nvml-gpu-ha/cmd/ha-gpu-ccd

ha-gpu-ccd - Home Assistant GPU CCD Temperature Monitor

A tool that retrieves GPU temperature information from Home Assistant MQTT and writes it to sysfs format files.

Features

• Listen to Home Assistant MQTT temperature topics (supports multiple subscription modes)
• Convert temperature from Celsius to millidegrees (e.g., 80.5°C -> 80500)
• Write temperature to /tmp/temp_{DEVICEID} files (sysfs format)
• Support for monitoring specific devices or all devices

Build

go build

Usage

./ha-gpu-ccd --mqtt-host <MQTT_BROKER> --mqtt-port <PORT> --mqtt-username <USERNAME> --mqtt-password <PASSWORD>

Parameters

• --mqtt-host: MQTT broker host (default: localhost)
• --mqtt-port: MQTT broker port (default: 1883)
• --mqtt-username: MQTT username (optional)
• --mqtt-password: MQTT password (optional)
• --temp-dir: Directory to write temperature files (default: /tmp)
• --device-id: Specific GPU device ID to monitor (leave empty to monitor all devices)

Examples

# Connect to local MQTT broker, monitor all GPUs
./ha-gpu-ccd

# Connect to remote MQTT broker with authentication
./ha-gpu-ccd --mqtt-host 192.168.1.100 --mqtt-username homeassistant --mqtt-password mypassword

# Monitor only specific GPU device
./ha-gpu-ccd --device-id 00_04_00_0

# Custom temperature file directory
./ha-gpu-ccd --temp-dir /var/lib/gpu-temps

# Combined usage: specific device with authentication
./ha-gpu-ccd --mqtt-host 192.168.1.100 --mqtt-username user --mqtt-password pass --device-id 00_04_00_0

Output Format

The tool creates a temperature file for each GPU device:

• File name: temp_{DEVICEID}
• Content: Temperature value in millidegrees (integer)
• Location: Default in /tmp/ directory

Example:

• GPU device ID: 00_04_00_0
• Temperature: 80.5°C
• File: /tmp/temp_00_04_00_0
• Content: 80500

MQTT Topic Format

The tool supports two subscription modes:

Monitor All Devices (Default Mode)

Subscribe topic: homeassistant/sensor/nvml-gpu/# Filter condition: Only process messages containing _temperature/state

Monitor Specific Device

Subscribe topic: homeassistant/sensor/nvml-gpu/{DEVICEID}_temperature/state

Temperature topic format:

homeassistant/sensor/nvml-gpu/{DEVICEID}_temperature/state

Where {DEVICEID} is the device ID generated by the nvml-gpu-ha main program, e.g., 00_04_00_0

Logging

The tool outputs the following log messages:

• Connection status
• Subscription status
• Received temperature data
• File write status
• Error messages

Troubleshooting

1. Connection Failed: Check if MQTT broker address and port are correct
2. Authentication Failed: Verify username and password are correct
3. Subscription Failed: Ensure MQTT broker supports wildcard subscriptions
4. No Data Received: Check if nvml-gpu-ha main program is running and sending data

Integration

This tool is designed for integration with other systems, such as:

• System monitoring software
• Temperature control systems
• Automation scripts

By reading the /tmp/temp_{DEVICEID} files, other programs can obtain real-time GPU temperature information.