main.py

"""
DaRUS MCP Server

This module provides a Model Context Protocol (MCP) server for interacting with the
DaRUS (Data Repository of the University of Stuttgart) Dataverse instance.

The server exposes Dataverse functionality through MCP tools, allowing AI assistants
to search, retrieve, and interact with datasets and metadata stored in DaRUS.

Features:
    - Search datasets and dataverses in DaRUS
    - Retrieve dataset metadata and files
    - Access publication information
    - Integration with MathModDB for mathematical model data

Usage:
    Run this server to expose DaRUS functionality via MCP:

    $ python main.py

    The server will start on http://0.0.0.0:8000 and provide MCP tools for:
    - Dataverse operations (search, metadata retrieval)
    - MathModDB integration (mathematical models and data)

Dependencies:
    - fastmcp: MCP server framework
    - pyDataverse: Dataverse API client
    - asyncio: Async runtime support
"""

import asyncio

from fastmcp import Client, FastMCP
from pyDataverse import Dataverse
from pyDataverse.mcp import DataverseMCP

# Initialize connection to the DaRUS Dataverse instance
# DaRUS is the institutional data repository of the University of Stuttgart
darus = Dataverse(
    base_url="https://darus.uni-stuttgart.de",
)


# Create the FastMCP application
# This serves as the main MCP server instance that will expose tools to clients
app = FastMCP(name="darus-mcp", version="0.1.0")

# Register Dataverse MCP tools with the application
# This adds all Dataverse-related tools (search, metadata, files) to the MCP server
DataverseMCP(dataverse=darus).to_mcp(app)

if __name__ == "__main__":
    """
    Main entry point for the DaRUS MCP server.
    
    Starts the MCP server using streamable HTTP transport on all interfaces
    at port 8000. The server will handle MCP requests and provide access to
    both DaRUS Dataverse tools and imported MathModDB tools.
    
    Server Configuration:
        - Transport: streamable-http (compatible with MCP clients)
        - Host: 0.0.0.0 (accepts connections from any interface)
        - Port: 8000 (standard development port)
    """
    # Start the MCP server
    app.run(
        transport="streamable-http",
        host="0.0.0.0",
        port=8000,
    )