Technology May 04, 2026 · 1 min read

Shipping TanStack Start and Bun to Railway

Railway's Nixpacks autobuild detects Bun projects fine, but it can't sequence the combination this site needs: prisma generate at build time, Vite + the TanStack Start plugin, a custom server.ts entry, and prisma migrate deploy on boot. A four-stage Dockerfile is simpler than teaching Nixpacks all o...

DE
DEV Community
by Andreas Bergström
Shipping TanStack Start and Bun to Railway

Railway's Nixpacks autobuild detects Bun projects fine, but it can't sequence the combination this site needs: prisma generate at build time, Vite + the TanStack Start plugin, a custom server.ts entry, and prisma migrate deploy on boot. A four-stage Dockerfile is simpler than teaching Nixpacks all of that.

The full post walks through the recipe: two parallel bun install stages (one full, one production-only), a build stage that runs prisma generate against a dummy DATABASE_URL, a lean runtime that layers the generated Prisma client on top of deps-prod's node_modules, and an entrypoint that runs migrations before exec-ing into Bun so the container's PID 1 shuts down cleanly on Railway's SIGTERM.

Plus the small Railway-side bits: how to wire the Postgres reference variable, why you should bind explicitly to 0.0.0.0, and the COPY-order detail that determines whether your runtime sees the right Prisma client.

Originally published at andreasbergstrom.dev — read the full post there.

DE
Source

This article was originally published by DEV Community and written by Andreas Bergström.

Read original article on DEV Community
Back to Discover

Reading List