← all repositories
mx-space/core

A CMS that actually uses AI for something useful

Personal blog backend with LLM-powered summaries, translation, and moderation wired into the content flow—not bolted on as an afterthought.

535 stars TypeScript Domain AppsOther AI
core
Velocity · 7d
+0.3
★ / day
Trend
steady
star history

What it does

MX Space Core is a headless CMS built on NestJS and PostgreSQL for personal blogs and creator sites. It handles posts, notes, comments, feeds, and search, but the distinguishing feature is a built-in AI workflow: automatic summary generation, multi-language translation, comment moderation, and writing assistance via pluggable LLM providers (OpenAI, Anthropic, OpenRouter, and OpenAI-compatible APIs).

The interesting bit

The project recently executed a hard cutover from MongoDB to PostgreSQL (v11 → v12), complete with a one-shot migration CLI. The API enforces a strict response envelope—every success wraps in { data, meta? }, every error in { error: { code, message } }—with CI-enforced checks to prevent developers from accidentally double-wrapping or leaking raw objects. CamelCase internally, snake_case on the wire, enforced by pipes and interceptors. It’s the kind of pedantic boundary discipline that saves frontend teams headaches.

Key highlights

  • 45 business modules in a NestJS monorepo, including a distributed Redis-backed task queue for AI jobs
  • WebSocket broadcasting via Socket.IO with Redis adapter for multi-instance setups
  • Auth via better-auth with sessions, passkeys, OAuth, and API keys (x-api-key header)
  • Admin SPA built in React 19 + Vite, served from the same server at /proxy/qaqdmin
  • Dual-licensed: AGPLv3 for the apps, MIT for packages and tooling
  • Docker multi-arch images (linux/amd64, linux/arm64) with ~unspecified pull count

Caveats

  • Requires Node.js ≥22, PostgreSQL 16, Redis 7—no shortcuts on dependencies
  • The MongoDB → PostgreSQL migration is one-way and mandatory for v12; plan accordingly
  • AI features depend on external LLM providers; no local model support mentioned

Verdict

Worth a look if you run a personal blog or creator site and want AI assistance integrated into your publishing flow rather than copy-pasting into ChatGPT. Skip it if you need a general-purpose enterprise CMS or want to avoid managing PostgreSQL + Redis + LLM API keys.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.