Custom Docker images built and published to ghcr.io/yoyozbi via GitHub Actions.

Image: ghcr.io/yoyozbi/images/postgres-backup

Based on postgres:17-alpine, this image bundles everything needed to back up a PostgreSQL 17 database: the matching pg_dump/pg_restore binaries (same major version as the target server), curl for shipping dumps to remote storage, and jq for processing JSON responses from backup APIs.

Using the same major version as the target server avoids pg_dump compatibility warnings and ensures the dump format is fully supported.

Image: ghcr.io/yoyozbi/images/super-sync

A self-hostable sync server for Super Productivity, built from the upstream packages/super-sync-server source. It exposes a REST API that the Super Productivity client uses to synchronise tasks across devices via an append-only operation log backed by PostgreSQL.

Patch applied: The upstream project was bootstrapped with prisma db push instead of prisma migrate dev, so no initial CREATE TABLE migration was ever committed. The upstream Dockerfile runs prisma migrate deploy on startup, which would fail on a fresh database because all existing migrations are ALTER TABLE statements that assume the base tables already exist.

The build workflow injects a baseline migration (20200101000000_init) before the Docker build that creates the four core tables (users, operations, user_sync_state, sync_devices) in the state they were in before the first upstream incremental migration. This makes the image fully self-initialising: just point it at an empty PostgreSQL database and it will bootstrap the schema automatically on first start.

See each image's workflow in .github/workflows/ for build inputs and tags. Images are triggered manually via workflow_dispatch with a ref input (tag or commit SHA from the upstream repo).