Skip to content
Get started

Quickstart

This page takes you from zero to your first task in about 10 minutes, plus model download time.

1. Check the requirements#

You do not need a GPU. If you have one, litecode uses it.

2. Install#

Windows (PowerShell):

irm https://raw.githubusercontent.com/konkomaji/litecode/main/install.ps1 | iex

Linux or macOS:

curl -fsSL https://raw.githubusercontent.com/konkomaji/litecode/main/install.sh | sh

The installer adds the litecode and lite commands to your PATH and starts the setup wizard. If the command is not found afterwards, open a new terminal.

3. Answer the setup wizard#

The wizard asks 4 things:

  1. Where to keep data (cache, memory, skills). Press Enter for the default.
  2. Where to keep models. Pick the disk with the most free space.
  3. Whether to download the engine. Answer yes. If you have an NVIDIA GPU, answer yes to the CUDA build.
  4. Whether to download the recommended models. It shows the size first.

4. Tune for your machine#

litecode tune

This takes about 5 minutes. It finds the fastest thread count and decoding mode for your CPU or GPU.

5. Run your first task#

cd path/to/your/project
litecode

Type a question or a task:

› what is a list comprehension
› explain what @main.py does
› add input validation to the signup function and run the tests
› /review src/

Questions get a direct answer. Tasks run the agent. litecode asks before it changes a file or runs a command: press y (yes), a (always, for this session), n (no), or e (no, and say what to do instead). Tab on an empty prompt switches to auto or plan mode.

6. Next steps#

Source: docs/QUICKSTART.md