English | 中文

CommandChat

commandchat provides the occ CLI for chatting with OpenAI or Azure OpenAI models from your terminal.

Features

• Interactive CLI chat with occ chat
• Profile-based configuration for multiple OpenAI / Azure OpenAI accounts
• Prompt template management with built-in and custom prompts
• Prompt selection per command or as a profile default
• Experimental image generation command

Requirements

• Python 3.8+
• A terminal with pip
• An OpenAI-compatible API key or Azure OpenAI configuration

Install

pip3 install commandchat

If you manage multiple Python versions locally:

python3 -m pip install commandchat

Quick start

1. Configure your first profile

occ configure

This opens the interactive configuration flow and writes local settings under ~/.occ/.

2. Start chatting

occ chat "Hello, introduce yourself briefly."

You can also start interactive multiline input:

occ chat

Inside interactive chat, you can use slash commands:

• / to open the shortcut command menu
• /pmp, /p, or /prompt to list prompts and switch the active prompt template
• /m or /model to list models and switch the active model
• /q to exit the chat session
• /help to show the help panel again

When you switch prompts with /pmp or models with /m, the CLI will ask whether you want to keep the current context or start a fresh chat session with a new session id.

Interactive chat also shows a bottom status bar with the current prompt, model, and session id.

3. Use prompt templates

occ prompt list
occ chat -pt translate "你好世界"
occ chat -pt improve "I wants to go to school yesterday"

4. Work with multiple profiles

occ configure profile -p work
occ configure list
occ chat -p work "Summarize today's priorities"

Common commands

Files created locally

• ~/.occ/config: profile and model configuration
• ~/.occ/prompts.json: custom prompt templates

Packaging and PyPI

This project now uses pyproject.toml as the single source of truth for packaging metadata.

• README.md is configured as project.readme, so PyPI renders this file on the project page.
• README-zh.md is kept for the GitHub repository and Chinese-language documentation.
• To update the PyPI project page content, update README.md, bump the version in pyproject.toml, and publish a new release.

Automatic publishing with GitHub Actions

The repository includes .github/workflows/publish-to-pypi.yml.

On pushes to master or manual runs, it will:

1. Check whether the version in pyproject.toml changed
2. Build the sdist and wheel from pyproject.toml
3. Run twine check dist/*
4. Publish to PyPI via Trusted Publishing (OIDC)

Because the package metadata points to README.md, each new PyPI release also refreshes the PyPI homepage description.

Manual build

python3 -m pip install build twine
python3 -m build
python3 -m twine check dist/*

Uninstall

pip3 uninstall commandchat

Feedback

• xxx.tao.c@gmail.com
• xoto@outlook.be

License

MIT