| name | event-driven-architecture |
| description | Designs event-driven systems with asynchronous flows, brokers, producers, consumers and contracts. Use when architecting asynchronous, decoupled, or event-first systems. |
Event-Driven Architecture
When to use
Use when designing asynchronous integration or event-first systems.
Objective
Produce a practical, concise, traceable architecture artefact that a coding agent can use to guide implementation or review.
Procedure
- Identify triggers and outcomes.
- Separate events, commands and queries.
- Identify producers/consumers.
- Define topics/streams.
- Define schemas/contracts.
- Define ordering, partitioning, idempotency and replay.
- Define failure handling.
- Define observability.
Required outputs
- Event flow
- Producers/consumers
- Topic/stream design
- Contracts
- Ordering/idempotency/replay decisions
- Operational controls
Best-practice alignment
Apply event-driven best practice: business event semantics, producer/consumer ownership, schema contracts, compatibility, ordering, partitioning, idempotency, replay, retention, lineage, observability and operational SLOs.
Use open standards where they fit: CloudEvents for a common event envelope and metadata (id, source, type, time, dataschema), and AsyncAPI for application-level channel, operation and message contracts. CloudEvents and AsyncAPI are complementary (envelope vs interface).
Quality checks
- Events have business meaning.
- Ownership is clear.
- Duplicate/replay behaviour is handled.
- Observability exists.
Avoid
Do not use events to disguise tight coupling.
Decision aid
Choose event-driven architecture when producers and consumers can evolve independently, consumers need replay, or business facts must be observed by multiple domains. Prefer request/response when the caller needs an immediate answer and no other consumer benefits. For each event, record owner, contract, idempotency key, ordering assumption, retry policy and observable failure path.
References
Verification
