migrations.py
- schema_migrations table tracks applied versions (version, description, applied_at)
- MIGRATIONS list is append-only; each entry is (version, description, sql)
- backup() dumps all user-data tables to a timestamped JSON file in backups/
before any schema changes so data can be recovered if something goes wrong
- run_migrations() is idempotent: already-applied versions are skipped
Integration
- app.py calls _run_startup_migrations() at module load so every restart
applies any pending migrations (no-op if schema is current)
- setup_routes.py calls run_migrations() after the initial setup form is
submitted so all tables exist before the user hits the main page for the
first time
- notes.py and admin.py: removed all per-request CREATE TABLE DDL; schema is
now owned entirely by the migration system
Docker
- docker-compose.dev.yml: add backups-data volume so JSON backups survive
container restarts and rebuilds
- Dockerfile: pre-create /app/backend/logs and /app/backend/backups so the
directories exist even before volumes are mounted
Adding future schema changes
- Append a new (version, description, sql) tuple to MIGRATIONS in migrations.py
- The next restart will detect it as pending, back up first, then apply it
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Nirodan