Sparkbase

Open-source, self-hosted backend services. Create a project, add only the services your app needs, and use them from one SDK with a project URL and key, or natively with a connection string.

Service	What you get
Postgres	A database with its own role. Native postgres:// URL plus SQL over HTTP
Cache	A Redis namespace with its own ACL user. Native redis:// plus a key/value API
Storage	A private bucket. Upload, download, list and signed URLs
Vector	Collections of embeddings with similarity search (pgvector, with a fallback)
Queues	Durable queues with visibility timeouts, retries and dead letters
Auth	Email and password sign-in for your app's users, with JWT sessions
Realtime	Broadcast channels and live table changes over Server-Sent Events
Cron	Scheduled jobs that call a URL, run SQL or send a queue message
Backups	Scheduled database dumps to this server or any S3 bucket
Monitor	Uptime checks with alerts and a webhook

Databases and caches run in their own containers with a memory limit and scale to zero when idle; the next connection wakes them in under a second. Every service has live logs and metrics, and databases and caches have a terminal.

import { createClient } from '@sparkbase/sdk'

const sb = createClient(process.env.SPARKBASE_URL, process.env.SPARKBASE_SECRET_KEY)
const { rows } = await sb.postgres('main').query('select * from todos')
await sb.queue('jobs').queue('emails').send({ to: 'ada@example.com' })

No paid tier, no hosted edition, no feature gating.

Install on a server

git clone https://github.com/shellhaki/sparkbase && cd sparkbase/deploy
cp .env.example .env && ./generate-secrets.sh   # set SPARKBASE_DOMAIN, TLS_MODE, ACME_EMAIL
docker compose up -d

Needs Docker: each database and cache runs as its own container. Open https://app.<your domain>/setup: the first account is the admin. Full guide: docs/self-hosting.md, and all docs at /docs in the dashboard.

Run it locally

Needs Go, Bun, Postgres 17+ and Redis installed (for example brew install go bun postgresql@18 redis). No Docker: local development uses shared hosts instead of containers.

make dev        # backend on :8787 + dashboard on :3000, Ctrl+C stops both

Or separately, in two terminals:

make backend    # API + worker, starts private Postgres and Redis
make frontend   # dashboard

make stop stops the private Postgres and Redis.

First run: open http://localhost:3000/setup and create your account. The first account is the admin; everyone else joins by invite.

Repository

Path	What
server/	Go: API, worker, container engine and drivers (cmd/sparkbase-server) and the CLI (cmd/sparkbase)
apps/dashboard/	Next.js dashboard on Bun, including the /docs site
apps/landing/	Public landing page and docs, deployable to Vercel
packages/sdk-js/	@sparkbase/sdk
scripts/	Dev services, dev runner, test helpers
deploy/	Docker install: compose, Traefik, Postgres host, Garage
docs/	Documentation (Markdown plus nav.json)
examples/	Example apps
DECISIONS.md	Decision log
waitlist/	Landing page (separate app)

Tests

scripts/dev-services.sh start
cd server && make test                                   # Go unit + integration, incl. isolation suites
cd packages/sdk-js && ../../scripts/with-server.sh bun test       # SDK against a live server
scripts/with-server.sh scripts/e2e-dashboard.sh          # dashboard smoke test
scripts/e2e-compose.sh                                   # full Docker install through Traefik

License

Server, drivers and dashboard: AGPL-3.0. SDK and CLI: MIT. See CONTRIBUTING.md.