If you love living inside the terminal but still want the reach of GPT models at your fingertips, let me introduce the latest release of Tux-GPT. It’s a command-line assistant (not a persistent background agent) that gives you conversational help, code guidance, and shell-ready automations without breaking your flow.
If you ask me, I’d say this is like a lite version of the Codex client.
Why Tux-GPT?
- Terminal-native experience: prompts, responses, and formatting stay inside your shell (rendered with
rich). - Context-aware conversations: history and preferences live in
~/.config/tux-gpt/, so the CLI remembers your last 20 exchanges. - System-aware commands: at startup, Tux-GPT inspects the host OS (distro, architecture, version) so suggestions match yo…
If you love living inside the terminal but still want the reach of GPT models at your fingertips, let me introduce the latest release of Tux-GPT. It’s a command-line assistant (not a persistent background agent) that gives you conversational help, code guidance, and shell-ready automations without breaking your flow.
If you ask me, I’d say this is like a lite version of the Codex client.
Why Tux-GPT?
- Terminal-native experience: prompts, responses, and formatting stay inside your shell (rendered with
rich). - Context-aware conversations: history and preferences live in
~/.config/tux-gpt/, so the CLI remembers your last 20 exchanges. - System-aware commands: at startup, Tux-GPT inspects the host OS (distro, architecture, version) so suggestions match your environment.
- Flexible interaction modes: free-form chat, single-shot questions, machine-readable JSON, or even “please craft & run this shell command”.
Installation
Most platforms:
pip install tux-gpt
Recent Debian/Ubuntu releases route pip to the system Python, so use pipx to keep your global environment clean:
pipx install tux-gpt
From source:
git clone https://github.com/fberbert/tux-gpt.git
cd tux-gpt
pip install -r requirements.txt
pip install .
Just set your OpenAI key (OPENAI_API_KEY), and you’re all set to go! For more detailed guidance, feel free to visit our GitHub repository.
Getting Started
Interactive session (multi-line input supported):
tux-gpt
- Press Ctrl+j to send (Enter adds a new line).
- Type Ctrl+c or Ctrl+d to quit.
- Conversation history resumes automatically next time.
Single-shot question:
tux-gpt -q "Summarize the latest Kubernetes news"
JSON-ready response (great for scripts):
tux-gpt --json -q "What are three underrated Linux CLI tools?"
You’ll receive:
{
"answer": "...",
"sources": [...]
}
Command crafting & execution:
tux-gpt -c "create a zip with every png and jpg in the current folder"
Tux-GPT will automatically execute the command in your terminal. If the command seems risky or needs root privileges, it’ll ask for your confirmation before proceeding.
Configuration & Memory
On first run, Tux-GPT sets up ~/.config/tux-gpt/ (%APPDATA%\tux-gpt\ on Windows):
- config.json # default model, spinner settings, etc.
- history.json # last 20 messages (user + assistant)
- input_history # command history for ↑/↓ navigation
Need a clean slate? Delete the files in that directory and Tux-GPT will recreate them.
Under the Hood
- Powered by the latest openai responses API with optional web-search tooling.
- Uses rich for Markdown output and prompt-toolkit for multiline input.
- Defaults to the GPT model defined in config.json (ships with gpt-4.1-mini). Change it whenever you want.
Roadmap & Contributions
This project is still young; new capabilities to come. Ideas, bug reports, or pull requests are always welcome. Take it for a spin, file issues, or fork it to roll your own terminal assistant.
GitHub: https://github.com/fberbert/tux-gpt PyPI: https://pypi.org/project/tux-gpt/
Stay in your shell, keep your flow, and let Tux-GPT do the heavy lifting. Happy hacking!