py-fixer 1.0.0

Run Python scripts with AI-powered error explanations

pyfixer

A drop-in replacement for python that runs your scripts normally — but when an error occurs, uses AI to explain what went wrong and suggest a fix, right in your terminal.

Supports Anthropic Claude and Google Gemini. The provider is detected automatically from your API key.

---

✨ Key Features

• AI Error Explanations: Get plain-English explanations of tracebacks powered by Claude or Gemini.
• VS Code Diff View: See suggested fixes as a side-by-side diff before applying anything.
• One-Click Apply: Accept or discard fixes interactively — originals are backed up automatically.
• Multi-Provider: Works with Anthropic (sk-ant-...) and Google Gemini (AIza...) keys.
• No lock-in: Use --no-explain to skip AI entirely and behave like plain python.

---

🚀 Installation

pip install py-fixer

---

🏁 Quick Start

1. Store your API key

pyfixer login

Paste your API key when prompted — you'll also choose a model. Credentials are saved to ~/.config/pyfixer/config.json (mode 600).

2. Run a script

pyfixer run script.py

When an error occurs, pyfixer:

1. Prints the full traceback
2. Explains the error in plain English
3. Opens a VS Code diff showing the suggested fix
4. Asks whether to apply it — original is backed up to .pyfixer-backup/

pyfixer run script.py arg1 arg2     # forward args to your script
pyfixer run script.py --no-explain  # traceback only, no AI

---

📚 Commands

Command	Description
pyfixer login	Save your API key and choose a model
pyfixer logout	Remove your stored API key and model
pyfixer run <script>	Run a script with AI error explanations
pyfixer set-model	Switch models without re-entering your key
pyfixer install-extension	Install the VS Code extension for inline fixes

---

Requirements

• Python 3.10+
• An Anthropic or Google Gemini API key
• VS Code (optional, for diff view)

---

🤝 Contributing

Bug reports and pull requests are welcome! Please open an issue at github.com/ayushisahu222/py-fixer/issues.

---

📄 License

This project is licensed under the MIT License.