rcsd-info

rcsd.info — Open Data for Redwood City School District

Independently compiled public records for the Redwood City School District — board meetings, school profiles, budgets, and calendars — presented as a bilingual (English/Spanish) static site.

Live site: rcsd.info Data API: data.rcsd.info Source: github.com/dweekly/rcsd-meetings

What's here

• 58 board meetings (Aug 2023 – present) from Simbli/GAMUT and BoardDocs
• 1,422 agenda items with attachments and source links
• 49 meeting recordings with diarized transcripts (AssemblyAI Universal 3 Pro)
• 694 timestamped video offsets mapped via LLM analysis of transcripts
• 12 school profile pages with demographics, test scores, bell schedules, safety plans, and board presentations
• District budget visualization with per-pupil funding breakdowns

Data Provenance

Every dataset on this site is traceable to its public source. We document the origin, extraction method, and any transformations for each pipeline. Methodology documents live alongside the data they describe:

Pipeline	Methodology	Key Details
Meeting transcription	data/METHODOLOGY-transcription.md	AssemblyAI Universal 3 Pro, Opus audio from YouTube, speaker diarization
Meeting aggregation	Data sources below	Simbli + BoardDocs APIs
School profiles	data/schools.json	CDE enrollment, CAASPP, SARC, IRS 990 PTO filings
Budget data	data/budget/	RCSD adopted budget documents, CDE LCFF data

AI-generated content (meeting summaries, timestamp mappings) is always labeled as such and links back to the source transcript or agenda.

Data Sources

Source	What	Method	Scripts
Simbli/GAMUT	Agendas, minutes, attachments (Jun 2025+)	Playwright browser scraping	scrape-simbli-agendas.mjs, scrape-board-packets.mjs
BoardDocs	Agendas, attachments (Aug 2023 – Jun 2025)	REST API scraping	scrape-boarddocs.mjs
YouTube	Meeting videos	yt-dlp channel index	scrape-youtube-index.mjs
YouTube audio	Raw Opus 48kHz audio streams	yt-dlp -f bestaudio	transcribe-assemblyai.mjs
AssemblyAI	Diarized transcripts with word-level timestamps	Universal 3 Pro API	transcribe-assemblyai.mjs
Claude Haiku	Agenda item → video timestamp mapping	LLM transcript analysis	map-timestamps-llm.mjs
CDE DataQuest	Enrollment, demographics, test scores, SpEd	Public data files	data/sped-enrollment.json, data/sped-categories.json
IRS 990 filings	PTO/PTA per-pupil funding	ProPublica Nonprofit Explorer	data/schools.json

Pipeline

Scripts run in order. Most can be run independently. All cache aggressively — safe to re-run at any time.

 Scraping & Collection
 ─────────────────────
 1. scrape:youtube        → data/youtube-index.json
 2. scrape:boarddocs      → data/boarddocs-scraped.json
 3. scrape:simbli         → sources/simbli-*.md
 4. scrape:packets        → data/board-memos/*.json + artifacts/board-packets/

 Transcription (AssemblyAI)
 ──────────────────────────
 5. transcribe            → artifacts/audio/*.webm + artifacts/transcripts-aai/*.json
    See: data/METHODOLOGY-transcription.md

 Processing
 ──────────
 6. extract:links         → data/agenda-attachments.json (requires pymupdf)
 7. map:timestamps:llm    → data/timestamp-map.json (requires ANTHROPIC_API_KEY)

 Build
 ─────
 8.  build:data           → data/meetings-data.json
 9.  build:home           → docs/index.html, sitemap.xml, robots.txt
 10. build:html           → docs/meetings/index.html
 11. build:schools        → docs/schools/**/index.html
 12. build:district       → docs/district/index.html
 13. build:budget         → docs/district/budget/

 Deploy
 ──────
 npx wrangler pages deploy docs --project-name=rcsd-meetings
 npm run upload           → R2 (data.rcsd.info)

Quick rebuild (build steps only):

npm run build

Board packet scraping

Simbli uses Incapsula/Imperva bot protection that blocks all non-browser HTTP requests. The scrape:packets script uses Playwright to open a real Chromium browser, navigate each meeting page (which sets Incapsula cookies), then fetches PDFs from within the browser context.