CMS Cloud
Visit ProjectCMS Cloud is a self-hosted, API-first headless CMS that I architected and built end to end — the admin dashboard, the APIs behind it, the delivery layer and the infrastructure it runs on. Its content APIs follow established headless-CMS conventions, which keeps migrating off an existing platform straightforward. The dashboard covers the full editorial surface: a content-type builder, entries with scheduling and per-locale translation, a media library, configurable search, spaces and environments with aliases, roles and permissions, API keys and CMA tokens, an audit log and schema migrations.
Underneath, it splits into independent services — an admin API, a read-only GraphQL delivery API, and a REST management API for programmatic writes — backed by PostgreSQL, Redis, and Cloudflare Workers that cache GraphQL at the edge and transform images on the fly. Three-layer caching with stale-while-revalidate and real-time invalidation keeps delivery fast, and every space gets its own performance view: request volume, cache hit ratio and p50/p95/p99 latency broken down per query. It runs in production today behind webook.com, ahlan.sa and scbraga.pt.
Because data ownership is rarely negotiable, CMS Cloud is deployed as a multi-zone, multi-region service: each customer’s content can be pinned to the region their regulator requires, which is what lets one platform satisfy GDPR in Europe and PDPL in Saudi Arabia — and any other residency regime a client needs. Search runs dual-layer: a self-hosted Meilisearch cluster handles lexical matching while OpenAI embeddings stored in Postgres through pgvector carry semantic meaning, so the same query returns sharper results.