PAIR is an AI-powered coding assistance REPL that pairs GPT-4 with you the developer to augment the best of both human and AI intelligence. It provides an interactive environment where users can input existing code, ask questions about the code or other open source projects or dependencies, receive helpful answers from the GPT-based programming assistant, add new code or refactor existing code, etc.
The REPL supports special commands for loading files and changing directories, and it can propose code changes as context diffs that can be processed automatically. Users have the option to accept or reject the proposed changes, making PAIR a flexible and powerful tool for developers.
Currently we are using GPT-4 to help build PAIR, and are very open to other collaborators who are similarly inclined to eat their own dogfood.
To install Pair AI, run the following command:
pip install pair_ai
After installing the package, you can use the pair
command in your terminal or command prompt to start the REPL:
pair [file1] [...]
One or more filenames can be specified on the command line to load into the model context. This tends to be more convenient than using the /file command in the repl loop.
In the REPL, enter your questions or guidance or /file to input local files into the context.
help
- Display this help message/file <path>
: Load a file's content into the model context by providing its path./cd <path>
: Change the current working directory to the specified path./url <url>
: Load the content of a URL into the context./status
: - Show the status of the OPENAI_API_KEY and the model being used.
To use the special commands, simply type the command followed by the appropriate path or command in the REPL.
Example:
/file /path/to/your/file.py
/cd /path/to/your/directory
- chatstack
- prompt_toolkit
The following environment variables are used to inject credentials and other required configuration for the major dependencies:
OpenAI
- OPENAI_API_KEY # your OpenAI API key
- PAIR_MODEL # one of "gpt-3.5-turbo" or "gpt-4", default to gpt-4