import { railway } from '@ruvyxa/adapter-railway'
import { config } from 'ruvyxa/config'

export default config({ adapter: railway() })

The build emits a standalone server at .ruvyxa/deploy/railway/server/index.mjs and a safe railway.json. Existing project configuration is never overwritten. The server honors Railway's PORT and supports SSR, SSG, CSR, ISR, PPR, API routes, and native realtime.

Client identity behind a proxy

The generated server believes X-Forwarded-For and X-Real-IP only from a connection whose peer is loopback or listed in security.trustedProxyIps. From any other peer both headers are dropped before the request is routed, because nothing in front of the server overwrote them and one caller rotating a value would otherwise collect a fresh bucket per request from the built-in rate middleware, the server-action rate limiter, and the action replay quota.

Behind a reverse proxy that is not on the same host — nginx or Traefik in another container, an ingress controller, a service mesh — list its address or range, or every visitor is counted as one client:

export default config({
  security: { trustedProxyIps: ['10.0.0.0/8'] },
})