使用说明

PLC-Bridge MCP Server

Industrial automation bridge connecting Beckhoff TwinCAT PLCs to AI assistants via the Model Context Protocol (MCP).

Features

Live PLC Integration: Read real-time data from Beckhoff TwinCAT PLCs via ADS/AMS
Interactive Web Dashboard: Modern web interface with conversational AI for querying PLC data
AI-Powered Dashboard Generator: Uses DeepSeek LLM to intelligently create Grafana dashboards
Local AI Chat: Interact with PLC data using natural language via DeepSeek (offline)
Autonomous Dashboards: Generate Grafana dashboards programmatically using grafanalib
Optimized Performance: Batch reads and symbol caching for large projects
Mock Mode: Test without hardware using simulated data

Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Run the Web Dashboard (Recommended)

.venv\Scripts\python.exe web_dashboard_server.py

Then open http://localhost:5000 in your browser.

Features:

Ask questions: "Show me all metrics", "What is the OEE?"
Live dashboard updates every 5 seconds
Modern dark theme with glassmorphism effects

3. Alternative: Run the AI Chat Host

.venv\Scripts\python.exe ollama_host.py

Default: LIVE mode. Configure the PLC address for your own lab environment; public examples should not contain production or customer network details.

To use MOCK mode instead, edit ollama_host.py and change:

server_env["PLC_MODE"] = "LIVE"  # Change to "MOCK"

Project Structure

Core/
├── web_dashboard_server.py  # Web dashboard server
├── static/                   # Web dashboard frontend
│   ├── index.html           # Main page
│   ├── styles.css           # Styling
│   └── app.js               # JavaScript logic
├── server.py                # MCP server with PLC tools
├── ollama_host.py           # Local AI chat interface
├── ai_dashboard_generator.py # AI-powered dashboard generator
├── dashboard_generator.py   # Template-based dashboard creation
├── drivers/
│   ├── beckhoff.py         # TwinCAT ADS driver
│   └── mock.py             # Simulated PLC driver
├── plc_settings/
│   └── Testing_Machine.json # Machine configuration
├── tests/
│   └── test_plc_connection.py  # Connection verification
└── dashboards/              # Generated dashboard JSONs

Available Interfaces

1. Web Dashboard

Modern web interface with conversational AI:

URL: http://localhost:5000
Features: Live metrics, chat interface, auto-refresh
Best for: Monitoring and visualization

2. AI Chat Host

Terminal-based chat interface:

Command: .venv\Scripts\python.exe ollama_host.py
Features: Natural language queries, tool calling
Best for: Advanced queries and automation

3. MCP Tools

When using the AI host, you can:

List machines: "Show me the machines"
List tags: "What tags are available?"
Read values: "Read all tag values" or "What is the OEE?"
Generate dashboards: "Create a dashboard for Testing_Machine"

Configuration

Edit plc_settings/Testing_Machine.json to configure:

Machine name and type
IP address
Tag definitions with types and descriptions

Documentation

LIVE_MODE.md - Guide for connecting to real PLCs
TESTING_DASHBOARDS.md - Dashboard testing guide
changelog.md - Version history
.agent/rules/plc-safety.md - Safety guidelines

Requirements

Python 3.12+
Ollama with DeepSeek model (for local AI)
TwinCAT ADS (for live PLC connection)
Grafanalib (for dashboard generation)
Flask, Flask-SocketIO (for web dashboard)

Web Dashboard Usage

Start the server:

.venv\Scripts\python.exe web_dashboard_server.py

Open browser: Navigate to http://localhost:5000
Ask questions:
- "Show me all metrics" - Displays all PLC tags
- "What is the OEE?" - Shows OEE metric
- "Show production data" - Displays production metrics
Watch live updates: Dashboard auto-refreshes every 5 seconds

Technology Stack

Backend: Flask + Flask-SocketIO for real-time communication
Frontend: HTML5 + CSS3 + Vanilla JavaScript
PLC Communication: pyads for TwinCAT ADS protocol
AI: DeepSeek via Ollama for intelligent analysis
Dashboard Generation: Grafanalib for programmatic dashboard creation