Kimari CLI

Command-line interface for Kimari Local AI
LLM inference gateway for consumer GPUs

---

Install

npm install -g @kimari-ai/cli

Then set up the Python backend:

git clone https://github.com/smouj/kimari-local-ai.git ~/.kimari/kimari-local-ai
cd ~/.kimari/kimari-local-ai
pip install -e .

Quick Start

kimari doctor              # System diagnostics
kimari server start        # Start inference server
kimari models download --model Qwen3-4B  # Download a model
kimari chat --profile default           # Start chatting
kimari dashboard           # Open web UI

Commands

Command	Description
kimari doctor	Full system diagnostics
kimari server start	Start llama.cpp inference server
kimari server stop	Stop server gracefully
kimari server status	Server health and uptime
kimari models list	List downloaded GGUF models
kimari models download	Download a model from HuggingFace
kimari chat	Interactive chat or single prompt
kimari benchmark	Performance benchmarks
kimari system	Live resource monitor
kimari profiles	Manage GPU profiles
kimari config	View/edit configuration
kimari dashboard	Open web dashboard

How It Works

npm (kimari) → Node.js wrapper → Python CLI → llama.cpp → GPU

The CLI is a lightweight Node.js wrapper delegating to the Python backend. npm handles distribution and updates; Python handles GPU inference and model management.

Requirements

• Linux (Ubuntu 20.04+, Debian 11+)
• NVIDIA GPU with CUDA 12+
• Python 3.10+ · Node.js 18+

Related

• Kimari Local AI — Main repo
• Kimari Dashboard — Web UI

License

Apache-2.0 © Kimari AI