UC Entrepreneurship Hub

A unified front door for entrepreneurship programs across all ten University of California campuses — incubators, accelerators, courses, funding, competitions, and maker spaces in one searchable, comparable catalog.

Stack

Built on Vite+ — the unified open-source toolchain that bundles Vite, Vitest, Oxlint, Oxfmt, Rolldown, and tsdown into a single vp CLI.

• React 18 + TypeScript (strict)
• React Router (HashRouter, for static-host deep links)
• CSS variables from the official UC design tokens (Source Serif 4 + Source Sans 3, UC Blue / Dark Blue / Gold, square corners, Dark Blue tinted shadows)

Develop

Requires Node.js 24 (the LTS as of early 2026). The crawler scripts run TypeScript natively under node — node scripts/crawl/run.ts only works because Node ≥22 strip-types is the default and explicit .ts import extensions resolve out of the box. Use vp env use 24 (Vite+ ships Node version management) or your usual version manager.

# install Vite+ globally if you haven't
curl -fsSL https://vite.plus | bash

# install deps
vp install

# dev server
vp dev

# format + lint + type-check + duplication scan (fallow)
vp run check

# auto-fix formatting and lint issues
vp check --fix

# tests
vp test

# production build
vp build

# preview production build
vp preview

A Git pre-commit hook runs vp run check (format, lint, type-check, and fallow) before every commit, so issues are caught locally instead of in CI. It activates automatically on vp install / pnpm install — the prepare script points core.hooksPath at .githooks/. Bypass a single commit with git commit --no-verify.

Project layout

src/
├── App.tsx                    # router + compare provider
├── main.tsx                   # entry
├── styles/
│   ├── tokens.css             # UC design tokens (verbatim from brand guide)
│   └── global.css             # resets + body defaults
├── data/
│   ├── types.ts               # Campus, Program, ProgramType, Spotlight
│   ├── campuses.ts            # 10 UC campuses with map coords, colors, hubUrl
│   ├── types-list.ts          # program types, industries, stages, eligibility, durations
│   ├── normalize.ts           # slug + canonical type/industry/stage helpers, mergePrograms
│   ├── programs.ts            # curated programs + spotlight stories (merges crawled data)
│   └── programs.generated.ts  # AUTO: written by `vp run build-data`
├── lib/                       # icons, compare cart context, program gradient
├── components/                # Nav, Footer, Page, ProgramCard, CaliforniaMap, etc.
└── pages/                     # HomePage, DiscoverPage, ProgramDetail, CampusPage, etc.

scripts/crawl/
├── sites.json                 # campus seed URLs
├── run.ts                     # Playwright crawler (worker pool over campuses)
├── extract.ts                 # link discovery + in-page program extractor
├── build-data.ts              # data/crawled/*.json → src/data/programs.generated.ts
└── campuses/                  # per-campus override hooks (allowlist, denylist, name filters)

data/crawled/                  # one JSON file per campus, written by `vp run crawl`

Data pipeline

The catalog is a merge of curated programs (hand-shaped in src/data/programs.ts) and crawled programs (auto-extracted from each campus's innovation hub). On conflict, the curated record wins; the crawled record fills in optional fields the curated copy is missing (website, applicationLink, sourceUrl, lastUpdated, longDescription).

Refresh the catalog

# install Playwright's chromium once
vp install
vp exec playwright install chromium

# crawl every campus (≈ 5–10 min over a fast connection)
vp run crawl

# scope to a subset
vp run crawl -- --campus=berkeley,la --limit=10

# walk the seed but skip extraction
vp run crawl -- --campus=sd --dry-run

# regenerate src/data/programs.generated.ts from data/crawled/*.json
vp run build-data

# the combined shortcut
vp run refresh-data

Weekly automation: .github/workflows/weekly-crawl.yml runs refresh-data every Sunday and commits the regenerated module. Trigger it on demand from the Actions tab via Run workflow.

