# Vault offload (serve protected files from S3 or R2)

*Collection: Access Control & Protection · Category: Access Control · Availability: FileDeck Pro · Canonical URL: https://www.getfiledeck.com/docs/vault-offload/*

A heavy protected library makes your web server do all the serving: every
vaulted download streams through PHP. With offload on, vaulted files are
mirrored into your connected S3-compatible bucket — Amazon S3, Cloudflare
R2, Wasabi, DigitalOcean Spaces, MinIO — and downloads redirect to
**short-lived presigned URLs**. The bucket does the bandwidth; FileDeck
keeps doing the access control.

> **FileDeck Pro** — **Documents → Settings → Privacy & integrity → Vault
> offload**. Needs the S3 connection (Documents → Cloud storage) with
> credentials that can **write** (`s3:PutObject`) — browsing and attaching
> only ever needed read.

## What actually happens

- The gates come first, always. Access rules, terms, quotas, bot checks —
  everything runs on your server before the redirect is minted, exactly as
  when streaming. The presigned URL a visitor receives lasts minutes and
  names one object.
- Vaulted files are **mirrored** into the bucket on a daily cycle (capped
  per day so a large vault drips across days; failures are logged and
  retried). A file not yet mirrored simply streams locally — availability
  never depends on the bucket.
- **The local file stays.** Thumbnails, content search, checksums and
  watermarking all read local bytes. What offload moves off your server is
  the recurring cost — download bandwidth and PHP time — not the storage.
- **Watermarked serving stays local**, deliberately: stamping happens in
  the stream, and a redirect would silently hand out unstamped files.
  Stamped beats offloaded.
