QA rubric

Apply these checks field by field against the scraper's JSON output. Each check has a pass/fail criteria and links to the relevant schema section.

Rubric report

Paste a spider's JSON output below to run the mechanical half of the rubric - required fields, datetime format, vocabularies, link order, duplicates. The checks reuse the same generated schema and duplicate detection the app uses.

Paste scraper output (JSON array)

title

Check: Official meeting name including the body name; no embedded date or time.
Pass: "City Council Regular Meeting".
Fail: "Regular Meeting" or "Board Meeting - January 15, 2026 at 6:00 PM".

description

Check: Present; usually empty unless the source provides one.
Pass: "" or a real source-provided description.
Fail: Missing field, or a synthesized copy of the title.

classification

Check: One of the city-scrapers-core classification constants - Advisory Committee, Board, City Council, Commission, Committee, Forum, Police Beat, Not classified.
Pass: "City Council" for a council meeting.
Fail: An invented type outside the vocabulary.

start

Check: Naive datetime string "YYYY-MM-DD HH:mm:ss" - no timezone suffix.
Pass: "2026-01-15 10:00:00".
Fail: "2026-01-15T10:00:00-08:00" (timezone-aware) or a missing field.

end

Check: Naive datetime, and after start when present.
Pass: "2026-01-15 12:00:00" (two hours after start).
Fail: Before start, or carrying a timezone suffix.

If end is not explicitly set by the spider, it defaults to start + 2 hours. Usually correct - but check it against the source, some meetings have different durations.

status

Check: One of passed, tentative, confirmed, cancelled, and consistent with start.
Pass: passed for past meetings, tentative or confirmed for upcoming ones, cancelled only when the source confirms cancellation.
Fail: passed for a future meeting, or a value outside the vocabulary.

location

Check: Not empty unless time_notes explains why.
Pass: { "name": "City Hall", "address": "202 C St, San Diego, CA" }.
Fail: Empty name and address with an empty time_notes.

all_day

Check: Correct for the meeting type.
Pass: true for all-day events with midnight start/end times.
Fail: true for a meeting with a specific start time.

duplicates

Check: No records with the same start time and a similar title.
Pass: All records are distinct.
Fail: Two records with the same start and titles like "City Council Meeting" and "City Council Regular Meeting" - the viewer's duplicate detection uses significant-word overlap, not exact match.

The viewer's duplicate detection lives in lib/duplicate-detection.ts. It groups records by start time and then checks significant-word overlap in titles - generic words like "the", "meeting", "regular" are ignored. The rubric report above runs the same algorithm.

QA Rubric