pipe-pai

pipe-pai is a simple command-line tool that allows you to interact with OpenAI models through Unix pipes. Use it like this: ... | pai - hence the name "pipe-pie".

Installation

Install via pipx (recommended):

pipx install pipe-pai-tool

Or with pip:

pip install pipe-pai-tool

Or locally:

git clone git@github.com:krzysztofarendt/pai.git
cd pai
pipx install .

Ensure your OpenAI API key is set in the environment:

export OPENAI_API_KEY="your_api_key_here"

Optionally, set a default model using the PAI_MODEL environment variable:

export PAI_MODEL="gpt-4"

Usage

Pipe a prompt to pai and receive the model's response on stdout:

echo "Write a haiku about the sea" | pai

Alternatively, specify a different model or API key:

echo "Tell me a joke" | pai --model gpt-4 --api-key $OPENAI_API_KEY

Chat mode

By default, pai runs in the pipe mode with no chat history. You can enter an interactive multi-turn chat by:

pai --chat
> Write a poem about the moon

--------------------
<AI response displayed here>

To exit chat mode, send an empty prompt press ctrl+c or type q, exit, or quit.