chatgpt-cli

README

ChatGPT Command-Line Interface (CLI)

ChatGPT CLI offers an efficient way to communicate with ChatGPT directly from the command line.

Installation

Using brew to install:

brew tap duanemay/tap
brew install chatgpt-cli

Setting Up

You need to set up a ChatGPT API key to use the CLI. Sign up on OpenAI platform to generate one.

It is recommended to create a .chatgpt-cli file in your home directory (or the directory where you'll run the CLI), and put your ChatGPT API key there, together with any other default settings you wish to use.

API_KEY=sk-mysupersecretAPIkey

Flags and Environment Variables

The configuration settings precedence order is:

command line flags > environment variables > configuration files > defaults

Every flag can be set in a config file by converting it to uppercase and replacing any - with _. For example, --api-key can be set as API_KEY. The config file is loaded from the current directory, then from the user's home directory. The configuration file is named .chatgpt-cli.

You can select a different configuration file with the --config flag. Settings in a config file are specified as KEY=VALUE pairs, one per line. Lines starting with # are treated as comments and ignored.

The full list of available flags and corresponding environment variables:

Flag	Short	Config File Key	Default	Description
--api-key	-k	API_KEY	Required	ChatGPT API Key
--config	-c	CONFIG	./.chatgpt-cli then $HOME/.chatgpt-cli	Config file to load
--verbose	-v	VERBOSE	false	Verbose logging
--eom		EOM	s	End of message marker
--eos		EOS	q	End of session marker
--session-file	-s	SESSION_FILE	Generated	Session file
--no-write-session		NO_WRITE_SESSION	false	Do not write or update session file
--model	-m	MODEL	gpt-4	Model to use
--role		ROLE	user	Role of User
--temperature	-m	TEMPERATURE	1.0	Temperature: 0-2
--max-tokens		MAX_TOKENS	0	Max tokens: 8192
--top-p		TOP_P	1.0	Top P: 0-1

For instance, if you want to change the end of message and end of session markers, you can modify them in your configuration file.

Usage

The basic chatgpt-cli command usage:

chatgpt-cli [command]

Available commands include:

• chat: Enter a chat session with ChatGPT
• completion: Generate the autocompletion script for the specified shell
• help: Help about any command
• list-models: Lists all models available to your account
• replay-session: Replay a chat session from a saved file
• version: Displays version information

chat

This command initiates a chat session with ChatGPT. You will be prompted to enter a message. Type in desired texts which could be multi-line to send. When your message is complete, you can either type CTRL+D or \s on a separate line to send. You can change the default end of message marker with the --eom flag.

ChatGPT will then respond with a message. You can continue the chat by inputting another message.

To exit the chat session, you can either type in CTRL+C or \q on a separate line. Any text inputted before exiting would not be sent. The default end of session marker can also be changed using the --eos flag.

The chat session will be saved in a session file. You can resume this session by specifying the file using the --session-file flag.

For instance:

chatgpt-cli chat --session-file session.json

You can also set Model, Role, Temperature, Max Tokens, and Top P with the --model, --role, --temperature, --max-tokens, and --top-p flags respectively. These can be changed when resuming a session with the --session-file flag.

replay-session

This command replays a chat session from the saved file, a great way to revisit a chat session in a more readable format than the raw JSON.

chatgpt-cli replay-session --session-file session.json

list-models

This command lists all models available for your account.

chatgpt-cli list-models

version

This command displays version information.

chatgpt-cli version

Build and Release

For a trial run, do a dry-run of the deployment process. This can be achieved by running a command with few flags that will hinder you from releasing it on GitHub.

goreleaser --snapshot --skip-publish --clean

After adding the GitHub token in ./.github_token, run:

goreleaser --clean

Examples for use

Improving README

echo "Rewrite this README file as a user guide. Make it easy to read and informative. Use a helpful and clear style" \
  | chatgpt-cli chat < README.md > README-new.md

Editing a directory full of notes

for file in notes/*.md; do
  printf "\nEditing '%s'\n============================\n" "$file"
  {
    printf "Revise my notes. "
    printf "Use Markdown format. "
    printf "Revise the text of the note below to use a clear and informative style. "
    printf "Use newlines to keep line length less than 120 characters.\n\n"
  } >"${file}.req"

  chatgpt-cli chat <"${file}.req" >"${file}.new"
  if [ $? -ne 0 ]; then
    printf "  Request Failed: '%s'\n" "${file}"
    rm "${file}.new"
  else
    mv "${file}.new" "${file}"
  fi
  rm "${file}.req"
done