accordsync-django

accordsync-django

The Accord sync server for Django: an app with the /v1/push, /v1/pull and /health URLs, management commands accord_migrate and accord_compact, and a system check. All behaviour comes from accordsync-server; this package only routes to it. Part of Accord.

Install

pip install accordsync-django

Configure

# settings.py
INSTALLED_APPS = [..., "accordsync_django"]
ACCORD_SERVER = "myapp.sync:server"  # a ServerDefinition made with define_server(...)
ACCORD_DATABASE_URL = "postgresql://..."  # optional, see below
ACCORD_POOL_SIZE = 20  # optional
ACCORD_SCHEDULE_COMPACTION = True  # optional

# urls.py
urlpatterns = [path("sync/", include("accordsync_django.urls")), ...]

The database is ACCORD_DATABASE_URL (setting, else environment variable), or else the default database's settings, which must then use django.db.backends.postgresql. The sync server uses its own psycopg connection pool, opened on the first request, never Django's ORM connections. The views are CSRF-exempt, opt out of ATOMIC_REQUESTS, and need no session or auth middleware (requests authenticate with the JWT in Authorization).

manage.py check reports accordsync.E001 when ACCORD_SERVER is missing, cannot be imported or is not a ServerDefinition, and accordsync.E002 when no PostgreSQL database is configured.

python manage.py accord_migrate

Compaction

With compaction.interval_ms > 0 in the definition (default one hour), each process compacts on that interval in a background thread started by its first sync request. With several worker processes, set ACCORD_SCHEDULE_COMPACTION = False and run from cron:

python manage.py accord_compact

Notes

• Serve with a threaded WSGI server (e.g. waitress, gunicorn --threads); each request holds a thread while it waits on PostgreSQL.
• Rate limits are per process (in memory).
• examples/django_app in this repository serves the conformance profile with waitress and passes the shared server conformance suite (68/68). Its control API is for tests only (ACCORD_CONTROL=1).