Deployment Architecture

How to run one Bellamy Book instance (self-hosted social network) in different environments.

Operator install guides: Self-Hosting.

Choose a path

Goal	Path	What you use
Customer self-host (no source build)	Pre-built images	dockerPublish kit — Self-host with pre-built images
Test the kit on a laptop	Same images, local override	Local testing
Production from monorepo	Build full stack	dockerProd Compose (API, workers, frontend, admin, docs, Traefik)
Develop features	Infra Docker + apps on host	dockerLocalENV + run API / Frontend / Admin / workers locally
CI/CD	Build → registry → deploy	Jenkins / Harbor (CICD/)

                 What do you need?
        ┌───────────────┼───────────────┐
        ▼               ▼               ▼
  Local develop   Customer install   Your prod VPS
  dockerLocalENV    dockerPublish      dockerProd
  + run apps        pull images        full compose

What a full deployment includes

Group	Components
Apps	Frontend (Nginx), Admin (Nginx), API, Docs site
Workers	Interaction, Scoring, Trending, Graph, Hashtag, ES Sync, Media, Chat, WebSocket, Web Push, Expo Push, Blog AI, Content Moderation
Data	PostgreSQL (± replica), MongoDB, Redis, RabbitMQ, Kafka, Elasticsearch, MinIO/R2, Neo4j
Edge	Traefik (DNS hostnames, TLS) — see Traefik when you deploy
Optional	LiveKit, monitoring (Prometheus/Grafana/…)

Topology (production)

Internet → Traefik (80/443)
              ├─ frontend host  → frontend
              ├─ admin host     → admin
              ├─ api host       → api
              ├─ /hubs/chat*    → chat-worker
              └─ /hubs/notification* → websocket-worker
                         │
              API + workers ↔ Postgres/Mongo/Redis/Kafka/Rabbit/ES/MinIO/Neo4j

Environment essentials

• One .env for self-host kit (domains, secrets, MinIO, JWT) — Environment
• Frontend / Admin public URLs must match CORS and Traefik hosts
• Run DB migration once (db-migration / DatabaseApplication)
• Activate license in Admin after first boot (CONSUMER)

Scaling notes

• Scale API and workers horizontally; keep Kafka consumer groups and RabbitMQ consumers in mind
• Media worker scales by polling load; chat/notification workers need Redis backplane for multi-instance SignalR
• Details: Scaling

Related

• Self-host intro
• Workers
• Security
• Monitoring
• Backups