omon

Ollama tells you names. omon tells you what they mean.

GitHub · PyPI · Homebrew tap · Issues

Ollama's CLI shows model tags and file sizes. omon decodes those tags, benchmarks performance on your hardware, tracks what's eating your RAM, and tells you what to run — or remove.

Why omon?

• Decode cryptic model names — 35b-a3b-coding-nvfp4 becomes "Qwen3.5 · 35B (3B active) · Code generation · NVIDIA FP4 · vision, thinking, tools"
• Benchmark on your machine — cold/warm load times, tok/s, memory footprint; compare models side-by-side
• Know what to run — hardware-aware suggestions, successor alerts, cleanup recommendations for stale models
• Local-first, zero dependencies — Python stdlib only. No venv conflicts, no supply chain risk, no cloud calls

Screenshots

Click an image for full size.

Status overview (omon)

Live TUI (omon watch)

Web dashboard (omon serve)

Install

Pick one — you only need a single install path.

PyPI (macOS, Linux; recommended):

pipx install omon
# or: pip install omon

Homebrew (macOS):

brew tap LightbridgeLab/omon
brew install omon

From source:

pipx install git+https://github.com/LightbridgeLab/OllamaMon.git

Requires Python 3.10+ and a running Ollama instance.

Upgrade: pipx upgrade omon or brew upgrade omon

Quick start

omon                          # status overview: server, models, RAM, pressure
omon list                     # decoded model inventory
omon bench llama3.2:3b        # benchmark load time and tok/s
omon watch                    # live TUI (press q to quit)

More commands: omon hw, omon suggest --task coding, omon cleanup, omon serve.

Developing from source

git clone https://github.com/LightbridgeLab/OllamaMon.git && cd OllamaMon
make install
make dev              # common local test commands
make check            # run tests

See AGENTS.md for architecture rules and the pipx/Homebrew vs .venv note.

Developing from source

git clone https://github.com/LightbridgeLab/OllamaMon.git && cd OllamaMon
make install
make dev              # common local test commands
make check            # run tests

See AGENTS.md for architecture rules and the pipx/Homebrew vs .venv note.

Documentation

• Command reference — full docs for every command, config, and completions
• Design history — how and why omon was built
• Roadmap — what's planned next
• Security — network exposure and data storage
• Contributing — architecture rules and conventions

Requirements

• Python 3.10+
• Ollama running locally (or --host for remote)
• macOS or Linux (Apple Silicon is the primary target)

License

MIT — see LICENSE.