In CQRS, an aggregate is not a read model. It only needs the state required to decide whether a command is valid and which events to emit.
Workflow
-
Identify each command the aggregate handles.
-
List the state needed to validate and execute those commands.
-
Remove query-only projections, history lists, counters, denormalized display fields, and report data from the command aggregate.
-
Put read concerns into read models updated from events.
-
Redraw aggregate boundaries around true invariants and command contention.
Modeling Rules
- Keep state minimal, invariant-bearing, and command-oriented.
-
Prefer IDs and event references over embedded external aggregates.
-
Model lifecycle as explicit state transitions when commands depend on it.
-
Treat unbounded collections as read-model material unless they enforce a true invariant.
Detailed Reference
For non-trivial implementation, review, or refactoring work, read references/details.md before giving final guidance. It contains the detailed rules, examples, smells, and migration notes that do not belong in the short invocation body.
Output
Return the command-state matrix, proposed aggregate boundary, events, read models, and migration risks.