🤖 AI Shell Agent

An AI-powered command-line assistant that converts natural language into Windows PowerShell commands using Google Gemini and LangChain. Before executing any command, the agent validates it and asks for user confirmation to ensure safe execution.

---

✨ Features

• 🤖 Natural language to PowerShell command generation
• 🧠 Powered by Google Gemini using LangChain
• 🛡️ Command validation before execution
• ✅ User confirmation before running commands
• ⚡ Automatic PowerShell command execution
• 💻 Interactive command-line interface
• 🪟 Designed for Windows PowerShell

---

🏗️ Project Architecture

User Input
     │
     ▼
 AI Shell Agent
     │
     ▼
 LangChain + Gemini
     │
     ▼
 PowerShell Command
     │
     ▼
 Command Validator
     │
     ▼
 User Confirmation
     │
     ▼
 Command Executor
     │
     ▼
 Windows PowerShell

---

📂 Project Structure

ai-shell-agent/
│
├── app/
│   ├── __init__.py
│   ├── main.py
│   ├── agent.py
│   ├── llm.py
│   ├── validator.py
│   ├── executor.py
│   └── prompts.py
│
├── README.md
├── LICENSE
├── pyproject.toml
├── .env.example
└── .gitignore

---

⚙️ Tech Stack

• Python 3.11+
• LangChain
• Google Gemini API
• python-dotenv
• subprocess (Python Standard Library)
• PowerShell

---

🚀 Installation

Clone the repository:

git clone https://github.com/Nishtha-06/ai-shell-agent
cd ai-shell-agent

Install dependencies using uv:

uv sync

or using pip:

pip install -e .

---

🔑 Environment Variables

Create a .env file in the project root.

GEMINI_API_KEY=your_gemini_api_key

You can get an API key from:

https://aistudio.google.com/app/apikey

---

▶️ Running the Agent

Run the application:

ai-shell

or

python -m app.main

---

💬 Example

You > Create a folder named Project

Generated Command:
mkdir Project

Execute this command? (y/n): y

Command executed successfully.

---

🔄 Workflow

Natural Language
        │
        ▼
Gemini (LangChain)
        │
        ▼
PowerShell Command
        │
        ▼
Validator
        │
        ▼
Confirmation
        │
        ▼
Executor
        │
        ▼
PowerShell

---

🛡️ Safety

Before executing any generated command, the agent:

• Validates the generated PowerShell command
• Blocks potentially unsafe commands (based on validator rules)
• Requests user confirmation before execution

This helps prevent accidental execution of harmful commands.

---

📌 Current Limitations

• Supports Windows PowerShell only
• Requires a Google Gemini API key
• Executes one command at a time
• Does not maintain conversation memory between commands

---

🚀 Future Improvements

• Conversation memory
• LangGraph integration
• Tool calling
• Command history
• Plugin support
• Persistent shell session
• Linux and macOS support
• Rich terminal interface

---

👩‍💻 Author

Nishtha Modi

---

📄 License

This project is licensed under the MIT License.