ForgeStack Documentation
ForgeStack is an opinionated full-stack starter for building production-ready systems. Here's what you get out of the box:
- A Next.js frontend that's already talking to the backend. Registration and login work, there's a translation system for multiple languages, and a design system you can build on top of.
- One or more NestJS backend services built around solid patterns: Domain-Driven Design, CQRS, Hexagonal Architecture, and an event-driven setup where bounded contexts talk to each other asynchronously over Kafka.
- MongoDB for storage and Redis for caching, both wired up (with PostgreSQL support coming soon). The backend uses ports and adapters, so swapping a datastore later doesn't ripple through your code.
- An observability stack that tracks resource usage, API health, command/query throughput, domain and integration events, and more, with logs and traces to go with it.
- A deployment setup with Docker Compose, Caddy for TLS, one-command deploys, and a script that provisions a fresh server for you.
The point of all this is simple: spend your time building the product, not fighting the usual headaches of a scalable distributed system.
These docs walk through every part of the system: how it's put together, the patterns it sticks to, and how to grow it into your own product.
What's inside
- Architecture — the request path from start to finish, and how the layers talk to each other.
- Project Structure — the monorepo layout, the shared libraries, and the TypeScript path aliases.
- Backend — the code patterns that make the backend predictable:
- Domain-Driven Design — aggregates, value objects, domain events.
- CQRS — commands, queries and their handlers.
- Event-Driven Architecture — the outbox/inbox pattern over Kafka.
- Repositories — the ports-and-adapters persistence layer.
- Authentication — JWT access/refresh, guards and revocation.
- Enforced Boundaries — the custom lint rules that actually enforce all of the above.
- Frontend — Next.js App Router, design tokens, i18n.
- Databases — MongoDB, Redis and the swappable datastores.
- Monitoring — metrics, logs and traces with Prometheus, Loki, Tempo and Grafana.
- Deployment — Docker Compose, Caddy TLS and one-command deploys.
The guiding idea
Every convention here exists to keep the codebase predictable, both for people and for AI coding agents. When the structure is consistent and the boundaries are enforced, you can open any bounded context and know right away where the business logic, the wiring, and the I/O live.
On the roadmap
Some things I'm planning but haven't shipped yet:
- PostgreSQL support. A relational adapter next to MongoDB, so you can pick whichever store fits a given bounded context. Thanks to ports and adapters, this swap stays contained.
- AI agent skills. A set of skills that teach coding agents the project's patterns, so they can extend things without breaking the boundaries.