di37/nlp-fastapi-mcp-server

Electrical NER API and MCP Service

This project provides service for Named Entity Recognition (NER) specifically tailored for the Electrical Engineering domains. It can be accessed both as a standard RESTful API and as a Model Context Protocol (MCP) service, making it easy to integrate into various workflows.

Features

• Dual Interface: Accessible via a standard RESTful API and as a Model Context Protocol (MCP) service.
• High Performance: Built with FastAPI and Uvicorn for fast, asynchronous request handling.
• Specialized NER Model: Utilizes a fine-tuned NER model specifically tailored for the Electrical Engineering domain.
• Seamless MCP Integration: Exposes core NER functions as an MCP service for easy use with agentic AI systems.

API Documentation

The API provides the following endpoints:

GET /

Root endpoint to verify that the service is running.

• Success Response (200):

  {
    "status": "running",
    "timestamp": "2023-10-27T10:00:00.000Z",
    "service": "Electrical NER API",
    "version": "1.0.0"
  }
  

GET /health

Basic health check endpoint.

• Success Response (200):

  {
    "status": "healthy",
    "timestamp": "2023-10-27T10:00:00.000Z",
    "service": "Electrical NER API",
    "version": "1.0.0"
  }
  

GET /all_entities

Retrieves a list of all possible entity groups that the model can recognize.

• Success Response (200):

  [
    "COMPONENT",
    "DESIGN_PARAM",
    "MATERIAL",
    "EQUIPMENT",
    "TECHNOLOGY",
    "SOFTWARE",
    "STANDARD",
    "VENDOR",
    "PRODUCT"
  ]
  

POST /predict

Extracts entities from the provided input text.

• Request Body:

  {
    "text": "The LM358 operational amplifier from Texas Instruments needs a dual power supply."
  }
  

• Success Response (200):

  {
    "success": true,
    "message": "Entities detected successfully",
    "entities": [
      {
        "entity_group": "PRODUCT",
        "word": "LM358"
      },
      {
        "entity_group": "COMPONENT",
        "word": "operational amplifier"
      },
      {
        "entity_group": "VENDOR",
        "word": "Texas Instruments"
      },
      {
        "entity_group": "DESIGN_PARAM",
        "word": "dual power supply"
      }
    ]
  }
  

• Error Response (500):

  {
    "detail": "Error message explaining the failure."
  }
  

MCP (Model Context Protocol) Integration

The service uses fastapi-mcp to expose its core functionality as MCP operations. This allows AI agents or other MCP-compatible systems to directly call the NER functions as tools.

Exposed Operations

• extract_entities: Corresponds to the POST /predict endpoint. Takes a text input and returns extracted entities.
• get_all_entities: Corresponds to the GET /all_entities endpoint. Returns a list of all possible entity types.

Setup and Installation

1. Clone the repository:

   git clone https://github.com/di37/nlp-fastapi-mcp-server.git
   cd nlp-fastapi-mcp-server
   

2. Create and activate a virtual environment:

   python -m venv .venv
   source .venv/bin/activate  # On Windows, use `.venv\Scripts\activate`
   

3. Install dependencies:

   pip install -r requirements.txt
   

Running the Application

To start the API server, run the following command:

python api_mcp.py

The server will be available at http://0.0.0.0:5152.

MCP Configuration

To connect to this service from an MCP client (like another AI agent), use the following configuration:

For Windsurf:

{
    "mcpServers": {
        "electrical-ner-mcp":{
            "serverUrl": "http://localhost:5152/mcp"
        }
    }
}

For Cursor:

{
    "mcpServers": {
        "electrical-ner-mcp":{
            "url": "http://localhost:5152/mcp",
            "transport": "streamable-http"
        }
    }
}

Claude Desktop still does not support url MCP server. Will update the documentation once it is supported.