qwenshell 0.2.2

Natural language Linux terminal assistant powered by Ollama

🤖 ShellAI

Natural language Linux terminal assistant powered by Ollama

Turn plain English into safe, confirmed shell commands — entirely local, zero cloud dependency.

$ ai find files larger than 1GB
→ Request: find files larger than 1GB

  ┌──────────────────────────────────────────────┐
  │  find . -type f -size +1G                    │
  └──────────────────────────────────────────────┘

Execute? [y/N] y

./Downloads/ubuntu.iso
./Videos/rawfootage.mp4

---

✨ Features

Feature	Details
🔤 Natural language input	Type what you want, get a shell command
🛡️ Safety filter	Blocks rm -rf /, fork bombs, disk wipes, and more
⚠️ Risk levels	Commands rated safe / medium / high — risky ones flagged automatically
📖 Explain mode	--explain any command in plain English
📜 History log	All requests and outcomes saved locally
🎨 Colored output	Clean, readable terminal UI
⚙️ Configurable	Switch models, URL, and behaviour with ai --set
📦 Zero dependencies	Pure Python stdlib — no requests, no click
🔒 100% local	Nothing leaves your machine

---

📋 Requirements

• Linux (tested on Ubuntu 22.04+, Arch, Fedora)
• Python ≥ 3.10
• Ollama running locally

---

🚀 Installation

Option 1 — pip (recommended)

pip install shellai

Option 2 — development install (editable)

git clone https://github.com/yourusername/shellai
cd shellai
pip install -e ".[dev]"

Pull a model

ollama pull qwen2.5:7b       # default — fast, good quality
# or
ollama pull deepseek-coder   # great for system/dev tasks

Start Ollama

ollama serve

---

🎮 Usage

Translate natural language → shell command

ai find duplicate files
ai compress this folder
ai show running docker containers
ai install nginx
ai show disk usage by folder
ai monitor CPU and memory
ai list all open ports

Explain a command

ai --explain tar -czvf backup.tar.gz ./myfolder
ai --explain find . -name "*.log" -mtime +7 -delete

Manage history

ai --history          # show last 20 commands
ai --clear-history    # wipe history

Model management

ai --models                         # list available Ollama models
ai --model deepseek-coder "..."     # use a different model for one request
ai --set model deepseek-coder:6.7b  # change default model permanently

Configuration

ai --config                         # show all settings
ai --set model qwen2.5:7b
ai --set ollama_url http://192.168.1.5:11434
ai --set timeout 180
ai --set stream_explain true

---

🛡️ Safety System

Commands are classified into four risk levels:

Level	Examples	Behaviour
✅ safe	ls, df, ps, grep	Confirm → execute
🟡 medium	rm file.txt, systemctl stop	Warn + explain + confirm
🟠 high	rm -rf ~/dir, curl … | bash	Warn prominently + explain + confirm
🔴 critical	rm -rf /, fork bomb, mkfs, dd wipe	Blocked — never executed

Patterns detected include:

• Recursive root deletion (rm -rf /)
• Disk format (mkfs)
• Disk wipe via dd
• Fork bomb :(){ :|:& };:
• Base64 obfuscated eval tricks
• Pipe-to-shell downloads (curl … | bash)
• System shutdown/reboot

---

⚙️ Configuration Reference

Config is stored in ~/.config/shellai/config.json.

Key	Default	Description
ollama_url	http://localhost:11434	Ollama API base URL
model	qwen2.5:7b	Default Ollama model
timeout	120	Seconds to wait for Ollama response
history_limit	500	Max entries kept in history
stream_explain	true	Stream tokens when explaining
confirm_safe	true	Ask confirmation for safe commands
max_retries	2	LLM retry attempts on bad output

---

📁 Project Structure

shellai/
├── shellai/
│   ├── __init__.py        # version
│   ├── cli.py             # entry point & sub-commands
│   ├── ollama_client.py   # Ollama REST API client
│   ├── safety.py          # command safety filter
│   ├── executor.py        # subprocess runner
│   ├── history.py         # JSONL history log
│   ├── display.py         # colours & terminal UI
│   ├── prompts.py         # LLM prompt templates
│   └── config.py          # config load/save
├── tests/
│   ├── test_safety.py
│   └── test_cli.py
├── pyproject.toml
├── setup.py
├── LICENSE
└── README.md

---

🔬 Development

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest tests/ -v

# Lint
ruff check shellai/

# Type-check
mypy shellai/

# Build for PyPI
python -m build

---

📦 Publishing to PyPI

# Install build tools
pip install build twine

# Build
python -m build

# Upload to TestPyPI first
twine upload --repository testpypi dist/*

# Upload to PyPI
twine upload dist/*

---

🗺️ Roadmap

• Shell completion (bash/zsh/fish)
• Plugin system for custom safety rules
• --dry-run flag (generate but never execute)
• --pipe mode: echo "find large files" | ai
• Multi-step task chaining
• Context awareness (detect OS, installed tools)
• Interactive REPL mode (ai shell)
• Support for remote Ollama instances

---

📄 License

MIT — see LICENSE

---

🙏 Credits

Built on Ollama — local LLM inference made easy.