Solution
GPU and AI server power monitoring
Power Usage is a FreePBX 17 report module that shows how much electricity the machine running your AI Agent's local speech and language models is drawing, where that power is going, and what it costs under your electricity tariff.
It is an add-on for the AI Agent module and does not work on its own: the readings are taken every second on the services host by the AI Agent control daemon, and Power Usage is the FreePBX page that reads, charts and prices them.
Who it is for
- AI Agent users running local models on their own GPU server
- Teams that need to know what their AI server costs to run
Products
Main features
Live and historical
A live chart of the last five minutes updated every second, and history from 1 hour to 1 year or any custom period.
Per GPU, CPU and whole PC
Each NVIDIA GPU is measured; CPU and whole-PC figures are estimates unless the CPU's energy counters can be read.
Top apps
Which programs, services, Docker containers and Ollama models the power goes to, split by CPU and GPU.
Tiered tariff cost
Price levels per kWh with a fine when a level is reached, the billing cycle so far and a projection to its end.
Use cases
- See what each GPU draws while the AI agent handles calls.
- Find which model or service uses the most power.
- Project this billing cycle's electricity cost under a tiered tariff.
Prices and requirements
| Product | Price (USD) | Requirements |
|---|---|---|
| Power Usage | $49 – One-time fee per FreePBX server | AI Agent module 1.53.0 or later with its control daemon running on the services host; NVIDIA GPUs with nvidia-smi for GPU readings |
| AI Agent for FreePBX | $799 – One-time fee per FreePBX server | FreePBX 17 / Asterisk 21; an LLM (Ollama server, or an OpenAI or Anthropic key); speech-to-text and text-to-speech services or cloud keys – not included |
Frequently asked questions
Which figures are measured and which are estimated?
GPUs are measured, one series per NVIDIA card, with nvidia-smi. CPU power is estimated from load and the CPU's rating unless its RAPL energy counters can be read. Whole-PC power is always an estimate.
Does it need the AI Agent module?
Yes. It is a hard dependency: Power Usage needs AI Agent 1.53.0 or later on the same FreePBX, and the AI Agent control daemon running on the services host, reachable from FreePBX on port 2705.
Which GPUs are supported?
NVIDIA GPUs with the NVIDIA driver. Without them, only the CPU and whole-PC estimates are shown.
How much history is kept?
About 400 days, kept on the services host, so it keeps building up while FreePBX is down or being upgraded.