Databases

Provision a managed engine in one command. Credentials are generated and injected automatically — you never plumb a connection string by hand.

Provisioning

conway add postgres        # → DATABASE_URL (+ PGHOST/PGPORT/PGUSER/PGPASSWORD/PGDATABASE)
conway add redis           # → REDIS_URL
conway add mongo           # → MONGO_URL  (FerretDB: Mongo wire protocol over Postgres)

Engine	Version	Notes	Injected var
Postgres	16	TLS on (sslmode=require)	DATABASE_URL
Redis	7	password auth, AOF persistence	REDIS_URL
Mongo	FerretDB 1.x	Mongo-compatible wire protocol	MONGO_URL

A custom --name foo injects FOO_URL instead, so you can run several of the same engine.

Adding a datastore triggers a rolling restart of running services so they pick up the new variable; the response lists restarted_services. conway db info <name> reports provisioning → ready.

Isolation

Datastores are ClusterIP-only: reachable from your project's services, never from the public internet. There's no exposed port and no way to accidentally leave one open.

Browse & query

Inspect data without leaving the platform:

# postgres — any SQL, results capped at 500 rows
conway db query postgres "SELECT * FROM users LIMIT 20"

# redis — a single command
conway db query redis "HGETALL session:abc"

The dashboard exposes the same thing as a Data tab on each datastore: a table/key browser plus a query console. Every query is recorded in the audit log.

Backups & restore

Each datastore is backed up nightly to object storage with 7-day retention. Trigger and restore on demand:

conway db backup postgres                       # back up now
conway db backups postgres                       # list backups
conway db restore postgres --backup <key>        # restore (postgres round-trips fully)

WARNING

In v1, Postgres restores round-trip completely (pg_dump --clean). Redis and Mongo backups are taken and stored, but live-restore is not yet wired.