Why Phase 2 is boring on purpose

Phase 1 shipped the architecture package: system design, ADRs 0001–0018, C4, contracts. The next temptation is SMS + templates + campaigns before the host can say no to a bad JWT.

Phase 2 stops there. Boot plumbing. No Twilio. No SendGrid. No cute provider yet.

What "foundation complete" means

Piece	Choice
Shape	Modular monolith (Xnp.Api + Xnp.Worker + SharedKernel + Infrastructure)
Database	PostgreSQL 16 + EF Core — tenants, subscribers, outbox/inbox, audit
Auth	JWT + tenant middleware — unsigned requests fail
Messaging	IMessageBus + in-memory bus; in-process outbox dispatcher (single writer)
Ops	/health/live, /health/ready, OpenTelemetry hooks
Local	docker compose for Postgres; full profile builds API + DB

Apps still must not call Twilio / SendGrid / FCM directly. They will talk to XNP later. Today XNP must exist as a process that can reject, accept, and dispatch outbox rows without lying about providers.

ADRs that pin Phase 2 implementation

• ADR-0019 — Phase 2 foundation skeleton (exit criteria: builds, tests, compose, JWT, health).
• ADR-0020 — Keep the outbox dispatcher in-process for Phase 2; Azure Service Bus remains the production broker (ADR-0003) for later phases.

What we did not ship

Subscriber CRUD vertical slice. Templates. SMS. Schedules. Campaigns. Admin UI. Production Service Bus consumers.

That is Phase 3+. Roadmap already says one vertical slice before horizontal fan-out.

Takeaway

Empty and honest beats half-wired channels. Phase 2 is the scaffolding you can refuse to skip when someone asks for "just one Twilio call."

Further reading: docs/architecture/ · ADR-0019 / ADR-0020 in the same repo.