chrome_s5uCyT1Vfj.mp4
A model explorer website for AI gateways. Auto-fetches the latest models from Vercel AI Gateway and lets you browse, compare, and filter them.
This monorepo contains:
apps/website: a Next.js model explorer that fetches models at runtime from supported AI gateways (currently Vercel AI Gateway), with caching and a build-time fallback snapshot
Prerequisites:
- Node.js >= 22
- pnpm (repo uses
packageManager: pnpm@10.x)
Install dependencies at the root:
pnpm installCommon scripts (powered by Turbo):
- develop everything:
pnpm dev - build everything:
pnpm build - typecheck:
pnpm check-types
Run the website:
pnpm --filter website devThe website fetches models on the server using unstable_cache (1-hour TTL, tag-based revalidation). The active gateway is resolved via a registry abstraction in apps/website/lib/ai/gateways/, currently providing Vercel AI Gateway. Adding more gateways is just a matter of implementing the GatewayProvider interface and registering it.
If the live fetch fails (e.g. during build with no network), a generated snapshot at apps/website/lib/ai/models.generated.ts is used as a fallback. Refresh it with:
pnpm fetch:modelsThe Vercel gateway requires an AI_GATEWAY_API_KEY (or VERCEL_OIDC_TOKEN) at runtime/build time when fetching live data.
MIT. See LICENSE for details.