cli_lm

Switch context less by using a really cheap LLM at the command line.

Installation

# Install from source
git clone https://github.com/davenpi/cli_lm.git
cd cli_lm
pip install -e .

# Install from PyPI
pip install cli-lm

Quick Start

The first run will prompt for your Open AI API key. Just paste the key in. Or create the key directly in ~/.cli_lm/.env with

echo "OPENAI_API_KEY=your_new_key" > ~/.cli_lm/.env

Note you'll have to pay Open AI to get an API key and start using their models. For now this package sets gpt-4o-mini as the model of choice. It costs 0.15¢/0.60¢ per 1 million input/output tokens. 1 token ~ 1 word.

# Directly ask a question
clm "How do I amend a commit in git?"

# Interactive mode (input in python interpreter)
clm
> How do I list the processes listening on a given port?

Chat is assumed to be multiturn (i.e., you'll always be asked for your next prompt). Just hit Ctrl+c or Ctrl+d to exit.

Configuration

The package looks for your Open AI API key in ~/.cli_lm/.env.

To quickly update your API key run:

echo "OPENAI_API_KEY=your_new_key" > ~/.cli_lm/.env

It works on a Mac, but it hasn't been tested on another system.

Ongoing Improvements

• Multi turn chat capability
• Improved multiline input
• Would be nice to arrow backward on input
• System prompting for really concise, helpful answers
• Make set up easier
• Add token/cost accounting
• Make it faster (imports can be slow)
• Add color

License

MIT