*.json
• data/raw/2023/{units,courses,aos}/*.json
• data/raw/2024/{units,courses,aos}/*.json
• data/raw/2025/{units,courses,aos}/*.json
• data/raw/2026/{units,courses,aos}/*.json
• data/manifest.json — buildId, scrapedAt, source=pg-export, per-year counts

Counts

Year	Units	Courses	AoS
2020	5,365	337	400
2021	5,463	382	392
2022	5,251	445	375
2023	5,137	455	383
2024	5,084	495	392
2025	5,155	485	408
2026	5,218	501	410

Source: snapshot of the production database, not a fresh scrape. Each unit/course/aos file is the original CourseLoop payload preserved in the raw jsonb column; offerings, requisites, and enrolment rules live inside raw and are re-extracted by pnpm ingest.

Supersedes 2026-scrape.

Handbook snapshot: 2026 academic year

Full 2026 corpus scraped via pnpm scrape.

Contents (tar -xzf to data/):

• data/raw/2026/aos/*.json — 410 areas of study
• data/raw/2026/courses/*.json — 501 courses
• data/raw/2026/units/*.json — 5,217 units
• data/manifest.json — buildId, scrapedAt timestamp, counts, errors

Coverage: 6,128 / 6,135 (99.9%). Eight ErrorPage404 entries are genuinely unpublished codes; see manifest.json → errors[] for the exact list.