🧠 MCP Docusaurus Toolkit

A modular Content Management Platform (MCP) to manage, embed, and search Docusaurus documentation using FastAPI, self-hosted embeddings, and PostgreSQL + pgvector.

🚀 Features

• 📁 Auto-generate site map from Docusaurus docs
• 📝 Create, update, and continue editing Markdown docs
• 🧠 Vectorize content using pluggable embedding models
• 🔍 Perform RAG-style semantic search across the docs
• 🧾 Track incomplete docs with watermarks
• 🔄 Sync external edits into the vector database
• 🎨 Apply style/theme transformations to docs
• 🧪 Built-in tool decorators for AI agent control

📂 Project Structure

.
├── app/
│   ├── main.py             # FastAPI entrypoint with FastMCP + tools
│   ├── routes/             # Modular endpoints (docs, search, styles, site\_map)
│   ├── embeddings/         # Embedding model configuration + logic
│   ├── database.py         # PostgreSQL + pgvector connection logic
│   ├── models/             # DB + Pydantic models for documents
│   ├─── utils/              # Sync tools, helpers
│   └── docs/             # Full editable Docusaurus source (used in dev mode)
│
├── data/
│   ├── docusaurus/         # Runtime-synced Markdown docs
│   └── embeddings/         # Persisted vector embeddings
│
├── Dockerfile              # Multi-stage (Node + Python)
├── docker-compose.yml      # Services: FastAPI, Docusaurus Dev, Postgres
├── requirements.txt
└── README.md

🐳 Getting Started (Docker)

# 1. Clone this repository
git clone https://github.com/your-org/mcp-docusaurus.git
cd mcp-docusaurus

# 2. Launch services
docker compose up --build

• 🌐 MCP API available at: http://localhost:8000
• 📚 Docusaurus live dev server: http://localhost:3000

🧰 Tooling Overview (MCP API)

➕ Create a New Document

POST /tool/create_document

{
  "path": "guides/new-feature.md",
  "title": "New Feature",
  "content": "## Work in Progress..."
}

✍️ Update Existing Document

POST /tool/update_docs

{
  "path": "guides/new-feature.md",
  "line_begin": 10,
  "line_end": 15,
  "new_text": "Updated explanation here."
}

🧠 Vector Search (RAG-Style)

POST /tool/search_docs

{
  "query": "How to authenticate a plugin?"
}

🎨 Apply CSS or Theme Style

POST /tool/apply_style

{
  "style_id": "dark-theme",
  "content": "..."
}

🔌 Embedding Model Support

Pluggable embedding model backend with simple config.

✅ Local (Default)

• sentence-transformers/all-MiniLM-L6-v2

🌐 Optional Remote Support

• OpenAI Embeddings (requires API key)

⚙️ Environment Configuration

Use .env file or Docker environment: block:

PG_HOST=postgres
PG_USER=postgres
PG_PASSWORD=postgres
PG_DATABASE=vector_db
EMBEDDING_MODEL=local

🔐 Security

• 🔒 Tool routes are gated with @tool decorators
• ✏️ Document writes are verified before embedding
• 🔑 Optional OAuth2 / API Token guardrails available

📘 Docusaurus Notes

• The doc/ directory holds the full Docusaurus source (used for live editing).
• The data/docusaurus/ directory contains generated Markdown synced with the vector database.

📊 Metrics & Health

• GET /tool/health_check – Check if the API is alive
• GET /tool/metrics – Prometheus/Grafana-compatible stats

📝 License

MIT License ©️ [guidry company]