kubrick-cli 0.1.2

A CLI tool for AI-assisted coding using Triton LLM backend

Kubrick CLI

An AI-assisted coding CLI tool powered by your own Triton LLM backend. Like Claude Code, but self-hosted.

Installation

Local Installation

pip install -e .
kubrick

Docker

cd /path/to/your/project
docker run --rm -it \
  --network host \
  -v ${HOME}:/home/kubrick \
  -v ${PWD}:/workspace \
  -v /etc/localtime:/etc/localtime:ro \
  -v /etc/timezone:/etc/timezone:ro \
  rcland12/kubrick-cli

Quick Start

# Basic usage
kubrick

# Custom Triton server
kubrick --triton-url my-server:8000

# Load previous conversation
kubrick --load 20240118_143022

Requirements

• Python 3.8+
• Triton Inference Server with streaming LLM model
• Default endpoint: http://localhost:8000
• Default model: llm_decoupled

Documentation

• WIKI.md - Complete features, commands, and usage guide
• TRITON.md - Triton backend setup and requirements
• DOCKER.md - Docker setup and troubleshooting

Configuration

Config stored at ~/.kubrick/config.json. Override with environment variables:

export TRITON_URL=localhost:8000
export TRITON_MODEL_NAME=llm_decoupled

License

MIT License - See LICENSE