Scrape TACC Events — tacc.org/tacc/events

[linear]

Description

Build a scraper for the Texas Advanced Computing Center events page at tacc.org/tacc/events. Lower volume than the other sources but valuable for research-oriented events, workshops, and training sessions.

Stack

• Language: TypeScript
• Runtime: Cloudflare Workers (scheduled cron trigger)
• Storage: Cloudflare D1
• Client surface: React Native app

Hi-Fi UI Requirements (what the design demands from the scraper)

Events from TACC render in the same flyer-first card system as all other sources. Many TACC events are multi-day workshops and often have no flyer, so the no-flyer fallback variant matters here.

• Flyer image is the hero when available. Cards support vertical, square, horizontal, and no-flyer variants. We need:
  • image_url at highest resolution available (prefer original upload over thumbnail).
  • image_width, image_height, and image_aspect_ratio (vertical / square / horizontal / none).
  • image_mime_type and image_alt_text when present.
  • Expect a meaningful share of TACC events to resolve to image_aspect_ratio = "none" — the no-flyer card must still render cleanly with the data below.
• "Posted by [Org]" is shown. host_organization defaults to "TACC" but should capture the sub-program (e.g., "Frontera", "Stampede3", "TACC Institute") when listed. Populate host_organization_slug for routing.
• Card shows date + short time. Store start_datetime / end_datetime as ISO 8601 (America/Chicago). Multi-day workshops must preserve full start/end — the client will render a range (e.g., Mon 3/2 – Fri 3/6).
• Card shows a short location string (e.g., ACES 2.302 or Virtual). Store location_short (≤ 40 chars) and location_full.
• Interest grouping. TACC events map to research-oriented interests (training, symposium, workshop). Capture categories accurately so they flow into the interests system.

Scope

• Crawl the TACC events listing and paginate through all upcoming events.
• Extract the following fields per event:
  • title
  • description
  • start_datetime / end_datetime (ISO 8601, America/Chicago — handle multi-day workshops)
  • location_short, location_full
  • host_organization (default: "TACC"; capture sub-program if listed), host_organization_slug
  • event_url
  • image_url, image_width, image_height, image_aspect_ratio, image_mime_type, image_alt_text
  • categories (e.g., training, symposium, workshop)
  • registration_url if separate
  • source = "tacc"
• Map into the shared D1 events schema.
• Deduplicate via a stable source_event_id.
• Upsert into D1.

Deliverables

• scrapers/tacc.ts worker module exporting a run() entrypoint.
• Unit tests with saved fixtures covering: single-day event, multi-day workshop, virtual event, event with no flyer (exercise no-flyer card variant).
• Dry-run script.
• Per-event error isolation.

Acceptance Criteria

• ≥ 95% of listed TACC events are captured per run.
• Multi-day events are stored with correct start/end datetimes and render as a date range on the client.
• image_aspect_ratio correctly classified (including "none") for ≥ 95% of events.
• location_short renders in ≤ 40 chars.
• No duplicates across repeat runs.
• D1 row count for source = "tacc" matches site listings (±5%).
• CI passes lint, typecheck, and tests.

Out of Scope

• UI work in the React Native app.
• Integration with TACC account / allocation data.
• Cross-source deduplication.

[linear]

LOOP-146