Frankfurt + Hamburg culture monorepo

Six Cloudflare Workers that aggregate cultural programming for Frankfurt and Hamburg — plus Landau in der Pfalz — into edge-rendered apps. Same shape across all of them: one hourly GitHub Action (scrape.yml) runs the hub scrape into packages/event-hub, then per-app derive steps regenerate typed bundles (src/scrape-data.ts) committed to the repo in a single commit; Cloudflare's git integration redeploys each worker on push; workers render from the bundled data with no DB on the read path (D1 is used only for request-time user writes).

Five of the six are multi-city culture verticals (museums, theater, concerts, lectures, cinema), each a single Worker serving Frankfurt and Hamburg under <city>.<apex> subdomains. landau-today is single-city.

Cities

Cities live in one source of truth — packages/core/src/cities.ts (CITIES) — currently Frankfurt, Hamburg, Darmstadt and Heidelberg.

• Hosts — each city is served at <city>.<apex> (e.g. frankfurt.konzert.haus / hamburg.konzert.haus). The bare apex (konzert.haus) 302-redirects to the nearest city by Cloudflare edge geolocation (cityMiddleware({ apexBehavior: "geo" }) in packages/core/src/city-routing.ts). museumsufer.app is an SEO-primary alias host pinned to Frankfurt (resolved without a redirect).
• Switcher — the masthead CitySwitch (packages/core/src/cityswitch.tsx) lists only the cities a vertical actually has data for (supportedCities, derived per bundle), and degrades to a plain locality label when a vertical covers a single city.
• Event → city — declarative first (a scraper/orchestrator may set event.city), else geometric: a bbox pre-filter plus an optional precise polygon (cityFor / cityOf), so neighbouring cities can share a bbox yet split cleanly.

Hamburg coverage is rolling out vertical by vertical — live in museums, theater and lectures; concerts and cinema are wired for both cities and fill in as programming lands. The Darmstadt + Heidelberg cluster (upper-Rhine / Bergstraße) is wired across all five verticals — theaters, concerts, lectures, cinema and museums — via declarative-city routing plus reused ticketing adapters (Reservix, ztix, Kinoheld, Kinopolis, Bilettix/Eventim webshops) and bespoke venue/museum scrapers.

Roadmap. Hamburg proved the machinery is city-agnostic — the city model, geo-routing, per-city custom domains and the masthead switcher carry any number of cities, so a new one is mostly data + scrapers, not architecture. A backlog of city-expansion discovery tickets now maps every German city over 150k — Berlin, Munich, Cologne, the Rhine-Ruhr, Saxony, and the northern, central and south-western hubs — into per-vertical venue inventories (with proposed geofence + reuse notes) for staged, vertical-by-vertical rollout.

Apps

apps/museumsufer → museumsufer.app · {frankfurt,hamburg}.ins.museum

Daily exhibitions and events for ~40 Frankfurt Museumsufer museums, now extending to Hamburg (Hamburger Kunsthalle, Museum für Kunst und Gewerbe, Deichtorhallen, SHMH houses, …). Aggregates from museumsufer.de + per-museum APIs (15+ deterministic parsers in api-scrapers.ts: Tribe Events, TYPO3 calendarize, schema.org Event microdata, WP REST + ACF, RSS, Kirby CMS, …). DeepL EN/FR translation runs in the same scrape pipeline; the cache rides in the bundle. Image proxy with edge caching, distance sorting via RMV, fuzzy search, PWA.

• Scrape: hourly via the shared scrape.yml (museums derive; museumsufer.de is fetched directly and runs continue-on-error)
• D1: likes + push_subscriptions (request-time user writes)

apps/ins-theater → {frankfurt,hamburg}.ins.theater

Hourly performance schedule for Frankfurt theaters — Schauspiel, Oper, Mousonturm, English Theatre, Komödie, Tigerpalast, the Reservix-fronted small houses, plus a dozen long-tail venues — alongside Hamburg houses (Ernst Deutsch Theater, Alma Hoppes Lustspielhaus, Centralkomitee, …). Each theater has its own scraper module under src/scrapers/ (Reservix HTML, Tribe Events REST, MEC plugin, schema.org microdata, custom CMSes, …). Editorial Programmheft styling — Fraunces serif, JetBrains Mono numerals, single brick-red accent; the masthead carries a city-neutral stage-door mark.

• Scrape: hourly via the shared scrape.yml (theaters derive)
• D1: feedback + push_subscriptions

apps/konzert-haus → {frankfurt,hamburg}.konzert.haus

Hourly concert schedule for classical, jazz, sacred, world, experimental, and chamber music across Frankfurt and the Rhein-Main region — Alte Oper, Ensemble Modern, hr-Bigband, Holzhausenschlösschen, and more. Multi-city under the <city>.konzert.haus host scheme.

