AI Agent MQTT Integration

This application demonstrates real-time AI agent inference using MQTT messaging. It consists of a message producer that generates simulated data, an MQTT broker that handles message routing, and an AI agent that processes the data and returns predictions.

Features

• MQTT-based messaging system for real-time data processing
• Dockerized setup with multiple services
• Automated message generation for testing
• Real-time agent predictions
• Support for MQTT monitoring tools

Project Structure

├── Dockerfile                # Dockerfile for the agent service
├── Dockerfile.producer       # Dockerfile for the message producer
├── model/                    # Directory for the trained model
│   └── agent.json            # Trained model file
├── agent_inference.py        # Main agent service that processes MQTT messages
├── mqtt_producer.py         # Script that generates test messages
├── docker-compose.yml       # Docker compose configuration
├── mosquitto/               # Mosquitto broker configuration
│   ├── config/             
│   │   └── mosquitto.conf   # MQTT broker configuration
│   ├── data/               # Persistent storage for MQTT messages
│   └── log/                # Broker logs
├── .env                    # Environment file (for license key)
├── requirements.txt        # Python dependencies for agent service
├── producer_requirements.txt # Python dependencies for producer
└── README.md              

Requirements

• Docker
• Docker Compose
• Python 3.10.x
• MQTT Explorer (optional, for monitoring)

Getting Started

1. Clone the Repository

git clone https://github.com/composabl/agent-mqtt-app.git
cd agent-mqtt-app

2. Set up Mosquitto Configuration Create the necessary directories and configuration:

mkdir -p mosquitto/data mosquitto/log

Create mosquitto/config/mosquitto.conf with:

listener 1883
allow_anonymous true
persistence true
persistence_location /mosquitto/data/
log_dest file /mosquitto/log/mosquitto.log
log_type all

3. Environment Variables Make sure to add your COMPOSABL_LICENSE key in the .env file:

COMPOSABL_LICENSE=your_license_key_here

4. Start the Services

docker-compose up --build

This will start:

• MQTT Broker (Mosquitto)
• Agent Service
• Message Producer

Testing the Application

Using MQTT Explorer

1. Download and install MQTT Explorer from mqtt-explorer.com
2. Connect to broker:
   • Host: localhost
   • Port: 1883
3. Subscribe to topics:
   • agent/observation (input data)
   • agent/action (agent predictions)

Manual Testing

You can publish test messages using mosquitto_pub:

mosquitto_pub -h localhost -t "agent/observation" -m '{
  "observation": {
    "T": 311.0,
    "Tc": 292.0,
    "Ca": 8.56,
    "Cref": 8.56,
    "Tref": 311.0,
    "Conc_Error": 0.0,
    "Eps_Yield": 0.0,
    "Cb_Prod": 0.0
  }
}'

Subscribe to responses:

mosquitto_sub -h localhost -t "agent/action" -v

Message Format

Input Message (agent/observation)

{
  "observation": {
    "T": 311.0,
    "Tc": 292.0,
    "Ca": 8.56,
    "Cref": 8.56,
    "Tref": 311.0,
    "Conc_Error": 0.0,
    "Eps_Yield": 0.0,
    "Cb_Prod": 0.0
  }
}

Output Message (agent/action)

{
  "action": [10.0]
}

Development Setup (Without Docker)

If you prefer to run the services locally:

1. Install Mosquitto broker:

# Ubuntu/Debian
sudo apt-get install mosquitto

# MacOS
brew install mosquitto

2. Install Python dependencies:

pip install -r requirements.txt
pip install -r producer_requirements.txt

3. Start the services:

# Terminal 1: Start Mosquitto
mosquitto -c mosquitto/config/mosquitto.conf

# Terminal 2: Start the agent
python agent_inference.py

# Terminal 3: Start the producer
python mqtt_producer.py

Monitoring and Debugging

View Logs

# All services
docker-compose logs -f

# Specific service
docker-compose logs -f composabl-mqtt-broker
docker-compose logs -f composabl-agent-mqtt
docker-compose logs -f composabl-mqtt-producer

Message Producer Options

The message producer supports several command-line arguments:

# Custom interval (e.g., 2 seconds between messages)
python mqtt_producer.py -i 2

# Verbose logging
python mqtt_producer.py -v

Notes

• The MQTT broker is configured for development use. Additional security measures should be implemented for production.
• The message producer generates random data within realistic ranges for testing.
• MQTT QoS level 2 is used for guaranteed message delivery.

License

Make sure you have a valid license for the AI agent (COMPOSABL_LICENSE), as this is required for the app to work.

Contributions

Feel free to contribute by submitting pull requests or opening issues!