MarkdownHub

Self-hosted collaborative markdown workspace with real-time editing, Git integration, and AI-powered code generation.

Features

• Three editor modes: WYSIWYG (Milkdown), raw markdown (CodeMirror-style), and split-pane with live preview
• File management: Tree-based file browser, folders, search, create/delete
• Git versioning: Auto-commit on save, full history, diff viewer, restore previous versions
• External Git sync: Per-user remotes (Gitea, GitHub, etc.), push/pull on demand
• Sharing: Share files/folders with other users (read-only or read-write)
• AI integration: Verify specs, generate prompts, summarize — via LiteLLM/vLLM
• AI Build Pipeline: Write a spec → verify → build entire project → push to Gitea
• 2FA: TOTP-based two-factor authentication
• Light/Dark theme: Toggle in UI, persisted in preferences
• Export: PDF, HTML, raw markdown download
• Offline-ready: PWA architecture (planned)
• Multi-device: Responsive, works on desktop and iPad
• Docker deployment: Single container, easy backup

Quick Start

Docker (recommended)

docker build -t markdownhub .
docker run -d --name markdownhub \
  -p 8080:8080 \
  -v ./data:/app/data \
  -e MH_SECRET=your-random-secret \
  markdownhub

Default admin: admin@localhost / admin

Development

Backend:

go run ./cmd/server

Frontend:

cd frontend
npm install
npm run dev

Frontend dev server proxies API calls to localhost:8080.

Configuration

Environment variables:

Variable	Default	Description
MH_PORT	8080	Server port
MH_DATA_DIR	./data	Data directory (SQLite + files)
MH_SECRET	dev-secret-change-me	JWT signing secret
MH_ADMIN_EMAIL	admin@localhost	Initial admin email
MH_ADMIN_PASSWORD	admin	Initial admin password
MH_AI_ENDPOINT	—	LiteLLM-compatible API URL
MH_AI_API_KEY	—	API key for AI endpoint
MH_AI_MODEL	gpt-4	Default model for AI features

Architecture

┌─────────────┐     ┌──────────────┐     ┌────────┐
│  Frontend   │────▶│  Go Backend  │────▶│ SQLite │
│  (Vue 3)    │◀────│  (net/http)  │     └────────┘
└─────────────┘     │              │────▶ Filesystem (.md files)
                    │              │────▶ Git (per-user repos)
                    └──────────────┘
                           │
                    ┌──────▼───────┐     ┌────────┐
                    │ Build Daemon │────▶│ Gitea  │
                    │  (Python)    │     └────────┘
                    └──────────────┘

• Backend: Go, stdlib net/http, SQLite (pure Go via modernc.org/sqlite)
• Frontend: Vue 3, Vite, Milkdown, CodeMirror 6, marked
• Auth: JWT + bcrypt + TOTP
• Git: Shell-based git operations (auto-commit, remotes, push/pull)
• AI: LiteLLM-compatible OpenAI API calls
• Build Daemon: Python (stdlib only), polls for build jobs, runs Pi coding agent

Project Structure

cmd/server/          — Web server entry point
cmd/mdsync/          — CLI sync tool (planned)
internal/api/        — HTTP handlers & routing
internal/auth/       — JWT, bcrypt, TOTP
internal/db/         — SQLite schema & connection
internal/files/      — File CRUD & search
internal/git/        — Git operations
internal/crypto/     — AES-256-GCM encryption (for private vault)
daemon/              — Build daemon (Python)
frontend/            — Vue 3 SPA

Build Daemon

The build daemon runs on a separate Linux machine and processes AI build jobs:

export MH_URL=http://your-markdownhub:8080
export MH_DAEMON_TOKEN=your-token
export MH_WORKSPACE=~/builds
python3 daemon/build_daemon.py

It polls MarkdownHub for pending jobs, creates Gitea repos, runs the Pi coding agent, and pushes results.

Backup & Restore

All data lives in the data/ volume:

• data/db/markdownhub.db — SQLite database
• data/files/{user_id}/ — User markdown files (also git repos)

Backup: Copy the data/ directory. Restore: Stop container, replace data/, start container.

License

MIT