Schema tables

This page describes the tables in the Deedbox schema. Deedbox owns only these tables; your read models stay in your own tables.

Table	Key	Holds
streams	(tenant_id, stream_id)	One row per stream: stream_type, version, stored state, state_version, state_at (the version the state reflects), deleted_at.
events	global_position	Every event: event_id, tenant_id, stream_id, version, stream_type, event_type, event_version, payload, metadata, occurred_at. Unique on event_id and on (tenant_id, stream_id, version).
position	one row	The global position counter.
event_types	(stream_type, event_type, event_version)	Every event type ever stored; the start-up check reads it.
checkpoints	name	One row per projection and subscription: position, mode, status, error.
jobs	id	Rebuilds, skips, erasures and snapshot rebuilds; the audit trail.
master_keys	(tenant_id, key_version)	Wrapped tenant keys; shredded tenants leave tombstone rows.
subject_keys	(tenant_id, subject_id)	Wrapped subject keys. Erasure deletes a row.
subject_streams	(tenant_id, subject_id, stream_id)	Which streams hold which subject’s data.
schema_version	version	Applied migrations.

On Postgres, JSON columns are jsonb. On SQL Server, they are nvarchar(max), and key columns use the Latin1_General_100_BIN2 collation, so IDs compare case-sensitively on both databases.

Migrations are numbered, idempotent and forward-only. See apply the schema.