• Scrape: hourly via the shared scrape.yml (konzert-haus derive)
• D1: push_subscriptions

apps/lehrhaus → {frankfurt,hamburg}.lehr.salon

Daily index of public lectures, readings, and discussions — Polytechnische Gesellschaft, Haus am Dom, Jüdische Gemeinde, Literaturhaus, Bürgeruniversität, Institut für Sozialforschung, Evangelische Akademie, Sigmund-Freud-Institut, Denkbar, and more — with Hamburg coverage rolling in. Three formats (Vortrag / Lesung / Diskussion), with a rolling next-7-days view and cross-imports of Vortrag-class events from the museums and theaters apps. Editorial "annotated quarto" identity — foxed paper, iron-gall ink, rubric red, pilcrow anchors.

• Scrape: hourly via the shared scrape.yml (derives from packages/event-hub)
• D1: push_subscriptions (Web Push digest opt-ins)

apps/lichtspiel-haus → {frankfurt,hamburg}.lichtspiel.haus

Daily film-screening programme for the Frankfurt arthouse + repertory cinemas — DFF Deutsches Filminstitut, Astor, Cinéma / Eldorado / Harmonie, Pupille, Mal seh'n, Murnau Filmtheater, Caligari, Filmforum Höchst, plus the long-tail Rhein-Main houses; wired for Hamburg under the <city>.lichtspiel.haus scheme. TMDb-enriched posters and synopses (DeepL EN fallback), OMDb-backed Rotten Tomatoes + IMDb ratings with canonical deep links, mark-as-seen state across films, film-strip date slider in a Jugendstil / Saul-Bass register.

• Scrape: hourly via the shared scrape.yml (derives from packages/event-hub)
• D1: push_subscriptions

apps/landau-today → landau.today

Daily events for Landau in der Pfalz and the Südliche Weinstraße. Six public sources stitched into a single SSR page with URL-bound category + date filters. The only single-city app in the monorepo (apex landau.today + www.landau.today), shipping a different display font + linked stylesheet pipeline.

• Scrape: hourly via the shared scrape.yml
• D1: push_subscriptions

Packages

• packages/core — shared utilities: the city model + geo-routing + masthead switcher (cities.ts, city-routing.ts, cityswitch.tsx), hash, calendar URLs, German formatting, theme FOUC bootstrap, manifest/robots/api-catalog builders, security headers, UTM, scrape logging, bundle writer, null-last comparator, hreflang + locale detection, HtmlHead with preload / preconnect hooks
• packages/event-hub — central scrape orchestrator that fans out to the per-venue scrapers in packages/scrapers, classifies events via packages/classify, and runs the TMDb/OMDb/DeepL enrichment passes. Each app reads from the resulting EVENTS array via its own scripts/scrape.ts derive step.
• packages/scrapers — per-venue scraper modules (Reservix HTML, Tribe Events REST, schema.org microdata, WP REST + ACF, RSS, Kirby CMS, …) consumed by event-hub.
• packages/classify — label-based event classifier (film:cinema, music:classical, talk:lecture, …) driving which app picks up which event.
• packages/config — shared tsconfig and biome presets

Scrapers that need to bypass datacenter-IP blocks, broken TLS chains, or Cloudflare challenges route through an external fetch proxy (FETCH_PROXY_URL + FETCH_PROXY_TOKEN); apps/fetch-proxy is no longer an in-repo worker.

Stack

• Cloudflare Workers (TypeScript)
• Hono v4 + JSX SSR; htmx for the partial-swap routes
• Tailwind v4 (museums) / hand-rolled lightningcss (everyone else)
• Bun for tooling — installs, scripts, the scrape pipeline (bun:sqlite-free, pure-function)
• Turborepo workspaces
• GitHub Actions: hourly hub scrape (scrape.yml), daily scraper-health audit that files under-delivering scrapers as a tracking issue (scraper-audit.yml), nightly Lighthouse + SEO budget enforcement (lighthouse.yml), daily manifest-screenshot + OG-image regen against prod (regen-assets.yml)
• Cloudflare git integration for deploys

Common commands

mise use -g bun@latest                                  # one-time
bun install                                             # from repo root
bun run dev                                             # all apps
bun run typecheck
bun run lint
bun scripts/regen-screenshots.ts --prod                 # manifest screenshots (per-city variants)
bun scripts/regen-og-images.ts                          # OG raster from public/og-image.svg
bun scripts/regen-icons.ts                              # PWA icon raster from public/favicon.svg
gh workflow run scrape.yml                              # trigger hub scrape + per-app derives
gh workflow run regen-assets.yml                        # screenshots + OG
gh workflow run lighthouse.yml                          # CWV / a11y / SEO budgets
gh workflow run scraper-audit.yml                       # audit committed bundle for under-delivering scrapers

Per-app docs live next to the app (apps/<slug>/README.md).