gpterm

module

v0.0.0-...-2f29de6 Latest Latest Go to latest Published: Nov 12, 2024 License: MIT

Details

gpterm is a terminal-based integration with the OpenAI chat API.

gpterm also renders Markdown and syntax-aware code blocks:

Requirements:

To install from source:

git clone https://github.com/collinvandyck/gpterm.git
cd gpterm
make install

Because it uses the OpenAI API, an API key is required before you can start:

# set api key
gpterm auth

Once the API key has been set, have fun!

# enter an interactive session
gpterm

Once you're in gpterm, there are a handful of controls you can use to tailor your experience.

Ctrl-y spawn an editor to craft your message instead of using the text widget.
Ctrl-p/Ctrl-n switch between previous and next conversations.
Ctrl-x drops the current conversation. Ctrl-x again to confirm.
F1/F2 change the amount of conversation context sent to OpenAI on each request. Higher values will result in more coherence but at a greater API cost.
F3 change the GPT model. Currently supports gpt-3.5-turbo and gpt-4. You must have gpt-4 access for that mode to work. Because costs between the models are quite different, gpterm remembers the amount of conversation context to send per-model.

Chat history and your API key are stored in a sqlite database in:

~/.config/gpterm

Users will be able to instruct OpenAI how to act. Currently it's hardcoded to You are a helpful assistant.

Additionally, users should be able to change the name of the assistant as rendered in the chatlog. Current it is ChatGPT.

Conversations should be able to be uploaded to a gist.

Actions should be able to be found to user-configured keys.