Program schema

Required fields stay stable for back-compat with the design prototype's curated cards:

interface Program {
  id: string; // also acts as the URL slug
  slug?: string; // explicit slug (defaults to id)
  name: string;
  campus: string; // CAMPUSES.id
  type: string; // canonical type (TYPES.id)
  desc: string; // card-length summary
  industries: string[]; // canonical industry labels
  stage: "Idea" | "Prototype" | "Pre-seed" | "Scaling";
  eligibility: string[];
  duration: string;
  funding: string;
  selectivity: string;
  cohortSize: number | null;
  deadline: string;

  // Optional, filled in by the crawler when available:
  longDescription?: string;
  deadlines?: { label?: string; date?: string }[];
  website?: string;
  applicationLink?: string;
  associatedCenter?: string;
  tags?: string[];
  lastUpdated?: string; // ISO 8601, set by the crawler
  sourceUrl?: string; // page the program was extracted from
  featured?: boolean;
  eyebrow?: string;
}

canonicalType / canonicalIndustries / canonicalStage in src/data/normalize.ts map noisy upstream phrasing ("startup accelerator", "AI/ML", "machine learning") onto the design's controlled vocabulary.

Adding a new campus

1. Append the campus to src/data/campuses.ts (id, color, map x/y, hubUrl).
2. Add the matching entry to scripts/crawl/sites.json.
3. Drop scripts/crawl/campuses/<id>.mjs re-exporting from _default.mjs. Override linkAllowlist, linkDenylist, allowName, or maxSubpages only when the defaults pull in junk.
4. Run vp run refresh-data.

If a campus's site is JS-heavy and the generic extractor returns nothing, the override module is the right place to bolt in a custom discover selector — keep that surgery campus-local.

Routes

Path	Page
/	Home
/discover	Program discovery (with ?q=, ?campus=, ?type=, etc.)
/program/:id	Program detail
/campus/:id	Campus page
/campuses	All campuses index
/compare	Comparison tool

Deployment

The site has two deploy targets:

Production — GitHub Pages

GitHub Actions workflow at .github/workflows/deploy.yml builds on every push to main and publishes to GitHub Pages. The workflow:

1. Installs Vite+ globally
2. Runs vp run check (format + lint + type-check + duplication scan)
3. Runs vp build with VITE_GH_PAGES=1 so assets resolve under /uc-entrepreneurship-hub/
4. Copies index.html to 404.html so hash-router deep links work
5. Uploads dist/ and deploys

To enable: in the GitHub repo settings, Settings → Pages → Source: GitHub Actions.

Per-PR previews — Cloudflare (Workers + Static Assets)

Cloudflare auto-deploys every branch push (including PRs) to its own preview URL. The static-asset deploy is configured by the wrangler.jsonc at the repo root — that file is the source of truth for how the Worker serves dist/. Connect the repo from the Cloudflare dashboard once and every PR gets a unique preview URL.

Build settings (Cloudflare dashboard → your project → Settings → Build):

Field	Value
Build command	pnpm run build
Deploy command	npx wrangler versions upload (default for Workers projects, leave as-is)
Root directory	(empty)

The Worker's name in wrangler.jsonc ("name": "uc-entrepreneurship-hub") must match your Cloudflare project name. Update either the file or the dashboard until they agree.

Environment variables (Settings → Environment variables, set on both Production + Preview):

Name	Value
NODE_VERSION	24
PNPM_CONFIG_OPTIONAL	false

PNPM_CONFIG_OPTIONAL=false skips the optionalDependencies block in package.json — playwright lives there because only the crawler needs it, and it's ~70MB of binary that the static build never touches. Skipping it cuts CF install time noticeably without affecting the build.

Notes:

• Don't set VITE_GH_PAGES=1 on Cloudflare — it's only for the GH Pages base path. CF Pages serves at /, which is vite.config.ts's default.
• The build command must be pnpm run build, not vp build directly. Cloudflare's build shell doesn't put node_modules/.bin on $PATH; going through the npm script does.
• The app uses HashRouter, so deep links work without a _redirects file or SPA fallback — every URL serves index.html and the fragment is read client-side.

PR preview URLs surface automatically as a comment on the PR (configurable in the CF dashboard).

Credits

• Design system: UC Brand Guide
• Design prototype: handed off from Claude Design
• Toolchain: Vite+