saxi.ai

Broad public API directory with a strong AI and agent focus.

V1 is intentionally simple:

• static, build-time rendered pages
• Cloudflare Worker + static assets
• public APIs with availability metadata
• multiple public GitHub sources
• strong category and capability landing pages
• lightweight client-side search on top of a crawlable archive

Stack

• TypeScript for the Worker and build pipeline
• Tailwind CSS v4
• vanilla browser JavaScript for search/filter behavior
• Wrangler for Cloudflare deployment

Data sources

• public-api-lists/public-api-lists
• public-apis/public-apis
• tools-collection/apis-collection

The build reads versioned snapshots from data/vendor/ instead of fetching these sources live. Run npm run sources:refresh to update the snapshots from upstream.

Scripts

• npm run build builds the full static site into dist/
• npm run typecheck runs TypeScript checks
• npm run check runs type checks plus wrangler types --check
• npm run cf:typegen regenerates worker-configuration.d.ts
• npm run sources:refresh refreshes versioned API source snapshots in data/vendor/
• npm run dev builds the site, then starts Wrangler local development
• npm run audit:links audits all target docs URLs with 10 parallel workers and refreshes the ignore list
• npm run screenshots:capture captures up to 60 missing or stale real screenshots into .cache/screenshots using Playwright

Screenshots

V1 supports two modes:

• default: generated SVG placeholders for every API entry
• local/CI capture mode: npm run screenshots:capture renders real screenshots with Playwright into .cache/screenshots/
• optional remote mode: if SCREENSHOT_API_BASE_URL is provided, the build tries to fetch real screenshots and caches them in .cache/screenshots/

If remote screenshot capture is unavailable, the build does not fail.

Recommended production path:

• run npm run audit:links first, so dead targets land in data/api-ignore-list.json
• run npm run screenshots:capture -- --concurrency=5 --limit=60 --stale-days=90
• then run npm run build

The deploy workflow restores and saves .cache/screenshots, installs Chromium, refreshes a bounded batch of up to 60 missing screenshots plus screenshots older than 90 days, and then deploys.

Deploy

The repository includes GitHub Actions workflows for CI and Cloudflare deployment.

Required secrets for deploy:

• CLOUDFLARE_API_TOKEN
• CLOUDFLARE_ACCOUNT_ID

Optional screenshot-related secrets:

• SCREENSHOT_API_BASE_URL
• SCREENSHOT_API_TOKEN