olmon

A lightweight CLI to monitor your Ollama models and usage in real time.

$ olmon top

╭─────────────────────────────────────── olmon top 01:09:32 ───────────────────────────────────────╮
│ Model                        VRAM              VRAM %            Expires In      Status          │
│ ornith:9b                    5.0 GB            ██░░░  41%        0m 22s          ● expiring      │
│ phi4-mini:latest             2.9 GB            █░░░░  23%        0m 17s          ● expiring      │
╰────────────────────────────── VRAM: 7.8 GB / 12.0 GB |  2 running ───────────────────────────────╯

Features

• 🔄 Real-time dashboard — live auto-refreshing view of running models
• 📋 Model listing — browse all installed models with size, family, and quantization
• 🔍 Model inspection — full details on any installed model
• 🟢 Status indicators — green / blue / red at a glance
• ⚙️ Configurable — set your API host and refresh interval
• 🪶 Lightweight — minimal dependencies, works over SSH on headless servers
• 🖥️ htop-style monitoring — olmon top with VRAM usage and expiry countdown
• 🛑 Model control — force unload models from VRAM
• ⚖️ Model comparison — side by side spec comparison of multiple models
• 🔧 Scripting friendly — --json flag and exit codes on every command

Requirements

• Python 3.13+
• Ollama installed and running

Installation

pip

pip install olmon

curl (Linux / macOS)

curl -fsSL https://raw.githubusercontent.com/glemiu6/olmon/master/scripts/install.sh | sh

From source

git clone https://github.com/glemiu6/olmon.git
cd olmon
uv pip install -e .

Windows

Native Windows binary is not currently supported.
Use WSL and follow the Linux installation instructions.

Usage

olmon status               # quick health check
olmon models               # list all installed models
olmon models --sort size   # sort by size
olmon models --filter llama  # filter by name or family
olmon inspect llama3:latest  # full details on a model
olmon ps                   # show currently running models
olmon watch                # live auto-refreshing dashboard
olmon watch --interval 5   # refresh every 5 seconds
olmon top                  # htop-style live monitoring
olmon stop qwen2.5:7b      # unload a model from VRAM
olmon compare qwen2.5:7b llama3.2:latest  # compare models
olmon --no-color models    # pipe-friendly output
olmon models --json        # output as JSON

Global flags

olmon --host http://192.168.1.10:11434 status   # connect to remote Ollama
olmon --version                                  # print version

Status Indicators

Configuration

olmon init          # create default config file

Config is stored at ~/.config/olmon/config.json:

{
  "host": "http://localhost:11434",
  "interval": 2,
  "no_color": false,
  "default_sort": "name"
}

Update & Uninstall

olmon update      # update to latest version
olmon uninstall   # remove olmon and config

Why olmon?

Most Ollama monitoring tools are GUI or system tray apps. olmon is built for:

• Headless Linux servers — no GUI required
• Remote monitoring — works over SSH
• Shell scripting — pipe-friendly with --json flag and exit codes
• DevOps workflows — integrate into scripts and cron jobs

GPU Support

• NVIDIA — full VRAM monitoring via nvidia-smi
• AMD — coming soon
• CPU only — VRAM stats not available

Roadmap

See ROADMAP.md for the full plan.

Contributing

Contributions are welcome. Feel free to open an issue or submit a pull request.

License

MIT — see LICENSE for details.

Made with ❤️ by Vlad Digori