Open-source object storage, on the Drive you already have

Use Google Drive
as object storage.

Store your app's photos, documents and backups in Google Drive instead of S3. It speaks the S3 API, so your existing @aws-sdk/client-s3 code works by changing endpoint.

View on GitHub Read the docs ↓

Set up drivebucket

Get a refresh token, check the setup and create your first bucket:

Terminal

$ npx @drivebucket/cli auth login --client-id ID --client-secret SECRET > .env
$ npx @drivebucket/cli --env-file .env doctor
$ npx @drivebucket/cli --env-file .env buckets create photos

Before you run it

In Google Cloud Console, enable the Google Drive API.
Create an OAuth client of type Desktop app.
Publish the consent screen. In Testing mode, refresh tokens expire after 7 days.

Full walkthrough: Quick start in the README →

GitHub repository →

Source, issues and releases. MIT licensed.

Documentation →

Quickstarts for Express, Next.js and Docker, plus S3 migration.

npm packages →

Four packages, published with provenance.

How it works

drivebucket maps the object-storage model onto Drive. Buckets become folders, objects become files, and your keys live in private file properties that only your app can see.

Bucket = folder

Each bucket is a folder under one root: a folder the app creates in My Drive, or a shared drive on Workspace.
Object = file

Files sit flat in the bucket folder. The key, bucket and a SHA-256 lookup hash are stored in appProperties, and a key → file-id cache (memory, Redis or MongoDB) skips the search.
Served by you, not Drive

Uploads are MD5-checked end to end. Reads stream through your server with byte ranges and resizing. Files are never shared publicly in Drive.

How a photo travels through drivebucket, in 54 seconds. Turn the sound on.

Four packages

Use the adapter directly, put the S3-compatible server in front of it, or drop the Next.js route into your app.

@drivebucket/core

The DriveAdapter: buckets, put/get/head/list/copy/delete, resumable uploads, retries with backoff and a pluggable key index.

npm Source

@drivebucket/server

Docker image with an S3-compatible API (SigV4, presigned URLs, multipart), a public read proxy, image resizing and Prometheus metrics.

npm Source Docker image

@drivebucket/next

Next.js App Router route for uploads and serving, plus browser-side resize and upload with progress.

npm Source

@drivebucket/cli

auth login with loopback OAuth, doctor to check your setup, and buckets, ls, cp, rm.

npm Source

Pick your integration

Three ways in. Each one has a full guide in the docs.

upload.ts

import { S3Client, PutObjectCommand } from '@aws-sdk/client-s3';

const s3 = new S3Client({
  endpoint: 'https://files.example.com', // your drivebucket-server
  forcePathStyle: true,
  region: 'us-east-1',
  credentials,
});

await s3.send(new PutObjectCommand({
  Bucket: 'photos', Key: 'cat.jpg', Body, ContentType: 'image/jpeg',
}));

Keep your S3 code

Point the AWS SDK at drivebucket-server. Signatures, presigned URLs and multipart uploads work as they do on S3.

It's a subset: path-style only, with no versioning, ACLs or object lock.

Migrating from S3 →

app/api/files/[[...key]]/route.ts

import { DriveAdapter, driveOptionsFromEnv } from '@drivebucket/core';
import { createDriveRoute } from '@drivebucket/next';

export const runtime = 'nodejs';

export const { GET, HEAD, POST } = createDriveRoute({
  adapter: () => new DriveAdapter(driveOptionsFromEnv()),
  bucket: 'photos',
  maxBytes: 12 * 1024 * 1024,
  allowedTypes: ['image/*'],
});

Upload from the browser

uploadFile() from @drivebucket/next/client downscales images to WebP before upload and returns the key to save in your database.

File types are checked from the bytes, so an HTML file can't pose as image/png.

Next.js quickstart →

Terminal

# server + Redis key index
$ cp docker-compose.example.yml docker-compose.yml
$ docker compose up -d
$ curl localhost:8080/-/ready

# resized, cacheable image URLs
https://files.example.com/photos/cat.jpg?w=640&format=webp

Put a CDN in front

Every uncached read is a Drive download costing 200 quota units; a cached one costs nothing. Cloudflare in front of the proxy keeps reads fast and cheap.

Docker + Cloudflare guide →

The limits, up front

Drive is not S3. drivebucket works well for an app with thousands of users and a CDN in front. Read these before you rely on it.

Limit	Value	What it means
Upload volume	750 GB / user / day	After that, uploads and copies fail for 24 hours.
API quota	325,000 units / min	An uncached download costs 200 units: about 27 per second per user.
Objects per bucket	~500,000	Drive's per-folder cap. Shard across buckets beyond that.
Listing	Scans the bucket	Drive can't filter by prefix. Fine for thousands of objects, costly for hundreds of thousands.
Consistency	No atomic conditional writes	`If-Match` is best effort; concurrent writers can race.
Object size	5 TB	Uploads above 5 MB use resumable sessions.