Event-Driven Architecture Overview

The MATIH Platform uses an event-driven architecture alongside synchronous REST APIs to handle asynchronous communication, cross-service coordination, and real-time updates. This section documents the event infrastructure, including Apache Kafka for durable event streaming and Redis Pub/Sub for ephemeral real-time notifications.

Event Infrastructure

System	Use Case	Delivery	Persistence
Apache Kafka	Durable events, cross-service coordination	At-least-once (idempotent producers)	Yes (configurable retention)
Redis Pub/Sub	Real-time notifications, config broadcasts	At-most-once (fire-and-forget)	No

The selection rule is straightforward: if the event must not be lost, use Kafka; if the event is ephemeral and latency matters more than durability, use Redis Pub/Sub.

Sections

Section	Description
Kafka Topology	Cluster configuration, topic design, partitioning strategy
Event Schemas	The `DataPlaneEvent` base class, categories, chaining
Redis Pub/Sub	Real-time channels, config propagation, AI streaming
WebSocket Architecture	Browser-to-server real-time communication
CDC Patterns	Change Data Capture for database synchronization

Event Flow Summary

Producer Service
  |
  +-- Durable events --> Kafka --> Consumer Services
  |                                (audit, billing, notification)
  |
  +-- Real-time events --> Redis Pub/Sub --> WebSocket Bridge --> Browser
  |                                        (AI streaming, config updates)
  |
  +-- Database changes --> CDC (Debezium) --> Kafka --> Flink/Consumers

Related Sections

Service Topology -- How events flow between services
Multi-Tenancy -- Tenant isolation in event streams
Data Flow -- Events in the context of request processing

Event-Driven Architecture Kafka Topology