Architecture

Components, planes, modules, and how requests flow through the platform.

Architecture overview

Design principles

Control plane vs data plane

Control plane vs data plane
Control planeData plane
Provider register / deregisterPublish, subscribe, consume
GET /api/v1/control/plane/api/v1/connections/...
UI cascade (which tabs to show)WebSocket /ws/stream
OpenAPI catalog per deploymentInspect, Kafka admin APIs
No TCP to brokersAll broker clients live here

Backend modules

eventore-core          Domain, SPI, control/dataplane registries
eventore-api-codegen   OpenAPI → Spring controllers (per stream)
eventore-provider-*    One module per protocol (connector + inspector)
eventore-server        REST, WebSocket, security, delegates

MCP server (optional third workload)

mcp/eventore-mcp is a Node MCP adapter for AI clients. It proxies to the backend only:

Deploy via docker/Dockerfile.mcp or Helm chart eventore-mcp. Full reference: MCP for AI agents.

Frontend application

React + React Query. Routes:

The side Streams panel persists subscriptions in cookies via StreamWorkspaceContext.

Stream workspace UI

Provider registration flow

  1. Spring Boot loads StreamProvider beans from enabled provider JARs.
  2. On startup, ControlPlaneBootstrap registers each allowed protocol.
  3. ControlPlaneCoordinator attaches the provider to DataPlaneRegistry.
  4. GET /config returns controlPlane.uiCascade for the UI.
  5. Admins may POST /control/providers/{protocol}/register or deregister at runtime (ADMIN).

API surfaces (generated + manual)

SurfacePath prefixPlane
Core/api/v1/config, connections, publishMixed
Control/api/v1/control/*Control
Inspect.../inspect/*Data
Kafka.../kafka/*Data
Stream/ws/streamData

Reference