Services and databases

MySQL, Redis, Elasticsearch and the rest come from a catalogue, are installed as instances, and are switched on per project. Three acts, three places.

The catalogue: services installed as instances, each with its own version and port.

1. Install from the catalogue

Catalogue → Available lists what a source publishes, with versions. Install puts the package on disk. Nothing runs yet.

StackVo carries no services inside itself; until a source is given nothing is available. The catalogue is signed — the registry with minisign, each package with a checksum — and an unsigned source is named as such.

2. Add an instance

Catalogue → Service instances → Add starts that version in this workspace. Each instance has its own data and its own port.

The moment to set the root password is now: an image reads it only when it first initialises an empty data directory. After that, the password is what it is.

Two versions of one service run side by side — MySQL 8.0 and 8.4 — and each keeps its own data. Make primary sets the default for projects that name no instance.

3. Switch it on for a project

Project → Services lists what the project declares and what this machine runs:

State	What it means
On here	Running, and the project can reach it.
Not on here	The project wants it and this machine does not run it. Enable starts it.
Suggested	A guess read from the project's own .env — DB_CONNECTION=pgsql, say. Never written on its own; Write to stackvo.json turns it into a declaration you commit.

The connection string sits in the same pane. The password appears on a click.

Two versions at once

Project A on MySQL 8.0, project B on 8.4, both running: services are instances, not one global copy. Each project connects to the one it asked for.

Removing

Remove on an instance deletes its data with it. Take a snapshot first if you want to keep it.