Ragdocs

🚀 RagDocs Document Knowledge Graph Server

A Model Context Protocol (MCP) server that provides Retrieval Augmented Generation (RAG) functionality using the Qdrant vector database and Ollama/OpenAI embedding models. This server enables semantic search and management of documents through vector similarity.

🚀 Quick Start

RagDocs is a powerful MCP server that leverages vector databases and embedding models to offer RAG capabilities. It allows for semantic search and efficient management of documents.

✨ Features

• Add documents with metadata.
• Search for documents by semantic similarity.
• List and organize documents.
• Delete documents.
• Support both Ollama (free) and OpenAI (paid) embedding models.
• Automatically perform text chunking and embedding generation.
• Use Qdrant for vector storage.

📦 Installation

Prerequisites

• Node.js 16 or higher.
• One of the following Qdrant setups:
  • A local instance using Docker (free).
  • A Qdrant Cloud account with an API key (hosted service).
• One of the following for the embedding model:
  • Locally running Ollama (default, free).
  • An OpenAI API key (optional, paid).

Installation Steps

npm install -g @mcpservers/ragdocs

Install Dependencies

npm install express pg jsonwebtoken cors dotenv sequelize

💻 Usage Examples

Available Tools

1. add_document

Add a document to the RAG system.

Parameters:

• url (required): Document URL/identifier.
• content (required): Document content.
• metadata (optional): Document metadata.
  • title: Document title.
  • contentType: Content type (e.g., "text/markdown").

2. search_documents

Search for stored documents using semantic similarity.

Parameters:

• query (required): Natural language search query.
• options (optional):
  • limit: Maximum number of results (1 - 20, default: 5).
  • scoreThreshold: Minimum similarity score (0 - 1, default: 0.7).
  • filters:
    • domain: Filter by domain.
    • hasCode: Filter documents containing code.
    • after: Filter documents after a specific date (ISO format).
    • before: Filter documents before a specific date (ISO format).

3. list_documents

List all stored documents, supporting pagination and grouping options.

Parameters (all optional):

• page: Page number (default: 1).
• pageSize: Number of documents per page (1 - 100, default: 20).
• groupByDomain: Group by domain (default: false).
• sortBy: Sorting field ("timestamp", "title", or "domain").
• sortOrder: Sorting order ("asc" or "desc").

4. delete_document

Delete a document from the RAG system.

Parameters:

• url (required): URL of the document to delete.

📚 Documentation

Configuration

MCP Server

Edit the config.json file:

{
    "mcp_port": 5000,
    "db_path": "./db",
    "log_level": "info"
}

Start the MCP Server

Run in the terminal:

node index.js

Environment Variables

• MCP_PORT: Specify the port for the MCP server (default: 5000).
• DB_PATH: Specify the storage path for the database file (default: ./db).
• LOG_LEVEL: Set the log level (info/debug/warn/error).

📄 License

Apache License 2.0