Deploying a Generated App

Generated projects ship with a Procfile at the root and a Dockerfile in each app package. There is no opinion baked in about where you deploy — the artifacts are standard.

Local development

Run an app directly inside the workspace:

bun run --filter '@pastebin/api' start
bun run --filter '@pastebin/ui' start

Both pick up the configured storage/auth backends from environment variables (see Extensions).

Docker

Each app package has its own multi-stage Dockerfile. From the project root:

docker build -f apps/api/Dockerfile -t pastebin-api .
docker run --rm -p 3000:3000 -e PASTEBIN_STORAGE=memory pastebin-api

The build context is the project root because the multi-stage build needs the workspace lockfile. The image is oven/bun:1-alpine based; expect ~80MB compressed.

For SQLite/Redis-backed deployments, set the relevant env var (e.g. PASTEBIN_STORAGE=redis, PASTEBIN_REDIS_URL=...) — see Extensions for the full list.

Procfile-based platforms (Fly.io, Railway, Render, Heroku)

The generated Procfile declares one process per app:

web: bun run --filter '@pastebin/api' start
ui: bun run --filter '@pastebin/ui' start

Most Procfile-aware platforms will pick this up automatically. For Fly.io specifically:

cd pastebin
fly launch                            # one-time, generates fly.toml
fly secrets set PASTEBIN_STORAGE=sqlite
fly deploy

For Railway: railway up from the project root, then set env vars in the dashboard.

Configuration

Every backend selection is an env var. Conventions:

  • <APP>_STORAGE — memory, jsonfile, sqlite, redis, or eventsourced
  • <APP>_AUTH — none, jwt, session, apikey, password
  • <APP>_EVENT_STORE — memory, jsonfile, redis
  • Backend-specific settings: <APP>_REDIS_URL, <APP>_SQLITE_PATH, etc.

<APP> is the schema’s domain name uppercased and underscored. Run with no env vars set and the app prints the env vars it expected.

What isn’t generated

  • A reverse proxy / TLS terminator — front the API with whatever your platform provides.
  • Database migrations beyond schema.json consumption — for SQLite the runtime creates tables on first run; for relational migrations beyond that you’re on your own.
  • Observability (metrics, traces) — logs go to stdout (Effect logger); wire them through your platform’s log pipeline.