Architecture Deep Dive
Purpose
This page explains Cyprob EE architecture boundaries and runtime flow for technical evaluators.
Validity Note
- Status:
CONFIRMED - Basis: Architecture and behavior claims are checked against the current stable release.
Layering Model
Cyprob EE follows strict DDD layering:
- Transport: HTTP handlers, DTO parsing, route security metadata
- Application: orchestration, transactions, workflow decisions
- Domain: business rules, entities, invariants
- Infrastructure: PostgreSQL, external systems, persistence
Runtime Topology
Core runtime components:
- API/control plane
- Worker execution plane
- PostgreSQL state store
- Optional runtime services (reverse proxy)
Orchestration Pattern
The scan orchestration model is queue-driven and phase-based:
- Scan requested
- Jobs enqueued per phase/chunk
- Workers execute and publish progress/results
- Phase completion advances orchestration
Reliability Controls
- Duplicate protection: scan completion is idempotent; next-phase task creation runs only on the leader node and removes duplicates within a run
- Heartbeat-based worker liveness tracking
- Retry/reassignment behavior for failed worker execution
- System-state gates (booting / setup-needed / operational / read-only / maintenance); read-only allows GET only, maintenance admits platform admins only
Security-Relevant Boundaries
- Org-scoped API boundaries enforced in route/middleware model
- Permission-scoped endpoints (
resource:action) - Secret retrieval separated from application business logic
Next Action
Continue with Scanning Lifecycle.